Leopold Aschenbrenner - China/US Super Intelligence Race, 2027 AGI, & The Return of History

Speaker 1 Okay, today I'm chatting with my friend Leopold Oschenbrenner.

Speaker 1 He grew up in Germany, graduated valedictorian of Columbia when he was 19, and then he had a very interesting gap year, which we'll talk about.

Speaker 1 And then he was on the OpenAI Super Alignment team, mate, rest in peace.

Speaker 1 And now he, with some anchor investments from Patrick and John Collison and Daniel Gross and Nat Friedman, is launching an investment firm.

Speaker 1 So Leopold, I know you're off to a slow start, but life is long and I wouldn't worry about it too much. You'll make up for it in due time.

Speaker 1 But thanks for coming on the podcast.

Speaker 2 Thank you. You know,

Speaker 2 I first discovered your podcast when your best episode had, you know, like a couple hundred views.

Speaker 2 And so it's just been, it's been amazing to follow your trajectory and it's a delight to be on.

Speaker 1 Yeah, yeah. Well, I think

Speaker 1 in the Shota and Trenton episode, I mentioned that a lot of the things I've learned about AI, I've learned from talking with them.

Speaker 1 And the third part of this triumvirate, probably the most significant in terms of the things that I've learned about AI, has been you. We'll get a lot of stuff on the record now.
Great.

Speaker 1 Okay, first thing not to get on record. Tell me about the trillion-dollar cluster.

Speaker 1 By the way, I should mention, so the context of this podcast is today

Speaker 1 you're releasing a series called Situational Awareness. We're going to get into it.
First question about that is, tell me about the trillion dollar cluster.

Speaker 2 Yeah. So,

Speaker 2 you know, unlike basically most things that have come out of Silicon Valley recently, you know, AI is kind of this industrial process.

Speaker 2 You know, the next model doesn't just require some code. It's building a giant new cluster.
Now it's building giant new power plants. Pretty soon it's going to be building giant new fabs.

Speaker 2 And since ChatGPT, this kind of extraordinary sort of techno capital acceleration has been set into motion. I mean, basically, exactly a year ago today,

Speaker 2 NVIDIA had their first kind of blockbuster earnings call, where it went up 25% after hours and everyone was like, oh my God, AI, it's a thing.

Speaker 2 I think within a year,

Speaker 2 NVIDIA, NVIDIA data center revenue has gone from like, you know, a few billion a quarter to like, you know, 20, 25 billion a quarter now and, you know, continuing to go up.

Speaker 2 Like, you know, big tech capex is skyrocketing. And,

Speaker 2 you know, it's funny because it's both, there's this sort of, this kind of crazy scramble going on, but in some sense, it's just the sort of continuation of straight lines on a graph, right?

Speaker 2 There's this kind of like long-run trend, basically almost a decade of sort of training compute of the sort of largest AI systems growing by about, you know, half an order of magnitude, you know, 0.5 booms a year.

Speaker 2 And you can just kind of play that forward, right? So, you know, GPT-4, you know, rumored or reported to have finished pre-training in 2022.

Speaker 2 You know, the sort of cluster size there was rumored to be about 25,800s, you know, sorry, A100s on semi-analysis.

Speaker 2 You know, that's roughly, you know, if you do the math on that, it's maybe like a $500 million cluster. You know, it's very roughly 10 megawatts.

Speaker 2 And, you know, just play that forward, half an oom a year, right? So then 2024, that's say, you know, that's a cluster that's 100 megawatts. That's like 100,8100 equivalents.
You know, that's,

Speaker 2 you know, costs in the billions. You know, play it forward, you know, two more years, 2026, that's a cluster that's a gigawatt.
You know, that's

Speaker 2 sort of a large nuclear reactor size. It's like the power of the Hoover Dam.
You know, that costs tens of billions of dollars. That's like a million H100 equivalents.

Speaker 2 You know, 2028, that's a cluster that's 10 gigawatts, right? That's more power than kind of like most US states.

Speaker 2 That's like 10 million H100s equivalents. It costs hundreds of billions of dollars.
And then 2030,

Speaker 2 trillion-dollar cluster, 100 gigawatts, over 20% of U.S. electricity production, you know, 100 million H100 equivalents.
And that's just the training cluster, right?

Speaker 2 That's like the one largest training cluster. And then there's more inference GPUs as well, right? Most of, you know, once there's products, most of them are going to be inference GPUs.
And so,

Speaker 2 you know, U.S. power production has barely grown for like, you know, decades.
And now we're really in for a ride.

Speaker 1 So, I mean, when I had Zuck on the podcast, he was claiming not a plateau per se, but that AI progress would be bottlenecked by specifically this constraint on energy and specifically like oh gigawatt data centers are going to build another three gorgeous dam or something i know that there's companies according to public reports who are planning things on the scale of a gigawatt data center 10 gigawatt data center who's going to be able to build that i mean a hundred gigawatt center like a state where you're getting are you going to pump that into one physical data center how is it going to be possible yeah what is zuck missing i mean you know i don't know i think 10 gigawatt you know like six months ago, you know, 10 gigawatts was the talk of the town.

Speaker 2 I mean, I think, I feel like now, you know, people have moved on. You know, 10 gigawatts is happening.

Speaker 2 I mean, I don't know, there's the information report on OpenAI and Microsoft planning a $100 billion cluster. So you got to, you know, if you.

Speaker 1 Is that a gigawatt or is that the 10 gigawatt?

Speaker 2 I mean, I don't know. But, you know, if you try to like map out, you know, how expensive would the 10 gigawatt cluster be, you know, that's maybe a couple hundred billion.

Speaker 2 So it's sort of on that scale.

Speaker 2 And they're planning it. They're working on it.
You know, so, so

Speaker 2 the,

Speaker 2 you know, it's not just sort of my crazy take. I mean, mean, AMD, AMD, I think, forecasted a $400 billion AI accelerator market by 27.

Speaker 2 You know, I think, I think it's, you know, and AI accelerators are only part of the expenditures.

Speaker 2 It's sort of, you know, I think sort of a trillion dollars of sort of like total AI investment by 2027 is sort of like, we're very much in track on it.

Speaker 2 I think the trillion-dollar cluster is going to take a bit more sort of acceleration. But, you know, we saw how much sort of chat GPT unleashed, right?

Speaker 2 And so like every generation, you know, the models are going to be kind of crazy and people, it's going to shift the overton window.

Speaker 2 And then, you know, obviously the revenue comes in, right? So these are forward-looking investments. And the question is, do they pay off?

Speaker 2 And so if we sort of estimated the, you know, the GPT-4 cluster at around 500 million,

Speaker 2 by the way, that's sort of a common mistake people make is they say, you know, people say like $100 million for GPT4, but that's just the rental price.

Speaker 2 They're like, ah, you rent the cluster for three months. But if you're building the biggest cluster,

Speaker 2 you got to build the whole cluster. You got to pay for the whole cluster.
You can't just rent it for three months.

Speaker 2 But I mean, really, once you're trying to get into the sort of hundreds of billions, eventually you got to get to like 100 billion a year right.

Speaker 2 I think this is where it gets really interesting for the big tech companies, right? Because their revenues are on order, you know, hundreds of billions, right?

Speaker 2 So it's like 10 billion, fine, you know, and it'll pay off the 2024 size training cluster.

Speaker 2 But you know, really, when sort of big tech, it'll be gangbusters is 100 billion a year. And so the question is sort of how feasible is 100 billion a year from AI revenue.
And

Speaker 2 it's a lot more than right now, but I think

Speaker 2 if you sort of believe in the trajectory of the AI systems as I do, and which we'll probably talk about, it's not that crazy, right?

Speaker 2 So there's, I think there's like 300 million-ish Microsoft Office subscribers, right?

Speaker 2 And so they have Copilot now, and I know what they're selling it for, but suppose you sold some sort of AI add-on for $100 a month

Speaker 2 and you sold that to a third of Microsoft Office subscribers subscribe to that. That'd be $100 billion right there.

Speaker 2 $100 a month is

Speaker 2 a lot. It's a lot.
It's a lot.

Speaker 1 For a third of Office subscribers?

Speaker 2 Yeah, but it's for the average knowledge worker, it's like a few hours of productivity a month.

Speaker 2 And it's kind of like you have to be expecting pretty lame AI progress to not hit some few hours of productivity a month

Speaker 2 of, yeah.

Speaker 1 Okay, sure. So let's assume all this.

Speaker 1 What happens in the next few years in terms of

Speaker 1 what is the one gigawatt training,

Speaker 1 the AI that's trained on the one gigawatt data center? What can it do, the one on the 10 gigawatt data center? Just map out the next few years of AI progress for me.

Speaker 2 Yeah, I think probably the sort of 10 gigawatt-ish range is sort of my best guess for when we get the sort of true AGI. I mean, yeah, I think it's sort of like one gigawatt data center.

Speaker 2 And again, I think actually compute is overrated, and we're going to talk about that, but we will talk about compute right now.

Speaker 2 So, you know, I think sort of 25, 26, we're going to get models that are, you know, basically smarter than most college graduates.

Speaker 2 I think sort of the practice, a lot of the economic usefulness, I think, really depends on sort of unhobbling. Basically, it's the models are kind of,

Speaker 2 they're smart, but they're limited, right? There's this chatbot, and things like being able to use a computer, things like being able to do kind of like gender, long horizon tasks.

Speaker 2 And then I think by 27, 28, if you extrapolate the trends, and we'll talk about that more later, and I talk about it in the series, I think we hit basically

Speaker 2 as smart as the smartest experts. I think the unhobbling trajectory kind of points to,

Speaker 2 it looks much more like an agent than a chat bot and much more almost like basically like a drop-in remote worker, right?

Speaker 2 So it's not like, I think basically, I mean, I think this is the sort of question on the economic returns.

Speaker 2 I think a lot of the intermediate AI systems could be really useful, but you know, it actually just takes a lot of schlep to integrate them, right?

Speaker 2 Like GVT-4, you know, whatever, 4.5, you know, probably there's a lot you could do with them in a business use case, but you know, you really got to change your workflows to make them useful.

Speaker 2 And it's just like, there's a lot of, you know, it's a very Tyler-Cowanist take. It just takes a long time to diffuse.
Yeah. It's like, you know, we're an SF and so we missed that or whatever.

Speaker 2 But I think in some sense,

Speaker 2 you know, the way a lot of these systems want to be integrated is you kind of get this sort of sonic boom where it's, you know, the sort of intermediate systems could have done it, but it would have taken schlap.

Speaker 2 And before you do the schlap to integrate them, you get much more powerful systems, much more powerful systems that are sort of on Hobbled.

Speaker 2 And so they're this agent and they're this drop-in remote worker. And, you know, then you're kind of interacting with them like a coworker, right?

Speaker 2 You know, you can take these Zoom calls with them and you're slacking them and you're like, ah, can you do this project?

Speaker 2 And then they go off and they go away for a week and write a first draft and get feedback on them and run tests on their code.

Speaker 2 And then they come back and you see it and you tell them a little bit more things.

Speaker 2 And that'll be much easier to integrate.

Speaker 2 And so

Speaker 2 it might be that actually you need a bit of overkill to make the sort of transition easy and to really harvest the gains. What do you mean by the overkill?

Speaker 1 Overkill on the model capabilities. Yeah, yeah.

Speaker 2 So basically the intermediate models could do it, but it would take a lot of schlep.

Speaker 2 And so then, you know, the like, actually, it's just the drop-in remote worker kind of AGI that can automate cognitive tasks that actually just ends up kind of like, you know, basically it's you're like, you know, the intermediate models would have made the software engineer more productive, but will the software engineer adopt it?

Speaker 2 And then the 27 model is, well, you know, you just don't need the software engineer. You can literally interact with it like a software engineer and it'll do the work of a software engineer.

Speaker 1 So the last episode I did was with John Shulman. Yeah.
And

Speaker 1 I was asking about basically this.

Speaker 1 And one of the questions I asked is, We have these models that have been coming out in the last year, and none of them seem to have significantly surpassed you before.

Speaker 1 And certainly not in the agentic way in which they are interacting with as a coworker. You know, they'll brag that they got a few extra points on MMLU or something.

Speaker 1 And even GPT-4.0, it's cool that they can talk like Scarlett Johansson or something, but like,

Speaker 1 and honestly, I'm going to use that.

Speaker 1 Not anymore. Not anymore.

Speaker 1 Okay, but the whole coworker thing. So

Speaker 1 this is going to be a run on question, but you can address it in any order. But the, it makes sense to me why they'd be good at answering questions.

Speaker 1 They have a bunch of data about how to complete Wikipedia text or whatever. Where is the equivalent training data that enables it to understand

Speaker 1 what's going on in the Zoom call? How does this connect with what they were talking about in the Slack? What is the cohesive project that they're going after based on all this context that I have?

Speaker 1 Where is that training data coming from?

Speaker 2 Yeah. So I think a really key question for sort of...
AI progress in the next few years is sort of how hard is it to do sort of unlock the test time compute overhang?

Speaker 2 So, you know, right now, GPT-4 answers a question and you know, it kind of can do a few hundred tokens of kind of chain of thought. And that's already a huge improvement, right?

Speaker 2 Sort of like, this is a big on hobbling. Before, you know, answer a math question, it's just shotgun.
And,

Speaker 2 you know, if you tried to kind of like answer a math question by saying the first thing that came to mind, you know, you wouldn't be very good. So, you know, GPT-4 thinks for a few hundred tokens.

Speaker 2 And, you know, if I thought for a few hundred, you know, if I think at like 100 tokens a minute and I thought for a few minutes.

Speaker 1 You think at much more than 100 tokens a minute,

Speaker 1 I don't know.

Speaker 2 If I thought for like 100 tokens a minute, you know, it's like what GPT-4 does, maybe it's like, you know, it's equivalent to me thinking for three minutes or whatever, right?

Speaker 2 You know, suppose GPT-4 could think for millions of tokens, right? That's sort of plus four oomes, plus four orders of magnitude on test time compute, just like on one problem.

Speaker 2 It can't do it right now. It kind of gets stuck, right? It like writes some code, even if, you know, it can do a little bit of iterative debugging, but eventually it just kind of like

Speaker 2 gets stuck in some thing. It can't correct its errors and so on.
And,

Speaker 2 you know, in a sense, there's this big overhang, right? And like other areas of ML, you know, there's this great paper on AlphaGo, right?

Speaker 2 Where you can trade off train time and test time compute and if you can use you know four ooms more test time compute that's almost like you know a three and a half oom bigger model just because again like you can you know if a hundred tokens a minute a few million tokens that's that's a few months of sort of working time there's a lot more you can do in a few months of working time than and then right now so the question is how hard is it to unlock that and i think the

Speaker 2 you know the sort of short timelines ai world is if it's not that hard.

Speaker 1 And the reason it might not be that hard is that

Speaker 2 there's only really a few extra tokens you need to learn, right?

Speaker 2 You need need to kind of learn the error correction tokens the tokens where you're like ah I think I made a mistake let me think about that again you need to learn the kind of planning tokens that's kind of like I'm gonna start by making a plan here's my plan of attack and then I'm gonna write a draft and I'm gonna like now I'm gonna critique my draft I'm gonna think about it and so it's not it's not things that models can do right now but you know the question is how hard is that um and in some sense also you know there's sort of two paths to agents right you know um when sholto was on your podcast you know he talked about kind of scaling leading to more nines of reliability um

Speaker 2 and so that's one path i think the other path is this sort of like unhobbling path where you, it needs to, it needs to learn this kind of like system two process.

Speaker 2 And if it can learn this sort of system two process, it can just use kind of millions of tokens and think for them and be cohesive and be coherent.

Speaker 2 You know, one analogy. So when you drive, here's an analogy.
When you drive, right? Okay, you're driving. And

Speaker 2 most of the time you're kind of on autopilot, right? You're just kind of driving and you're doing well. And then,

Speaker 2 but sometimes you hit like a weird construction zone or a weird intersection, you know, and then I sometimes like, you know, my passenger seat, my girlfriend, I'm kind of like, ah, be quiet for a moment.

Speaker 2 I need to figure out what's going on.

Speaker 2 And that's sort of like, you know, you go from autopilot to like the system two is jumping in, and you're thinking about how to do it. And so the scaling is improving that system one autopilot.

Speaker 2 And I think it's sort of the brute force way to get to kind of agents. If you just improve that sort of system, but if you can get that system two working, then I think you could quite quickly jump

Speaker 2 to sort of this like more identified, you know, test time compute overhang is unlocked.

Speaker 1 What's the reason to think that is an easy win in the sense that oh, you just get the there's like some loss function that easily enables you to train it to enable the system to thinking?

Speaker 1 Yeah, there's not a lot of animals that have system two thinking, you know, it like took a long time for evolution to give us system two thinking.

Speaker 1 Yeah, the free training, it like listen, I get it, you got like trillions of tokens of internet text.

Speaker 1 I get that, like, yeah, you like match that and you get all these, yeah, um, all this free training capabilities. What's the reason to think that this is an easy and hobbling?

Speaker 2 Yeah, so okay, a bunch of things.

Speaker 1 So,

Speaker 1 first of all, pre-training is magical, right? And

Speaker 2 it gave us this huge advantage

Speaker 2 for models of general intelligence because

Speaker 2 you just predict the next token, but predicting the next token, I mean, it's sort of a common misconception. But what it does is it lets this model learn these incredibly rich representations, right?

Speaker 2 Like these sort of representation learning properties are the magic of deep learning.

Speaker 2 You have these models, and instead of learning just kind of like, you know, whatever, statistical artifacts or whatever, it learns sort of these models of the world.

Speaker 2 You know, that's also why they can kind of like generalize, right? Because it learned the right representations.

Speaker 2 And so you pre-train these models and you have this sort of like raw bundle of capabilities that's really useful. And sort of this almost unformed raw mass.

Speaker 2 And sort of the unhobbling we've done over sort of like GPT2 to GP4 was you kind of took this sort of like raw mass and then you like RLHF'd it into really good chatbot.

Speaker 2 And that was a huge win, right? Like going, going, you know,

Speaker 2 in the original, I think, in StructGPT paper, you know, RLHF versus non-RLHF model, it's like 100x model size win on sort of human preference ratings.

Speaker 2 You know, it started started to be able to do like simple chain of thought and so on. But you still have this advantage of all these kind of like raw capabilities.

Speaker 2 And I think there's still like a huge amount that you're not doing with them. And by the way, I think this sort of this pre-training advantage is also sort of the difference to robotics, right?

Speaker 2 Where I think robotics, you know,

Speaker 2 I think you people used to say it was a hardware problem, but I think the hardware stuff is getting solved.

Speaker 2 But the thing we have right now is you don't have this sort of huge advantage of being able to bootstrap yourself with pre-training. You don't have all this sort of unsupervised learning you can do.

Speaker 2 You have to start right away with the sort of RL self-play and so on. All right, so now the question is: why,

Speaker 2 you know, why might some of this on-hobbing and RL and so on work?

Speaker 2 And again, there's sort of this advantage of bootstrapping, right? So, you know, your Twitter bio is being pre-trained, right?

Speaker 1 But you're actually not being pre-trained anymore. You are not being pre-trained anymore.

Speaker 2 You are pre-trained in like grade school and high school. At some point, you transition to being able to learn by yourself, right?

Speaker 2 You weren't able to do that in elementary school.

Speaker 2 I don't know, middle school, probably, high school is maybe when it's sort of started. You need some guidance.

Speaker 2 College,

Speaker 2 if you're smart, you can kind of teach yourself. And sort of models are just starting to enter that regime.
And so it's sort of like, it's a little bit, it's probably a little bit more scaling.

Speaker 2 And then you got to figure out what goes on top. And it won't be trivial.
So

Speaker 2 a lot of deep learning is sort of like, you know, it sort of seems very obvious in retrospect. And there's sort of this some obvious cluster of ideas, right?

Speaker 2 There's sort of some kind of like thing that seems a little dumb, but that just kind of works. But there's a lot of details you have to get right.

Speaker 2 So I'm not saying this, you know, we're going to get this, you know, next month or whatever. I think it's going to take a while to like really figure out the details.

Speaker 1 A while for you is like half a year or something.

Speaker 1 I don't know. I think next month, six months.

Speaker 1 Between six months and three years, you know?

Speaker 2 But, you know, I think it's possible. And I think there's,

Speaker 2 you know, I think, and this is, I think it's also very related to the sort of issue of the data wall. But I mean, I think the,

Speaker 2 you know, one intuition on the sort of like learning, learning. learning by yourself, right, is sort of pre-training is kind of the words are flying by.
Yeah. Right.

Speaker 2 You know, and, and um, or it's like you know, the teacher is lecturing to you, and the models, you know, the words are flying by, you know, they're taking, they're just getting a little bit from it, um, but that's sort of not what you do when you learn from yourself, right?

Speaker 2 When you learn by yourself, you know, so you're reading a dense math textbook, you're not just kind of like skimming through it once, you know, you wouldn't learn that much from it.

Speaker 2 I mean, some word cells just skim through, you know, reread and reread the math textbook, and then they memorize. And that's sort of, you know, like if you just repeated the data, then they memorize.

Speaker 2 What you do is you kind of like you read a page, you kind of think about it, you have some internal monologue going on, you have a conversation with a study buddy, you try a practice problem, you know, you fail a bunch of times, and at some point it clicks, and then you're like, this made sense.

Speaker 2 Then you read a few more pages. And so we've kind of bootstrapped our way to being able to do that now with models, or like just starting to be able to do that.

Speaker 2 And then the question is, you know, being able to read it, think about it, you know, try problems.

Speaker 2 And the question is, can you, you know, all this sort of self-play synthetic data RL is kind of like making that thing work.

Speaker 2 So basically

Speaker 2 translating like in context, like right, like right now, there's like in context learning, right? Super sample efficient. There's that, you know, in the Gemini paper, right?

Speaker 2 It just like learns a language in context.

Speaker 2 And then you have pre-training, not at all sample efficient.

Speaker 2 But you know, what humans do is they kind of like they do in context learning. You read a book, you think about it until eventually it collects.

Speaker 2 But then you somehow distill that back into the weights.

Speaker 2 And in some sense, that's sort of like what RL is trying to do. And like when RL is super finicky, but when RL works, RL is kind of magical because it's sort of the best possible data for the model.

Speaker 2 It's like when you try a practice problem and it, you know, and then you fail, and at some point you kind of figure it out in a way that makes sense to you.

Speaker 2 That's sort of like the best possible data for you because like the way you would have solved the problem. And that's sort of, that's what RL is.

Speaker 2 Rather than just, you know, you kind of read how somebody else solved the problem and doesn't, you know, traditionally click.

Speaker 1 Yeah.

Speaker 1 By the way, if that take sounds familiar, because it was like part of the question I asked John Shulman, that goes to illustrate the thing I said in the intro, where like a bunch of the things I've learned about AI, just like we do these dinners before the interviews and

Speaker 1 me with Sholto and a couple and like what should I ask John Shulman what should I ask Dario

Speaker 2 okay suppose this is the way things go and we get these enhoblings yeah now and the scaling right so it's like you have this baseline just enormous force of scaling right where it's like gpt2 to gpt4 you know gp2 it could kind of like it was amazing right it could string together plausible sentences um but you know it could it could barely do anything it was kind of like preschooler and then gpt4 is you know it's writing code it like you know can do hard math it's so sort of of like smart high school.

Speaker 2 And so this big jump, and, you know, in sort of the essay series, I go through and kind of count the orders of magnitude of compute scale up, of algorithmic progress.

Speaker 2 And so sort of scaling alone, you know, sort of by 27, 28 is going to do another kind of preschool to high school jump on top of GPT-4.

Speaker 2 And so that'll already be just like at a per token level, just incredibly smart. That'll get you some more reliability.

Speaker 2 And then you'll add these on hoblings that make it look much less like a chatbot, more like this agent, like a drop-in remote worker.

Speaker 2 And, you know, that's when things really get going.

Speaker 1 Okay.

Speaker 1 I want to ask more questions about this. I think.

Speaker 1 Yeah, yeah.

Speaker 1 Let's zoom out. Okay.
So suppose you're right about this. Yeah.

Speaker 1 And I guess you, this is because of the 2027 cluster, we've got 10 gigawatt, 2027, 10 gigawatts.

Speaker 2 28 is the 10 gigawatts. Yeah.

Speaker 1 Maybe it'll be pulled forward. Okay, sure.
Something. Yeah.
And so I guess that's like 5.5 level by 2027. Like whatever that's called, right?

Speaker 1 What does the world look like at that point?

Speaker 1 You have these remote workers who can replace people. What is the reaction to that in terms of the economy, politics, geopolitics?

Speaker 2 Yeah, so,

Speaker 2 you know, I think 2023 was kind of a really interesting year to experience as somebody who was like, you know, really following the AI stuff. Where, you know, before that,

Speaker 1 what were you doing in 2023?

Speaker 2 I mean, open AI.

Speaker 1 Oh, yeah, yeah, yeah, yeah.

Speaker 2 And,

Speaker 2 you know, it kind of went, you know, I mean, you know, I was been thinking about this and, you know, like talking to a lot of people, you know, in the years before, and it was this kind of weird thing, you know, you almost didn't want to talk about AI or AGI.

Speaker 2 You know, it was kind of a dirty word, right? And then 2023, you know, people saw ChatGPT for the first time and they saw GPT4 and it just like exploded, right?

Speaker 2 And it triggered this kind of like, you know, you know, huge sort of capital expenditures from all these firms and, you know, the explosion in revenue from NVIDIA and so on. And,

Speaker 2 you know, things have been quiet since then, but you know, the next thing has been in the oven. And I sort of expect sort of every generation, these kind of like G-forces to intensify, right?

Speaker 2 It's like people see the models.

Speaker 2 There's like, you know, people haven't counted the homes, so they're going to be surprised. It'll be kind of crazy.
And then, you know, revenue is going to accelerate.

Speaker 2 You know, suppose you do hit the 10 billion, you know, end of this year. Suppose it like just continues on this sort of doubling trajectory of, you know, like every six months of revenue doubling.

Speaker 2 You know, it's like, you're not actually that far from 100 billion. You know, maybe that's like 26.

Speaker 2 And so, you know, at some point, you know, like, you know, sort of what happened to NVIDIA is going to happen to big tech. You know, like their stocks, you know, that's going to explode.

Speaker 2 And I mean, I think a lot more people are going to feel it, right? I mean,

Speaker 2 2023 was the sort of moment for me where it went from kind of AGI as a sort of theoretical abstract thing, and you'd make the models to like, I see it, I feel it.

Speaker 2 And like, I see the path, I see where it's going. I like, I think I can see the cluster where it's trained on, like, the rough combination of algorithms, the people, like how it's happening.

Speaker 2 And I think, you know, most of the world is not, you know, most of the people who feel it are like right here, you know, right?

Speaker 2 But, but, you know, I think a lot more of the world is going to start feeling it.

Speaker 2 And I think that's going to start being kind of intense.

Speaker 1 Okay. So right now, who feels it? You can go on Twitter and there's these GPT rapper companies like, whoa, GPT 4.0 is going to change our business.

Speaker 2 I'm so bearish on the rapper companies, right? Because they're the ones that are going to be like the rapper companies are betting on stagnation, right?

Speaker 2 The rapper companies are betting like you have these intermediate models and it takes so much stuff to integrate them.

Speaker 2 And I'm kind of like, I'm really bearish because I'm like, we're just going to sonic voom you, you know, and we're going to get the unhobblings, we're going to get the drop-in remote worker, and then, you know, your stuff is not going to matter.

Speaker 2 Okay, sure, sure.

Speaker 1 So that's done. Now,

Speaker 1 who, so

Speaker 1 SF is paying attention now, or this crowd here is paying attention, who is going to be paying attention 2026, 2027?

Speaker 1 And presumably, these are years in which the hundreds of billions of CapEx is being spent on AI.

Speaker 2 I mean, I think the national security state is going to be starting to pay a lot of attention.

Speaker 2 And I hope we get to talk about that a lot.

Speaker 1 Okay, let's talk about it now.

Speaker 1 Like, what is the sort of political reaction immediately? Yeah. And even like internationally, like what people see, like right now, I don't know if Xi Jinping reads the news and sees like,

Speaker 1 oh my God, like mmlu score on that what are you doing about this comrade um yeah

Speaker 1 um and so what happens when the the like what the gbd he's like sees a remote replacement and it has a hundred billion dollars in revenue there's a lot of businesses that have a hundred billion dollars in revenue and people don't like aren't staying up all night talking about it um the question i think the question is when when does the ccp and when does the sort of american national security establishment realize that super intelligence is going to be absolutely decisive for national power, right?

Speaker 2 And this is where, you know, the sort of intelligence explosion stuff comes in, which we should also talk about later.

Speaker 2 It's sort of like you have AGI, you have this sort of drop-in remote worker that can replace you or me, at least at sort of remote jobs, kind of jobs.

Speaker 2 And then

Speaker 2 I think fairly quickly,

Speaker 2 by default, you turn the crank one or two more times, and then you get a thing that's smarter than humans. But I think even more than just turning the cramp a few more times, crank a few more times,

Speaker 2 I think one of the first jobs to be automated is going to be that of sort of an AI researcher engineer. And if if you can automate AI research, you know, I think things can start going very fast.

Speaker 2 You know, right now there's already this trend of half an order of magnitude a year of algorithmic progress.

Speaker 2 You know, suppose, you know, at this point, you know, you're going to have GPU fleets and the tens of millions for inference or more.

Speaker 2 And you're going to be able to run like 100 million human equivalents of these sort of automated AI researchers.

Speaker 2 And if you can do that, you know, you can maybe do a decade's worth of sort of ML research progress in a year. You know, you get some sort of 10x speed up.

Speaker 2 And if you can do that, I think you can make the jump to kind of like AI that is vastly smarter than humans, you know, within a year, a couple years. And then, you know, that broadens, right?

Speaker 2 So you have this, you have this sort of initial acceleration of AI research. That broadens to like you apply R D to a bunch of other fields of technology.

Speaker 2 And the sort of like extremes, you know, at this point, you have like a billion just super intelligent researchers, engineers, technicians, everything, you know, superbly competent, all the things.

Speaker 2 You know, they're going to figure out robotics, right? We talked about it being a software problem.

Speaker 2 Well, you know, you have have a billion of super smart, smarter than the smartest human researchers, AI researchers on your cluster, you know, at some point during the intelligence explosion, they're going to be able to figure out robotics, you know, and then again, that expands.

Speaker 2 And,

Speaker 2 you know, I think if you play this picture forward, I think it is fairly unlike any other technology in that

Speaker 2 it will, I think, you know, a couple years of lead could be utterly decisive in, say, like military competition. Right.

Speaker 2 You know, if you look at like Gulf War I, right, Gulf War I, you know, like the Western coalition forces, you know, they had, you know, like a hundred to one kill ratio, right?

Speaker 2 And that was like, they had better sensors on their tanks, you know, and they had better, you know, more precision missiles, right?

Speaker 2 Like GPS, and they had, you know, stealth, and they had sort of a few, you know, maybe 20, 30 years of technological lead, right?

Speaker 2 And they, you know, just completely crushed them.

Speaker 2 Super intelligence applied to sort of broad fields of R ⁇ D. And then, you know, the sort of industrial explosion as well, you have the robots, you're just making lots of material.

Speaker 2 You know, I think that could compress, I mean, basically compressed kind kind of like a century worth of technological progress into less than a decade.

Speaker 2 And that means that, you know, a couple years could mean a sort of Gulf War one style like, you know, advantage in military affairs.

Speaker 2 And,

Speaker 2 you know, including like, you know, a decisive advantage that even like preempts nukes, right? Suppose, like, you know, how do you find the stealthy nuclear submarines?

Speaker 2 Like, right now, that's a problem of like, you have sensors, you have the software to like detect where they are. You know, you can do that.
You can find them.

Speaker 2 You have kind of like millions or billions of like mosquito-like, you know, sized drones.

Speaker 2 And, you know, they take out the the nuclear submarines, they take out the mobile launchers, they take out the other nukes.

Speaker 2 And anyway, so I think enormously destabilizing, enormously important for national power.

Speaker 2 And at some point, I think people are going to realize that. Not yet, but they will.
And when they will,

Speaker 2 I think there will be sort of,

Speaker 2 you know, I don't think it'll just be the sort of AI researchers in charge.

Speaker 2 And, you know, I think on the, you know, the CCP is going to, you know, have sort of an all-out effort to like infiltrate American AI labs, right?

Speaker 2 You know, like billions of dollars, thousands of people, you know, full force of the sort of Ministry of State Security. CCP is going to try to outbuild us, right?

Speaker 2 Like they, you know, their power in China, you know, like the electric grid, you know,

Speaker 2 they added a U.S. as you know, a complete, like, they added as much power in the last decade as the sort of entire U.S.
electric grid.

Speaker 2 So like the 100 gigawatt cluster, at least the 100 gigawatts is going to be a lot easier for them to get.

Speaker 2 And so I think sort of, you know, by this point, I think it's going to be like an extremely intense sort of international competition.

Speaker 1 Okay, so in this picture,

Speaker 1 one thing I'm uncertain about is whether it's more like what you say, where it's more of an implosion of you have developed an AGI and then you make it into an AI researcher.

Speaker 1 And for a while, a year or something, you're only using this ability to make hundreds of millions of other AI researchers. And then, like, the thing that comes out of this

Speaker 1 really frenetic process is a super intelligence.

Speaker 2 And then that goes out in the world and is developing robotics and helping you take over other countries and and whatever I think it's a little bit more you know it's a little bit more kind of like you know it's not like you know on and off it's a little bit more gradual but it's sort of like it's an explosion that starts narrowly it can do cognitive jobs you know the highest RI used for cognitive jobs is make the AI better like solve robotics you know and as as as you solve robotics now you can do RD and you know like biology and other technology

Speaker 2 you know initially you start with the factory workers you know they're wearing the glasses and the AirPods you know and the AI is instructing them right because you know you kind of make any worker into a skilled technician and then you have the robots come in and anyway so it sort of expands this process expands

Speaker 1 Meta's Ray-Vans are a compliment to their llama. You know, it's like, whatever.

Speaker 2 Like, you know, the fabs in the US, the constrained to skilled workers, right?

Speaker 2 You have, you have the, even if you don't have robots yet, you have the cognitive superintelligence, and you know, you can kind of make them all into skilled workers immediately.

Speaker 2 But that's, you know, that's a very brief period. You know, robots will come soon.
Sure. Okay.

Speaker 1 Okay. So suppose this is actually how the tech progresses.
In the United States, maybe because these companies are already experiencing hundreds of billions of dollars of AI revenue.

Speaker 2 At this point, you know, companies are borrowing hundreds of billions or more in the corporate debt markets.

Speaker 1 But why is a CCP bureaucrat, some 60-year-old guy, he looks at this and he's like, oh, it's like Copilot has gotten better now.

Speaker 1 Why are they now

Speaker 2 than Copilot has gotten better now?

Speaker 1 I mean, at this point, to them,

Speaker 1 like, yeah, so they're, because to shift the production of an entire country

Speaker 1 to dislocate energy that is otherwise being used for

Speaker 1 consumer goods or something and to make that all feed into the data centers. What

Speaker 1 would that be obviously? Because part of this whole story is you realize the super intelligence is coming soon, right? And I guess you realize it, maybe I realize it.

Speaker 1 I'm not sure how much I realize it. But

Speaker 1 will the national security outputs in the United States and will the CCP realize it?

Speaker 2 Yeah, I mean, look, I think in some sense, this is a really key question.

Speaker 2 I think we have sort of a few more years of mid-game, basically, and where you have a few more 2023s, and that just starts updating more and more people.

Speaker 2 And, you know, I think the trend lines will become clear.

Speaker 2 You know, I think you will see some amount of the sort of COVID dynamic, right?

Speaker 2 You know, like COVID was like, you know, February, February of 2020, you know, it's like, honestly, feels a lot like today, you know, where it's like, you know, it feels like this utterly crazy thing is happening, is about, you know, is impending, is coming.

Speaker 2 You kind of see the exponential, and yet most of the world just doesn't realize, right?

Speaker 2 The mayor of New York is like, go out to the shows, and this is just, you know, like Asian racism or whatever, you know, and and

Speaker 2 but you know, at some point the exponential, like, you know, at some point people saw it. And then, you know, like just kind of crazy radical reactions came.

Speaker 1 Right. Okay.
So, by the way, what were you doing during COVID? Or when like February? Okay.

Speaker 1 Like, freshman, sophomore, what?

Speaker 2 Junior. Yeah.

Speaker 1 But still, like, what were we like, 17-year-old junior or something?

Speaker 1 And

Speaker 1 then you bought, like, did you short the market or something?

Speaker 2 Yeah, yeah, yeah.

Speaker 1 Did you, did you, but did you sell at the right time? Yeah. Okay.

Speaker 1 Yeah. So there will be like a March 2020 moment that the thing that was COVID, but here.

Speaker 1 Now,

Speaker 1 then you can like make the analogy that you make in a series that this will then

Speaker 1 cause the reaction of like, we got to do the Manhattan Project for America here.

Speaker 1 I wonder what the politics of this will be like, because the difference here is it's not just like, we need the bomb to beat the Nazis.

Speaker 1 It's we're building this thing that's making all our energy prices rise a bunch and it's automating a bunch of our jobs.

Speaker 1 And the climate change stuff, like people are going to be like, oh my God, it's making climate change worse. And it's helping big tech.

Speaker 1 Like, politically, this doesn't seem like a dynamic where the national security apparatus or the president is like,

Speaker 1 we have to step on the gas here and make sure America wins.

Speaker 2 Yeah. I mean, again, I think a lot of this really depends on sort of how much people are feeling it, how much people are seeing it.

Speaker 2 You know, I think there's a thing where, you know, kind of basically our generation, right?

Speaker 2 We're kind of so used to kind of, you know, basically peace and like, you know, the world, you know, American hegemony and nothing matters.

Speaker 2 But, you know, the sort of like extremely intense and these extraordinary things happening in the world

Speaker 2 and like intense international competition is like very much the historical norm. Like in some sense, it's like,

Speaker 2 you know, sort of this, there's this sort of 20-year, very unique period. But like, you know,

Speaker 2 the history of the world is like, you know,

Speaker 2 like in World War II, right, it was like 50% of GDP went to, you know, like, you know, warped in production. The U.S.
borrowed over 60% of GDP.

Speaker 2 You know, and in, you know, I think Germany and Japan over 100%. World War I, you know, UK, Japan, sorry, UK, France, Germany all borrowed over 100% of GDP.

Speaker 2 And,

Speaker 2 you know, I think the sort of

Speaker 2 much more was on the line, right?

Speaker 2 Like, you know, and you know, people talk about World War I being so destructive and, you know, like 20 million Soviet soldiers dying and like 20% of Poland, but you know, that was just the sort of like that happened all the time, right?

Speaker 2 You know, like seven years' war, you know, like whatever, 20, 30% of Prussia died, you know, like 30 years' war, you know, like

Speaker 2 I think, you know, up to 50% of like large swaths of Germany died.

Speaker 2 And,

Speaker 2 you know, I think the question is, will these sort of like,

Speaker 2 will people see that the stakes here are really, really high? And that basically sort of like history is actually back.

Speaker 2 And I think, you know, I think the American National Security State thinks very seriously about stuff like this. They think very seriously about competition with China.

Speaker 2 I think China very much thinks of itself on this as our historical mission and rejuvenation of the Tiny's nation. They think a lot about national power.
They think a lot about like the world order.

Speaker 2 And then, you know,

Speaker 2 I think there's a real question on timing, right? Like, do they, do they start taking this seriously, right?

Speaker 2 Like, when the intelligence explosion is already happening, like quite late, or do they start taking this seriously like two years earlier? And that matters a lot for how things play out.

Speaker 2 But at some point, they will. And at some point, they will realize that this will be sort of utterly decisive

Speaker 2 for

Speaker 2 you know, not just kind of like some proxy war somewhere, but, you know, like whether liberal democracy can continue to thrive, whether, you know, whether the CCP will continue existing.

Speaker 2 And I think that will activate sort of forces that we haven't seen in a long time.

Speaker 1 The great conflict, the great power conflict thing definitely seems compelling.

Speaker 1 I think just all kinds of different things seem much more likely when you think from a historical perspective, when you zoom out beyond the liberal democracy that we've been living in, had the pleasure to live in America, let's say 80 years,

Speaker 1 including dictatorships, including

Speaker 1 obviously war, famine, whatever.

Speaker 1 I was reading the Gulag Archipelago, and and one of the chapters begins with Sojenitsin saying, if you would have told a Russian citizen under the Tsars that because of all these new technologies, we wouldn't see some great Russian revival or becomes a great power and the citizens are

Speaker 1 made wealthy.

Speaker 1 But instead, what you would see is tens of millions of Soviet citizens tortured by millions of beasts in the worst possible ways, and that this is what would be the result of the 20th century.

Speaker 1 They wouldn't have believed you. They'd have called you a slanderer.

Speaker 2 Yeah. And, you know, the,

Speaker 2 you know, the possibilities for dictatorship with superintelligence are sort of even crazier, right? I think, you know, imagine you have a perfectly loyal military and security force, right?

Speaker 2 That's it. No more, no more rebellions, right? No more popular uprisings, you know, perfectly loyal.
You know, you have, you know, perfect lie detection, you know, you have surveillance of everybody.

Speaker 2 You know, you can perfectly figure out who's the dissenter, weed them out. You know, no Gorbachev would have ever risen to power, who had some doubts about the system.

Speaker 2 You know, no military coup would have ever happened. And I think you, I mean,

Speaker 2 you know, I I think there's a real way in which,

Speaker 2 you know, part of why things have worked out is that

Speaker 2 you know, ideas can evolve. And, you know, there's sort of like some sense in which sort of time heals a lot of wounds and time, you know, and

Speaker 2 solves, you know, a lot of debates.

Speaker 2 And a lot of people had really strong convictions, but you know, a lot of those have been overturned by time because there's been this continued pluralism and evolution.

Speaker 2 I think there's a way in which kind of like, you know, if you take a CCP-like approach to kind of like truth, truth is what the party says, and you supercharge that with super intelligence.

Speaker 2 I think there's a way in which that could just be like locked in and enshrined for

Speaker 2 a long time. And I think the possibilities are pretty terrifying.

Speaker 2 You know, your point about history and sort of like living in America for the past eight years, you know,

Speaker 2 I think this is one of the things I sort of took away from growing up in Germany is a lot of this stuff feels more visceral, right?

Speaker 2 Like, you know, my mother grew up in the former East, my father in the former West. They like met shortly after the wall fell, right?

Speaker 2 Like the end of the Cold War was this sort of extremely pivotal moment for me because it's, you know, it's the reason I exist, right?

Speaker 2 And then, you know, growing up in Berlin and, and, you know, the former wall, you know,

Speaker 2 my great-grandmother, who is, you know, still alive, is very important in my life. You know, she was born in 34, you know, grew up, you know, during the Nazi era, during, you know, all that.

Speaker 2 You know, then World War II, you know, like saw the fire bombing of Dresden from the sort of, you know, country cottage or whatever where, you know, they as kids were, you know, then, and then, you know, then spends most of her life in sort of the East German communist dictatorship.

Speaker 2 You know, she'd tell me about, you know, in like 54, when there's like the popular uprising, you know, Soviet tanks came in, you know, her, her husband was telling her to get home really quickly, you know, get off off the streets, you know, had a, had a son who, who tried to, you know, ride a motorcycle across the Iron Curtain and then was put in Nestasi prison for a while.

Speaker 2 You know, and then finally, you know, when she's almost 60, you know, it was the first time she lives in, you know, a free country

Speaker 2 and a wealthy country. And,

Speaker 2 you know, when I was a kid, she was,

Speaker 2 she, the thing she always really didn't want me to do was like get involved in politics because like joining a political party was just, you know, it was very bad connotations for her.

Speaker 2 Anyway, and she, and she sort of raised me when I was young, you know, and so it,

Speaker 2 you know, it doesn't feel that long ago. It feels very close.

Speaker 1 Yeah.

Speaker 1 So I wonder when we're talking today about the CCP, listen, the people in China who will be doing

Speaker 1 their version of the project will be AI researchers who are somewhat westernized, who interact with,

Speaker 1 either got educated in the West West or have colleagues in the West.

Speaker 1 Are they going to

Speaker 1 sign up for the CCP project that's going to hand over

Speaker 1 control to Xi Jinping?

Speaker 1 What's your sense on, I mean, you're just like, fundamentally, they're just people, right? Like, can't you like convince them about the dangers of super intelligence? Will they be in charge, though?

Speaker 2 I mean, in some sense, this is, I mean, this is also the case, you know,

Speaker 2 you know, in the U.S. or whatever.
This is sort of like rapidly depreciating influence of the lab employees.

Speaker 2 Like right now, the sort of AI lab employees have so much power, right, over this, you know, like you said. But they're going to get automated and then

Speaker 2 you saw this November event, so much power, right? But both, I mean, both are going to get automated and they're going to lose all their power.

Speaker 2 And it'll just be, you know, kind of like a few people in charge with their sort of armies of automated eyes. But also,

Speaker 2 you know, it's sort of like the politicians and the generals and the sort of national security state.

Speaker 2 You know, a lot, you know, I mean, there's sort of, this is the sort of some of these classic scenes from the Oppenheimer movies.

Speaker 2 You know, the scientists built it and then it was kind of, you know, then the bomb was shipped away and it was out of their hands.

Speaker 2 You know, I actually, I think, I actually think it's good for like lab employees to be aware of this. It's like, you have a lot of power now, but

Speaker 2 maybe not for that long. And, you know, use it wisely.

Speaker 2 Yeah, I do think they would benefit from some more organs of representative democracy.

Speaker 1 What do you mean by that?

Speaker 2 Oh, I mean,

Speaker 2 in the open AI board events, employee power was exercised in a very sort of direct democracy way. And I feel like that's some of how that went about.

Speaker 2 I think really highlighted the benefits of representative democracy and having some deliberative organs. Interesting.

Speaker 1 Let's go back to the 100 billion revenue, whatever. And so these companies.

Speaker 1 Yeah. The companies are deploying, trying to build clusters that are this big.
Yeah. Where are they building it?

Speaker 1 Because if you say it's the amount of energy that would be required for a small or medium-sized U.S.

Speaker 1 state, is it then Colorado gets no power and it's happening in the United States or is it happening somewhere else?

Speaker 2 Oh, I mean, I think that, I mean, in some sense, this is the thing that I always find funny is: you know, you talk about Colorado gets no power.

Speaker 2 You know, the easy way to get the power would be like, you know, displace less economically useful stuff.

Speaker 2 You know, it's like whatever, buy up the aluminum smelting plant and, you know, that has a gigawatt and, you know, we're going to replace it with the data center because that's important.

Speaker 2 I mean, that's not actually happening because a lot of these power contracts are really sort of long-term locked in. You know, there's obviously people don't like things like this.

Speaker 2 And so it sort of, it seems like in practice, what it's, what it's requiring, at least right now, is building new power.

Speaker 2 That might change. And I think that that's when things get really interesting, when it's like, no, we're just dedicating all of the power to the AGI.
But anyway, so right now it's building new power.

Speaker 2 10 gigawatt, I think, quite doable.

Speaker 2 You know, it's like a few percent of like US natural gas production.

Speaker 2 You know, when you have the 10 gigawatt chaining cluster, you have a lot more inference. So that starts getting more, you know, I think 100 gigawatt that starts getting pretty wild.

Speaker 2 You know, that's, you know, again, it's like over 20% of U.S. electricity production.

Speaker 2 I think it's pretty doable, especially if you're willing to go for like natural gas.

Speaker 2 But

Speaker 2 I do think it is incredibly important, incredibly important that these clusters are in the United States.

Speaker 1 And why does it matter? It's in the US?

Speaker 2 I mean, look, I think there's some people who are

Speaker 2 trying to build clusters elsewhere. And there's like a lot of of free-flowing middle eastern money that's trying to build clusters elsewhere um

Speaker 2 i think this comes back to the sort of like national security question we talked about earlier like would you i mean would you do the manhattan project in the uae right and i think i think basically like putting putting the clusters you know i think you can put them in the us you can put them in sort of like ally democracies but i think once you put them in kind of like you know dictatorships authoritarian dictatorships you kind of create this you know irreversible security risk right so i mean one cluster is there, much easier for them to exfiltrate the weights.

Speaker 2 You know, they can like literally steal the AGI, the super intelligence. It's like they got a copy of the, you know, of the atomic bomb, you know, and they just got a direct replica of that.

Speaker 2 And it makes it much easier to them. I mean, we are ties to China.
You can ship that to China. So that's a huge risk.
Another thing is they can just seize the compute, right?

Speaker 2 Like maybe right now they just think of this.

Speaker 2 I mean, in general, I think people, you know, I think the issue here is people are thinking of this as the chat GPT big tech product clusters, but I think the clusters being planned now, three to five years out, like may well be the like AGI superintelligence clusters.

Speaker 2 And so anyway, so like when things get hot, you know, they might just seize the compute.

Speaker 2 And I don't know, suppose we put like, you know, 25% of the compute capacity in these sort of Middle Eastern dictatorships.

Speaker 2 Well, they seize that, and now it's sort of a ratio of compute of three to one.

Speaker 2 And, you know, we still have some more, but even like, even, even only, only 25% of compute there, like, I think it starts getting pretty hairy.

Speaker 2 You know, I think three to one is like not that great of a ratio. You can do a lot with that amount of compute.
And then, look, even if they don't actually do this, right?

Speaker 2 Even if they don't actually seize the compute, even if they actually don't steal the weights,

Speaker 2 there's just a lot of implicit leverage you get, right? They get the seat at the AGI table.

Speaker 2 And,

Speaker 2 I don't know why we're giving authoritarian dictatorships the seat at the AGI table.

Speaker 1 Okay, so there's going to be a lot of compute in the Middle East if these deals go through. First of all,

Speaker 1 is it just like every single big tech company is just trying to figure out what they have?

Speaker 1 Okay, okay.

Speaker 1 I guess there's reports. I think Microsoft or

Speaker 1 which we'll get into.

Speaker 1 UAE gets a bunch of compute because we're building the clusters there.

Speaker 1 And why? So let's say they have 25% of the, why does a compute ratio matter?

Speaker 1 Is it, if it's about them being able to kick off the intelligence explosion, isn't it just some threshold where you have 100 million AI researchers or you don't?

Speaker 2 I mean, you can do a lot with, you know, 33 million extremely smart scientists.

Speaker 2 And, you know, again, a lot of the stuff, you know, so first of all, it's like, you know, that might be enough to build the crazy bioweapons, right?

Speaker 2 And then you're in a situation where like, now, wow, we've just like, they stole the weights, they seized the compute. Now they can make, you know,

Speaker 2 they can build these crazy new WMDs WMDs that, you know, will be possible with super intelligence. And now you've just kind of like proliferated the stuff.
And, you know, it'll be really powerful.

Speaker 2 And also, I mean, I think, you know, three to three acts on compute isn't actually that much. And so the,

Speaker 2 you know, the,

Speaker 2 you know, I think a thing I worry a lot about is

Speaker 2 I think everything, I think the riskiest situation is if we're in some sort of like really tight-neck, feverish international struggle, right?

Speaker 2 If we're like really close with the CCP and we're like months apart,

Speaker 2 I think the situation we want to be in, we could be in if we played our cards right, is a little bit more like, you know, the US, you know, building the atomic bomb versus the German project, way behind, you know, years behind.

Speaker 2 And if we have that, I think we just have so much more wiggle room, like to get safety right.

Speaker 2 We're going to be building like, you know, there's going to be these crazy new WMDs, you know, things that completely undermine, you know, nuclear deterrence, you know, intense competition.

Speaker 2 And that's so much easier to deal with if, you know, you're like, you know, it's not just, you know, you don't have somebody right on your tails. You got to go, go, go.
You got to go at maximum speed.

Speaker 2 You have no wiggle room.

Speaker 2 You're worried that at any time they can overtake you. I mean, they can also just try to outbuild you, right? Like they can might, they might literally win.

Speaker 2 Like China might literally win if they can steal the weights because they can outbuild you.

Speaker 2 And they maybe have less caution, both, you know, good and bad caution, you know, kind of like whatever unreasonable regulations we have.

Speaker 2 Or you're just in this really tight race.

Speaker 2 And I think it's that sort of like, if you're in this really tight race, this sort of feverish struggle, I think that's when sort of there's the greatest peril of self-destruction.

Speaker 1 So then presumably the companies that are trying to build clusters in the Middle East realize this. But is it just that it's impossible to do this in America?

Speaker 1 And if you want American companies to do this at all, then you do it in the Middle East or not at all. And then you just like have China build a three gorgeous damn cluster.

Speaker 2 I mean, there's a few reasons. One of them is just like people aren't thinking about this as the AGI superintelligence cluster.

Speaker 2 They're just like, ah, you know, like cool clusters for my, you know, for my chat GPS.

Speaker 1 But so they're building in the plans right now are clusters which are ones that are like, because if you're doing once for inference, presumably you could like spread them out across the country or something.

Speaker 1 But the ones they're building, they realize we're going to do one training run in this thing we're building.

Speaker 2 I just think it's harder to distinguish between inference and training compute.

Speaker 2 And so people can claim it's training compute, but I think they might realize that actually, you know, this is going to be useful for

Speaker 2 actually it's useful for training compute too.

Speaker 1 Anyway, because of synthetic data and things like that.

Speaker 2 Yeah, the future of trending, you know, like RL looks a lot like inference, for example, right? Or you just kind of like end up connecting them, you know, in time.

Speaker 2 You know, it's that you have this like a lot of raw material. You know, it's like, you know, it's placing your uranium refinement facilities there.
Sure. Anyway, so a few reasons, right?

Speaker 2 One is just like they don't think about this as the the AGI cluster. Another is just like easy money from the Middle East, right?

Speaker 2 Another one is like, you know,

Speaker 2 people saying, some people think that, you know, you can't do it in the US. And,

Speaker 2 you know, I think we actually face a sort of real system competition here.

Speaker 2 Cause again, some people think there's only autocracies that can do this, that can kind of like top down, mobilize the sort of industrial capacity, the power, you know, get the stuff done fast.

Speaker 2 And again, this is the sort of thing, you know, we haven't faced in a while.

Speaker 2 But, you know, during the Cold War, like we really, there was this sort of intense system competition, right? Like East-West Germany was this, right?

Speaker 2 Like West Germany kind of like liberal democratic capitalism versus kind of communist, state-planned.

Speaker 2 And

Speaker 2 now it's obvious that the sort of

Speaker 2 free world would win. But even as late as 61, Paul Samuelson was predicting that the Soviet Union would outgrow the United States because they were able to sort of mobilize industry better.

Speaker 2 And so yeah, there's some people who, you know, they shit post about loving America by day, but then in private, they're betting against America. They're betting against the liberal order.

Speaker 2 And I I think, I basically just think it's a bad bet. And the reason I think it's a bad bet is I think this stuff is just really possible in the US.

Speaker 2 And so to make it possible in the US, there's some amount that we have to get our act together, right? So I think there's basically two paths to doing it in the US.

Speaker 2 One is you just got to be willing to do natural gas. And there's ample natural gas, right? You put your cluster in West Texas, you put it in Southwest Pennsylvania by the Marcella Shale.

Speaker 2 10 gigawatt cluster is super easy. 100 gigawatt cluster, also pretty doable.

Speaker 2 I think natural gas production in the United States is almost doubled in a decade. If you do that one more time time over the next

Speaker 2 seven years or whatever, you could power multiple trillion-dollar data centers.

Speaker 2 But the issue there is a lot of people have sort of these made these climate commitments.

Speaker 2 They're not just government, it's actually the private companies themselves, the Microsoft, the Amazons, and so on. They have these climate commitments.
So they won't do natural gas.

Speaker 2 And I admire the climate commitments, but I think at some point,

Speaker 2 the national interest and national security kind of is more important.

Speaker 2 The other path is like, you can do the sort of green energy mega projects, right? You do the solar and the batteries and the you know, the SMRs and geothermal.

Speaker 2 But if we want to do that, there needs to be sort of a sort of broad deregulatory push, right? So like you can't have permitting take a decade, right? So you got to reform FERC.

Speaker 2 You got to have blanket NEPA exemptions for this stuff. There's like inane state-level regulations that are like, yeah,

Speaker 2 you can build the solar panels and batteries next to your data center, but it'll still take years because

Speaker 2 you actually have to hook it up to the state electrical grid.

Speaker 2 And you have to use governmental powers to create rights of way to kind of like, you know, have multiple clusters and connect them, you know, and have thick cables, basically.

Speaker 2 And so, look, I mean, ideally, we do both, right? Ideally, we do natural gas and the broad regulatory agenda. I think we have to do at least one.

Speaker 2 And then I think this possible stuff is just possible in the United States. Yeah.

Speaker 1 I think a good analogy for this, by the way, before the conversation, I was reading,

Speaker 1 there's a good book about World War II industrial mobilization in the United States called Freedom's Forge.

Speaker 1 And

Speaker 1 I guess when we think back on that period, especially if you're from, if you read like the Patrick Collins and Fast and the Progress Studies stuff, it's like you had state capacity back then and people just got shit done, but now it's a cluster file.

Speaker 1 Is that the case? No, so it was really interesting. So you have people who are from the Detroit auto industry side, like Knutson, who are running mobilization for the United States.

Speaker 1 And they were extremely competent.

Speaker 1 But then at the same time, you had labor organization and agitation, which is actually very analogous to the climate pledges and climate change concern we have today. Yeah.

Speaker 1 Where they would have these strikes while literally into 1941 that would cost millions of man hours worth of time when we're trying to make tens of millions, sorry, tens of thousands of planes a month or something.

Speaker 1 And they would just debilitate factories. And for trivial like pennies on the dollar kind of concessions from capital.

Speaker 1 And it was concerns that, oh, the auto companies are trying to use the pretext of a potential war to actually prevent paying labor the money he deserves. And so

Speaker 1 what climate change is today, like you'd think, ah, fuck, America's fucked. Like, we're not going to be able to build this shit.

Speaker 1 Like, if you, uh, if you look at NEPA or something, but I didn't realize how debilitating labor was in like World War II.

Speaker 2 Right. It was just, you know, before at the, you know, sort of like 39 or whatever, the American military was in total shambles, right?

Speaker 2 You read about it, and it reads a little bit like, you know, the German military today, right? It's like, you know, military expenditures, I think, were less than 2% of GDP.

Speaker 2 You know, all all the European countries had gone, even in peacetime, you know, like above 10% of GDP, sort of this like rapid mobilization.

Speaker 2 There's nothing, you know, like we're making kind of like no planes. There's no military contracts.
Everything had been starved during the Great Depression, but there was this latent capacity.

Speaker 2 And at some point, the United States got their act together.

Speaker 2 I mean, the thing I'll say is I think, you know, the supply is sort of the other way around too, to basically to China, right?

Speaker 2 And I think sometimes people are, you know, they kind of count them out a little bit and like the export controls and so on.

Speaker 2 And, you know, they're able to make seven nanometer new chips now.

Speaker 2 I think there's a question of like how many could they make, but you know, I think there's at least a possibility that they're going to be able to mature that ability and make a lot of seven nanometer chips.

Speaker 2 And there's a lot of latent industrial capacity in China, and they are able to like, you know, build a lot of power fast. And maybe that isn't activated for AI yet.

Speaker 2 But at some point, you know, the same way the United States and like, you know, a lot of people in the US and the United States government is going to wake up.

Speaker 2 You know, at some point, the CCP is going to wake up. Yep.

Speaker 1 Okay.

Speaker 1 Going back to the question of presumably companies,

Speaker 1 are they blind to the fact that there's going to be be some sort of, well, okay, so they realize that there's going, they realize scaling is a thing, right?

Speaker 1 Obviously, their whole plans are contingent on scaling. And so they understand that we're going to be in 2028 building this in gigawatt data centers.

Speaker 1 And at this point, that the people who can keep up are big tech, just potentially at like the edge of their capabilities,

Speaker 1 then sovereign wealth fund funded things. Yeah.
And also big major countries like America, China, whatever.

Speaker 1 So what's their plan? If you look at like these AI labs, what's their plan given this landscape?

Speaker 1 Do they not want the leverage of having being in the United States?

Speaker 2 I mean, I think, I don't know. I think, I mean, one thing the Middle East does offer is capital, but it's like America has plenty of capital, right?

Speaker 2 It's like, you know, we have trillion-dollar companies. Like, what are these Middle Eastern states? They're kind of like trillion-dollar oil companies.

Speaker 2 We have trillion-dollar companies and we have very deep financial markets. And it's like, you know, Microsoft could issue hundreds of billions of dollars of bonds and they can pay for these clusters.

Speaker 2 I mean, look, I think another argument being made, and I think it's worth taking seriously, is an argument that, look, if we don't work with the UAE or with these Middle Eastern countries,

Speaker 2 they're just going to go to China, right? And so, you know, we, you know, they're going to build data centers, they're going to pour money into AI regardless.

Speaker 2 And if we don't work with them, you know, they'll just support China. And

Speaker 2 look, I mean, I think

Speaker 2 there's some merit to the argument in the sense that I think we should be doing basically benefit sharing with them, right?

Speaker 2 I think we should talk about this later, but I think basically sort of on the road to AGI, there should be kind of like two tiers of coalitions.

Speaker 2 It should be the sort of narrow coalition of democracies, that's sort of the coalition that's developing AGI.

Speaker 2 And then there should be a broader coalition where we kind of go to other countries, including dictatorships, and we're willing to offer them,

Speaker 2 you know, we're willing to offer them some of the benefits of AI, some of the sharing. And so it's like, look,

Speaker 2 if the UAE wants to use AI products, if they want to run, you know, meta-recommendation engines, if they want to run, you know, like the last generation models, that's fine.

Speaker 2 I think by default, they just wouldn't have had this seat at the AGI table, right? And so it's like, yeah, they have some money, but a lot of people have money.

Speaker 2 and um you know the only reason they're getting this sort of coarsey at the agi table the only reason we're giving these dictators will have this enormous amount of leverage um over this extremely national security relevant technology is because we're um you know we're kind of getting them excited and offering it to them um

Speaker 1 you know i think the other yeah who like who specifically is doing this like just the companies who are going there to fundraise are like this is the agi is happening and you can fund it or you can't reported that yeah it's been reported that you know sam is trying to raise you know seven trillion or whatever for a chip project.

Speaker 2 And it's unclear how many of the clusters will be there and so on. But it's

Speaker 2 definitely stuff is happening. I mean, look, I think another reason I'm a little bit at least suspicious of this argument of like, look, if the U.S.
doesn't work with them, they'll go to China is.

Speaker 2 You know, I've heard from multiple people, and this wasn't from my time at OpenAI, and I haven't seen the memo, but I have heard from multiple people that

Speaker 2 at some point several years ago, OpenAI leadership had sort of laid out a plan to fund and sell AGI by starting a bidding war between the governments of the United States, China, and Russia.

Speaker 2 And so it's kind of surprising to me that they're willing to sell AGI to the Chinese and Russian governments.

Speaker 2 But also, there's something that sort of feels a bit eerily familiar about kind of starting this bidding war and then kind of like playing them off each other.

Speaker 2 And well, if you don't do this, China will do it.

Speaker 1 So anyway. Interesting.
Okay, so that's pretty fucked up.

Speaker 1 But given that that's okay, so suppose that you're right about we ended up in this place because we got one the way one of our friends put it is that the Middle East has like no other place in the world billions of dollars or trillions of dollars up for persuasion.

Speaker 1 And

Speaker 1 this is true. And what we have at the form of standards, accountability, then like you know, the Microsoft board.

Speaker 2 It's only the dictator.

Speaker 1 Yeah, yeah, yeah. Okay, but so let's say you're right that you shouldn't have gotten them excited about AGI in the first place.
But now we're in a place where they are excited about AGI. Yeah.

Speaker 1 And they're like, fuck, we want us to have GPD-5 while you're going to be off building super intelligence. This Adams or Peace thing doesn't work for us.

Speaker 1 And if you're in this place,

Speaker 1 don't they already have the leverage? Aren't you like, and you might as well just say nothing?

Speaker 2 I think the UAE on its own is not competitive, right? It's like, I mean, they're already export-controlled.

Speaker 2 Like, you know, we're not, you know, there's like, you're not actually supposed to ship NVIDIA chips over there, right? You know, it's not like they have any of the leading AI labs.

Speaker 2 You know, it's like they have money, but you know, it's actually hard to just translate money into like.

Speaker 1 But the other things you've been saying about laying out your vision is very much there's this almost industrial process of you put in the compute and then you put in the algorithms. Yes.

Speaker 1 You add that up and you get AGA on the other end.

Speaker 1 If it's something more like that, then the case for somebody being able to catch up rapidly seems more compelling than if it's some bespoke.

Speaker 2 Well, well, if they can steal the algorithms and if they can steal the way. And that's really, that's really where sort of, I mean, we should talk about this.
This is really important.

Speaker 2 And I think, you know.

Speaker 1 So like right now, how easy would it be for

Speaker 1 an actor to steal the things that are like,

Speaker 1 not the things that are released about Scarlett Johansson's voice, but the RL things are talking about the unhoblings.

Speaker 2 I mean, I mean, all extremely easy, right? You know, I, you know, DeepMind even, like, you know, they don't make a claim that it's hard, right?

Speaker 2 DeepMind put out their like, whatever, frontier safety something, and they, like, lay out security levels, and they, you know, security levels zero to four, and four is the resilient resistant to state actors, and they say we're at level zero, right?

Speaker 2 And then, you know, I mean, just recently, there's like an indictment of a guy who just like stole the code, a bunch of like really important AI code and went to China with it.

Speaker 2 And, you know, all he had to do to steal the code was, you know, copy the code and put it into Apple Notes and then export it as PDF. And that got past their monitoring.
Right.

Speaker 2 And Google is the best security of any of the ILABs, probably, because they have the Google infrastructure.

Speaker 2 I mean, I think, I don't know, roughly, I would think of this as like, you know, security of a startup, right? And like, what does security of a startup look like? Right. You know, it's not that good.

Speaker 1 It's easy to steal. So even if that's the case.
Yeah.

Speaker 1 A lot of your posts is making the argument that, oh,

Speaker 1 you know, where are we going to get the intelligence explosion?

Speaker 1 Because if we have somebody with the intuition of an Alec Radford to become able to come up with all these ideas, ideas, that intuition is extremely valuable and you scale that up.

Speaker 1 But if it's a matter of these,

Speaker 1 if it's just in the code,

Speaker 1 if it's just the intuition, then that's not going to be just in the code, right? And also because of export controls, these countries are going to have slightly different hardware.

Speaker 1 You're going to have to make different trade-offs and probably rewrite things to be able to be compatible with that.

Speaker 1 Including all these things, is it just a matter of getting the right pen drive and you plug it into the gigawatt data center next to the Three Gorges Dam and then you're off to the races?

Speaker 2 I mean, look, there's a few different things, right? So one, one threat model is just stealing the weights themselves. And the weights one is sort of particularly insane, right?

Speaker 2 Because they can just like steal the literal like end product, right? Just like make a replica of the atomic bomb, and then they're just like ready to go.

Speaker 2 And, you know, I think that one just is, you know, extremely important around the time we have AGI and superintelligence, right? Because it's, you know, China can build a big cluster.

Speaker 1 By default, we'd have a big lead, right?

Speaker 2 Because we have the better scientists, but we make the super intelligence. They just steal it.
They're off to the races.

Speaker 2 Weights are a little bit less important right now because, you know, who cares if they steal the GP4 weights, right? Like, whatever.

Speaker 2 And so, you know, we still have to get started on weight security now. Because, you know, look, if we think AGI by 27, you know, this stuff is going to take a while.
And

Speaker 2 it's not just going to be like, oh, we do some access control. It's going to, you know, if you actually want to be resistant to sort of Chinese espionage, you know, it needs to be much more intense.

Speaker 2 The thing, though, that I think people aren't paying enough attention to is the secrets, as you say. And,

Speaker 2 you know I think this is you know the compute stuff is sexy you know we talk about it but you know I think that you know I think people underrate the secrets

Speaker 2 because they're you know I think they're you know the half an order of magnitude a year just by default sort of algorithmic progress that's huge you know if we have a few year lead by by default you know that's 10 30 x 100x bare cluster if we protected them um and then there's this additional layer of the data wall right and so we have to get through the data wall that means we actually have to figure out some sort of basic new paradigm sort of the alpha go step two right alpha go step one is learns from human imitation AlphaGo step two is the sort of self-play RL.

Speaker 2 And everyone's working on that right now. And maybe we're going to crack it.

Speaker 2 And, you know, if China can't steal that, then

Speaker 2 they,

Speaker 2 you know, then they're stuck. If they can't steal it, they're off to the races.

Speaker 1 But whatever that thing is. Yes.

Speaker 1 Is it like literally, I can write down on the back of a napkin? Because if it's that easy, then why is it that hard for them to figure it out?

Speaker 1 And if it's more about the intuitions, then don't you just have to hire Alec Bradford? Like, what are you copying down?

Speaker 2 Well, I think there's a few layers to this, right? So I think at the top is kind of like sort of the,

Speaker 2 you know,

Speaker 2 fundamental approach, right? And sort of like, I don't know, on pre-training, it might be, you know, like, you know, unsupervised learning, next token protection, train on the entire internet.

Speaker 2 You actually get a lot of juice out of that already.

Speaker 2 That one's very quick to communicate. Then there's like, there's a lot of details that matter.
And you were talking about this earlier, right?

Speaker 2 It's like, probably the way that thing people are going to figure out is going to be like somewhat obvious.

Speaker 2 There's going to be some kind of like clear, you know, not that complicated thing that'll work, but there's going to be a lot of details to getting that.

Speaker 1 But if that's true, then again,

Speaker 1 why are we even why do we think that getting state-level security in these stars will prevent China from catching up?

Speaker 1 If it's just like, oh, we know some sort of self-play RL will be required to get past the data wall.

Speaker 1 And if it's as easy, as you say, in this fundamental sense. I mean, again, but it's going to be solved by 2027, you say, like, right? So it's like not that hard.

Speaker 2 I just think, you know, the US and the sort of, I mean, all the leading AI labs are in the United States, and they have this huge lead. I mean, by default, China actually has some good LLMs.

Speaker 2 Why do they have good LLMs? They're just using the sort of open source code, right? You know, LLAMA or whatever. And so

Speaker 2 I think people really underrate the sort of both the sort of divergence on algorithmic progress and the lead the US would have by default.

Speaker 2 Because by, you know, all this stuff was published until recently, right? Like Chinchilla scaling laws were published. You know, there's a bunch of MOE papers.
There's, you know, transformers.

Speaker 2 And, you know, all that stuff was published. And so that's why open source is good.
That's why China can make some good models. That stuff is now, I mean, at least they're not publishing it anymore.

Speaker 2 And if we actually kept it secret, it would be this huge edge.

Speaker 2 To your point about sort of like some tacit knowledge, Alec Bradford, you know, there's another layer at the bottom that is something about like, you know, large-scale engineering work to make these big training ones work.

Speaker 2 I think that is a little bit more tacit knowledge. So I think that, but I think China will be able to figure that out.
That's like sort of engineering schlap.

Speaker 2 They're going to figure out how to do that.

Speaker 1 Oh, I can figure that out, but not how to get the RL thing working.

Speaker 2 I mean, look, I don't know.

Speaker 2 Germany during World War II, you know, they went down the wrong path. They did heavy water, and that was wrong.

Speaker 2 And there's actually, there's an amazing anecdote in the making of the atomic bomb on this, right? So secrety is actually one of the most contentious issues, you know, early on as well.

Speaker 2 And, you know, part part of it was sort of, you know, Zillard or whatever really thought this sort of nuclear chain reaction was possible.

Speaker 2 And so, an atomic bomb was possible. And he went around and he was like, this is going to be of enormous strategic importance, military importance.

Speaker 2 And a lot of people didn't believe it, or they're kind of like, well, maybe this is possible, but I'm going to act as though it's not possible.

Speaker 2 And, you know, science should be open and all these things.

Speaker 2 And anyway, so in these early days, so there had been some sort of incorrect measurements made on graphite as a moderator and that Germany had.

Speaker 2 And so they thought, you know, graphite was not going to work. We have to do heavy water.

Speaker 2 But then Fermi made some new measurements on graphite, and they indicated that graphite would work. You know, this is really important.

Speaker 2 And then, you know, Zillard kind of assaulted Fermi with another secrecy appeal. And Fermi was just kind of, he was pissed off, you know, had a temper tantrum.

Speaker 2 You know, he was like, he thought it was absurd. You know, like, come on, this is crazy.

Speaker 2 But, you know, you know, Zillard persisted. I think they roped in another guy, Pegram.
And then Fermi didn't publish it.

Speaker 2 And that was just in time, because Fermi not publishing it meant that the Nazis didn't figure out graphite would work. They went down this path of heavy water, and that was the wrong path.

Speaker 2 That was why

Speaker 2 this is a key reason why the sort of German project didn't work out. They were kind of way behind.

Speaker 2 And,

Speaker 2 you know, I think we face a similar situation on are we, are we just going to instantly leak the sort of how do we get past the data wall? What's the next paradigm? Or are we not?

Speaker 1 So, and the reason this would matter is if there's like being one year ahead would be a huge advantage.

Speaker 1 In the world where it's like you deploy AI over time, and they're just like, ah, they're going to catch up anyway. I mean, I interviewed Richard Rhodes, the guy who wrote the making

Speaker 1 an

Speaker 1 atomic bomb. Yeah.
And one of the anecdotes he had was

Speaker 1 when, so they'd realize America had the bomb. Obviously, we dropped it in Japan.
Yeah. And Beria goes, the guy who ran the NKBD,

Speaker 1 just a famously ruthless guy, just evil. And he goes to, I forgot the night name, but the guy, the Soviet scientist who was running their version of the Manhattan Project, he says,

Speaker 1 Comrade, you will get us the American bomb.

Speaker 1 And the guy says, well, listen, their implosion device actually is not optimal. We should make it a different way.

Speaker 1 And Barrier says, no, you will get us the American bomb or your family will be camp dust.

Speaker 1 But the thing that's relevant about that anecdote is actually the Soviets would have had a better bomb if they hadn't copied the American design, at least initially.

Speaker 1 And which suggests that often in history, this is something that's not just really the Manhattan Project, but there's this pattern of parallel invention where because the tech tree implies that the certain thing is next, in this case, a self-play, RL, whatever.

Speaker 1 Um,

Speaker 1 then people are just like working on that, and like people are going to figure out around the same time. There's not, there's not going to be that much gap in who gets it first.

Speaker 1 Um, was it like famously that a bunch of people were invented something like the light bulb around the same time and so forth?

Speaker 1 Yeah, so but is it just that, like, yeah, that might be true, but it'll with the one year or the six months or whatever.

Speaker 2 Two years makes all the difference. I don't know if it'll be two years, though.
Like, right? I mean, I actually, I mean, I actually think if if we lock down the labs, we have much better scientists.

Speaker 2 We're way ahead. It would be two years.
But even, I think, even, I think, I think whether you, I think, yeah, I think even six months, a year would make a huge difference.

Speaker 2 And this gets back to the sort of intelligence exploit dynamics.

Speaker 2 Like a year might be the difference between, you know, a system that's sort of like human level and a system that is like vastly superhuman, right? It might be like five

Speaker 2 ooms. You know, I mean, even on the current pace, right? We went from, you know, I think on the math benchmark recently, right?

Speaker 2 Like, you know, three years ago on the math benchmark, we, you know, that was, you know, this is sort of really difficult high school competition math problems.

Speaker 2 You know, we were at, you know, a few percent, couldn't solve anything, now it's solved.

Speaker 2 And that was sort of at the normal pace of AI progress. You didn't have sort of a billion superintelligent resources researchers.
So like a year is a huge difference.

Speaker 2 And then, particularly after super intelligence, right, once this is applied to sort of lots of elements of RD, once you get the sort of like industrial explosion with robots and so on,

Speaker 2 you know, I think a year, you know, a couple years might be kind of like decades worth of technological progress. And might, you know, again, it's like Gulf War one, right?

Speaker 2 20, 30 years of technological lead, totally decisive. You know, I think it really matters.
The other reason it really matters is,

Speaker 2 you know, suppose, suppose they steal the weight, suppose they steal the algorithms, and, you know, they're close on our tails. Suppose we still pull out ahead, right?

Speaker 2 We just kind of, we're a little bit faster, you know, we're three months ahead.

Speaker 2 I think the sort of like world in which we're really neck and neck, you know, we only have a three month lead, are incredibly dangerous, right?

Speaker 2 And we're in this like feverish struggle where like, if they get ahead, they get to dominate, you know,

Speaker 2 sort of maybe they get a decisive advantage. They're building clusters like crazy.

Speaker 2 They're willing to throw all caution to the wind. We have to keep up.
There's some crazy new WMDs popping up.

Speaker 2 And then we're going to be in the situation where it's like, you know, crazy new military technology, crazy new WMDs, you know, like deterrence, mutually disturbed instruction like keeps changing, you know, every few weeks.

Speaker 2 And it's like, you know, completely unstable, volatile situation. That is incredibly dangerous.

Speaker 2 So it's, I think, I think, you know, both, both from just the technologies are dangerous from the alignment point of view.

Speaker 2 You know, I think it might be really important during the intelligence explosion to have this sort of six-month

Speaker 2 wiggle room to be like look we're gonna like dedicate more compute to alignment during this period because we have to get it right we're feeling uneasy about how it's going and um

Speaker 2 so i think in some sense that like one of the most important inputs to whether we will kind of destroy ourselves or whether we will get through this just incredibly crazy period is whether we have that buffer um

Speaker 1 why so before we go further object level in this i think it's very much worth noting yeah that almost nobody at least nobody i talked to yeah thinks about the geopolitical implications of AI.

Speaker 1 And I think I have some object-level disagreements that we'll get into,

Speaker 1 or at least things I want to iron out. I may not disagree in the end.

Speaker 1 But

Speaker 1 the basic premise that obviously, if you keep scaling, and obviously, if people realize that this is where intelligence is headed, it's not just going to be like

Speaker 1 the same old world where, like, what model are we deploying tomorrow? And what is the latest? Like,

Speaker 1 if people on Twitter are like, oh, the GPT 4.0 is going to shake your expectations or whatever.

Speaker 1 You know, COVID is really interesting because

Speaker 1 before a year or something, when March 2020 hit,

Speaker 1 it became clear to the world, like...

Speaker 1 president, CEOs, media, average person, there's other things happening in the world right now, but the main thing we as a world are dealing with right now is COVID.

Speaker 2 Soon on AGI. Yeah.

Speaker 1 Okay. And then so.
This is the quiet period.

Speaker 2 You know, if you want to go on vacation, you know, you want to be like, you want to, yeah, you want to have, you know, maybe like now is the last time you can have some kids.

Speaker 2 You know, my girlfriend sometimes complains, you know, that I,

Speaker 2 you know, when I'm like, you know, off doing work or whatever, and she's like, I'm not spending time with her. She's like, you know,

Speaker 2 she threatens to replace me with like, you know, GB6 or whatever. And I'm like, you know, GB6 will also be too busy with doing AI research.

Speaker 1 Okay, anyway, so let's get to the question of why, why, why, why are now talking national security?

Speaker 2 I made this mistake with COVID, right?

Speaker 2 So I, you know, February of 2020, and I, um, you know, I thought just it was going to sweep the world and all the hospitals would collapse and it would be crazy and then, and then, you know, and then it'd be over.

Speaker 2 Um, and a lot of people thought this kind of the beginning of COVID. They shut down their offices a month or whatever.
I think the thing I just really didn't price in was the societal reaction, right?

Speaker 2 And within weeks, you know, Congress spent over 10% of GDP on like COVID measures, right? The entire country was shut down. That was crazy.

Speaker 2 And so, I don't know, I didn't price it in with COVID sufficiently.

Speaker 2 I don't know. why do people underrate it? I mean, I think

Speaker 2 there's a sort of way in which being kind of in the trenches actually kind of, I think,

Speaker 2 gives you a less clear picture of the trend lines. You actually have to zoom out that much, only like a few years, right?

Speaker 2 But, you know, you're in the trenches, you're like trying to get the next model to work. You know, there's always something that's hard.
You know, for example, you might underrate.

Speaker 2 algorithmic progress because you're like, ah, things are hard right now, or you know, data wall or whatever.

Speaker 2 But, you know, you zoom out just a few years and you actually try to like count up how much algorithmic progress made in the last, you know, last few years, and it's enormous. Um,

Speaker 2 but I also just don't think people think about this stuff. Like, I think smart people really underrate espionage, right?

Speaker 2 And, you know, I think part of the security issue is I think people don't realize like how intense state-level espionage can be, right?

Speaker 2 Like, you know, you know, this railway company had software that could just zero-click hack any iPhone, right?

Speaker 2 They just put in your number, and then it's just like straight download of everything, right? Like, the United States infiltrated an air gapped atomic weapons program, right? Wild, you know, like

Speaker 2 yeah, yeah, yeah. Um, you know, the you know, you know, intelligence agencies have just stockpiles of zero days.

Speaker 2 You know, when things get really hot, you know, I don't know, maybe they'll send special forces, right?

Speaker 2 To like, you know, get, go to the data center or something that's, you know, or, you know, I mean, China does this, they threaten people's families, right?

Speaker 2 And they're like, look, if you don't cooperate, if you don't give us the intel.

Speaker 2 There's a good book, you know, along the lines of the Gulag Archipelago, you know, the Inside the Aquarium, which is by a Soviet GRU defector.

Speaker 2 GRU was like military intelligence. Ilya recommended this book to me.

Speaker 2 And

Speaker 2 I think reading that, I was just kind of like shocked that I have 10 sort of state-level espionages.

Speaker 2 The whole book was about like they go to these European countries and they try to get all the technology and recruit all these people to get the technology.

Speaker 2 I mean, yeah, maybe one anecdote, you know, so when, so the spy, he, you know, this eventual defector, you know, so he's being trained, he goes to the kind of GRU spy academy.

Speaker 2 And so then to graduate from the spy academy, sort of before you're sent abroad, you kind of had to pass a test to show that you can do this.

Speaker 2 And the test was, you know, you had to, in Moscow, recruit a Soviet scientist and recruit them to give you information, sort of like you would do in the foreign country.

Speaker 2 But of course, for whomever you recruited, the penalty for giving away sort of secret information was death.

Speaker 2 And so to graduate from the Soviet spy, this GRU spy academy, you had to condemn a countryman to death.

Speaker 2 States do this stuff.

Speaker 1 I started reading the book on We as Aside in the series. Yeah.
And I was actually wondering,

Speaker 1 the fact that you use this anecdote

Speaker 1 and then you're like,

Speaker 1 a book recommended by Ilya, is this some sort of

Speaker 1 is this some sort of Easter egg?

Speaker 1 We'll leave that for an exercise for the reader.

Speaker 1 Okay, so the beatings will continue until the

Speaker 1 same time.

Speaker 1 So suppose that we live in the world in which these secrets are locked down, but China still realizes that this progress is happening in America.

Speaker 1 So, in that world,

Speaker 1 especially if they realize, and I guess it's a very interesting open question. They probably won't be locked down.
Okay, but suppose they're probably going to live in the bad world. Yeah.

Speaker 2 It's going to be really bad.

Speaker 1 Why are you so confident that they won't be locked down?

Speaker 2 I mean, I'm not confident that they won't be locked down, but I think it's just it's not happening.

Speaker 1 And so, tomorrow, the lab leaders get the message.

Speaker 1 How hard, like, what do they have to do?

Speaker 2 They get more security guards they like air gap the well you know what do they do so again i think i think basically it's you know i think people there's kind of like two two reactions there which is like it's you know we're already secure you know not um and there's um you know fatalism it's impossible i think the thing you need to do is you kind of got to stay ahead of the curve of basically how egi appells the ccp yeah right so like right now you've got to be resistant to kind of like normal economic espionage um

Speaker 2 they're not right i mean i probably wouldn't be talking about this stuff if the labs were right because i wouldn't want to wake them up more the CCP, but they're not.

Speaker 2 You know, this is like, this stuff is like really trivial for them to do right now. I mean, it's also, anyway, so they're not resistant to that.

Speaker 2 I think it would be possible for a private company to be resistant to it, right? So, you know, both of us have, you know, friends in the kind of like quantitative trading world, right?

Speaker 2 And, you know, I think actually those secrets are shaped kind of similarly where it's like, you know,

Speaker 2 you know, they've said, you know, yeah, if I got on a call for an hour with somebody from a competitor firm, I could, most of our alpha would be gone.

Speaker 2 And that's sort of like, that's the like list of details of like really how to how to make you're gonna have to worry about that pretty soon.

Speaker 1 You're going to have to worry about that pretty soon. Yeah.

Speaker 2 Well, anyway, and so all the alpha could be gone. But in fact, their alpha persists, right? And, you know, often, often for many years and decades.
And so this doesn't seem to happen.

Speaker 2 And so I think there's like, you know, I think there's a lot you could go if you went from kind of current startup security.

Speaker 2 You know, you just got to look through the window and you can look at the slides, you know, to kind of like, you know,

Speaker 2 good private sector security hedge funds, you know, the way Google treats, you know, customer data or whatever.

Speaker 2 That'd be good right now. The issue is, you you know, basically the CCP will also get more AGI build.

Speaker 2 And

Speaker 2 at some point, we're going to face kind of the full force of the Ministry of State Security.

Speaker 2 And again, you're talking about smart people underrating espionage and the sort of insane capabilities of states. I mean, this stuff is wild, right?

Speaker 2 You know, they can get like, you know, there's papers about, you know, you can find out the location of like where you are on a video game map just from sounds, right?

Speaker 2 Like states can do a lot with like electromagnetic emanations.

Speaker 2 You know, like, you know, at some point, like, you have to be working from a SCF, like your cluster needs to be air-gapped and basically be a military base.

Speaker 2 It's like, you know, you need to have, you know, intense kind of security clearance procedures for employees. You know, they have to be like, you know, all this shit is monitored.

Speaker 2 You know, they're, you know, they basically have security guards. You know, it's, you know, you, you can't use any kind of like, you know, other dependencies.
It's all got to be like intensely vetted.

Speaker 2 You know, all your hardware has to be intensely vetted.

Speaker 2 And,

Speaker 2 you know, I think basically if they actually really face the full force of state-level espionage, I don't really think this is a thing private companies can do. Both, I mean, empirically, right?

Speaker 2 Like, you know, Microsoft recently had executives' emails hacked by Russian hackers and government emails they posted hacked by government actors. But also,

Speaker 2 you know, it's basically there's just a lot of stuff that only kind of, you know, the people behind the security currences know and only they deal with.

Speaker 2 And so, you know, I think to actually kind of resist the sort of full force of espionage, you're going to need the government.

Speaker 2 Anyway, so I think basically we could, we could do it by always being ahead of the curve. I think we're just going to always be behind the curve.

Speaker 2 And I think, you know, maybe unless we get the sort of government project.

Speaker 1 Okay, so going back to the naive perspective of we're very much coming at this from there's going to be a race in the CCP, we must win.

Speaker 1 And listen, I understand like bad people are in charge of the Chinese government, like the CCP and everything.

Speaker 1 But just stepping back in a sort of galactic perspective,

Speaker 1 humanity is developing AGI. And do we want to come at this from the perspective of we need to beat China to this?

Speaker 1 Our super intelligent Jupiter brain descendants want to know who China, like China will be some like distant memory that they have.

Speaker 1 America too so shouldn't it be a more the initial approach just come to them like listen we this is super intelligence this is something like we come from a cooperative

Speaker 2 perspective why why immediately sort of rush into it from a hawkish competitive perspective I mean look I mean one thing I want to say is like a lot of the stuff I talk about in the series is you know is sort of primarily you know descriptive right and so I think that on the China stuff it's like you know yeah in some ideal world you know we we you know it's just all you know merry-go-round merry-go-round and cooperation.

Speaker 2 But again, it's sort of, I think, I think people wake up to AGI. I think the issue particularly on sort of like, can we make a deal? Can we make an international treaty?

Speaker 2 I think it really relates to sort of what is the stability of sort of international arms control agreements, right? And so we did very successful arms control on nuclear weapons in the 80s, right?

Speaker 2 And the reason it was successful is because the sort of new equilibrium was stable, right? So you take, go down from, you know, whatever, 60,000 nukes to 10,000 nukes.

Speaker 2 You know, when you have 10,000 nukes, you know, basically breakout, breakout doesn't matter that much, right? Suppose the other guy now tried to make 20,000 nukes. Well, it's like, who cares, right?

Speaker 2 You know, like, it's still mutually assured destruction. Suppose a rogue state kind of went from zero nukes to one nukes.
It's like, who cares? We still have way more nukes than you.

Speaker 2 I mean, it's still not ideal for destabilization, but it's, you know, it'd be very different if the arms control agreement had been zero nukes, right?

Speaker 2 Because if it had been zero nukes, then it's just like one rogue stake makes one nuke. The whole thing is destabilized.
Breakout is very easy.

Speaker 2 You know, your adversary state starts making nukes.

Speaker 2 And so basically, when you're going to sort of like very low levels of arms or when you're going to kind of, and you're in a sort of very dynamic technological situation,

Speaker 2 arms control is really tough because breakout is easy. You know, there's, there's, I mean, there's some other sort of stories about this in sort of like 1920s, 1930s.

Speaker 2 You know, it's like, you know, all the European states had done disarmament and Germany was kind of did this like crash program to build the Luftwaffe.

Speaker 2 And that was able to massively destabilize things because not that, you know, they were the first, they were able to like pretty easily build kind of a modern, you know, air force because the others didn't really have one.

Speaker 2 And that, you know, that really destabilized things.

Speaker 2 And so I i think the issue with agi and super intelligence is the explosiveness of it right so if you have an intelligence explosion if you're able to go from kind of agi to super intelligence if that super intelligence is decisive like either you know like a year after because you developed some crazy wmd or because you have some like you know super hacking ability that lets you you kind of um you know completely deactivate the sort of enemy arsenal um

Speaker 2 that means like suppose suppose you're trying to like put in a break you know like we both we're both gonna like cooperate and we're gonna go slower you know on the cusp of agi or whatever there is going to be such an enormous incentive to kind of race ahead, to break out.

Speaker 2 We're just going to do the intelligence explosion. If we can get three months ahead, we win.

Speaker 2 I think that makes it basically, I think any sort of arms control agreement that comes at a situation where it's close, very unstable.

Speaker 1 That's really interesting. This is very analogous to.

Speaker 1 kind of a debate I had with Rose on the podcast, where he argued for nuclear disarmament.

Speaker 1 But if some country tries to break out and starts developing nuclear weapons, the six months or whatever that you would get is enough to get international consensus and invade the country and prevent them from getting nukes.

Speaker 1 And I thought that was sort of, that's not a safe low reason. This seems really tough, yeah.

Speaker 1 But so on this, right? So like, maybe it's a bit easier because you have AGI and so like you can monitor the other person's cluster or something like that.

Speaker 1 Data centers, you can see them from space, actually. You can see the energy draw they're getting.

Speaker 1 There's a lot of things, as you were saying, there's a lot of ways to get information from an environment if you're really dedicated.

Speaker 1 And also because unlike a nukes, the data centers are nukes, you have obviously the submarines, planes, you have bunkers, mountains, whatever. You have so many different places.

Speaker 1 A data center, your 100-gigawatt data center, we can blow that shit up if you're like, we're concerned, right? Like just some cruise missile or something that's like very vulnerable to sabotage.

Speaker 2 That gets to the sort of, I mean, that gets to the sort of insane

Speaker 2 volatility of this period post-superintelligence, right?

Speaker 2 Because basically, I think, so you have the intelligence explosion, you have these like vastly superhuman things on your cluster, but you're like, you haven't done the industrial explosion yet.

Speaker 2 You don't have your robots yet, you haven't kind of, you haven't covered the desert in like robot factories yet.

Speaker 2 And that is the sort of crazy moment where, you know, say, say the United States is ahead, the CCP is somewhat behind. There's actually an enormous incentive for first strike, right?

Speaker 2 Because if they can take out your data center,

Speaker 2 they know you're about to have just this commanding, decisive lead. They know if we can just take out this data center,

Speaker 2 then we can stop it. And they might get desperate.
And,

Speaker 2 you know, so I think basically we're going to get into a position. It's actually, I think it's going to be pretty hard to defend early on.

Speaker 2 I think we're basically going to be in a position where we're protecting data centers with the threat of nuclear retaliation, which maybe sounds kind of crazy, though.

Speaker 1 Is this the inverse of the Eliezer? We got to

Speaker 1 the data centers and nuclear weapons.

Speaker 2 Nuclear deterrence for data centers.

Speaker 2 I mean, this is Berlin, in the late 50s, early 60s, both Eisenhower and Kennedy multiple times kind of made the threat of full-on nuclear war against the Soviets if they tried to encroach on West Berlin.

Speaker 2 It's sort of insane. It's kind of insane that that went well.
But basically, I think that's going to be the only option for the data centers.

Speaker 2 It's a terrible option this whole scheme is terrible right like being being being in this like neck and neck race sort of at this point is terrible and you know it's also the you know I think I have some uncertainty basically on how easy that decisive advantage will be.

Speaker 2 I'm pretty confident that if you have super intelligence, you have two years, you have the robots, you're able to get that 30-year lead. Look, then you're in this like Gulf War One situation.

Speaker 2 You have your millions or billions of like mosquito-sized drones that can just take it out. I think there's even a possibility you can kind of get a decisive advantage earlier.

Speaker 2 So, you know, there's these stories, you know, about these as well about, you know, like colonization and like the sort of 1500s where it was

Speaker 2 like a few hundred kind of span yards were able to like topple the Aztec Empire, you know, a couple, I think, a couple other empires as well. You know, each of these had a few million people.

Speaker 2 And it was not like godlike technological advantage. It was some technological advantage.
It was, I mean, it was some amount of disease. And then it was kind of like cunning strategic play.

Speaker 2 And so I think there's

Speaker 2 the possibility that even sort of early on, you know, you haven't gone through the full industrial explosion yet.

Speaker 2 You have super intelligence, but you know, you're able to kind of like manipulate the opposing generals, claim you're allying with them. Then

Speaker 2 you have sort of like some crazy new bioweapons. Maybe there's even some way to pretty easily get a paradigm that deactivates enemy nukes.
Anyway, so I think this stuff could get pretty wild.

Speaker 1 Here's what I think we should do.

Speaker 2 I really don't want this volatile period. And so a deal with China would be nice.
It's going to be really tough if you're in this unstable equilibrium.

Speaker 2 I think... Basically, we want to get in a position where it is clear that the United States, that a sort of coalition of democratic allies will win.

Speaker 2 It is clear to the United States, it is clear to China. You know, that will require having locked down the secrets.

Speaker 2 That will require having built the 100 gigawatt cluster in the United States and having done the natural gas and doing what's necessary.

Speaker 2 And then when it is clear that the Democratic coalition is well ahead, then you go to China and then you offer them a deal. And, you know, China will know they're going to win.
This is going to be,

Speaker 2 they're very scared of what's going to happen.

Speaker 2 We're going to know we're going to win, but we're also very scared of what's going to happen because we really want to avoid this kind of like breakneck, breakneck race right at the end and where things could really go awry.

Speaker 2 And,

Speaker 2 you know, and then and then so then we offer them a deal. I think there's an incentive to come to the table.
I think there's a sort of more stable arrangement you can do.

Speaker 2 It's a sort of an atoms for peace arrangement. And we're like, look, we're going to respect you.
We're not, we're not going to like, we're not going to use super intelligence against you.

Speaker 2 You can do what you want. You're going to get your, like, you're going to get your slice of the galaxy.

Speaker 2 We're going to like, we're going to benefit share with you.

Speaker 2 We're going to have some like compute agreement where it's like, there's some ratio of compute that you're allowed to have, and that's like enforced with our opposing AIs or whatever. And

Speaker 2 we're just not going to do, we're just not going to do this kind of like volatile sort of WMD arms race to the death.

Speaker 2 We're going to, and sort of, it's like a new world order that's US-led, that's sort of democratic-led, but that respects China, lets them do what they want.

Speaker 1 Okay, there's so much to

Speaker 1 there's so much there.

Speaker 1 First on the galaxies thing, I think it's just a funny anecdote. I want to kind of want to tell it.
And this, we were at an event and I'm respecting Chatham House rules here.

Speaker 1 I'm not revealing anything about it, but we're talking to somebody

Speaker 1 or Leopold was talking to somebody influential.

Speaker 1 And afterwards, that person asked the group, Leopold told me that he wants, he's not going to spend any money on consumption until he's ready to buy galaxies.

Speaker 1 And he goes, the guy goes, I honestly don't know if he meant galaxies like the brand of private plane galaxy or the physical galaxies. And there was an actual debate.

Speaker 1 Like he, he went away to the restroom, and there was an actual debate among people who are very influential about, he can't have meant galaxies. And

Speaker 1 the other people who knew you better have been like, no, he means galaxies. I mean the galaxies.
I mean the galaxies.

Speaker 1 I mean, I think it'll be interesting.

Speaker 2 I mean, yeah, I think there's a, I mean, there's two ways to buy the galaxies. One is like at some point, you know, it's like post-superintelligence.

Speaker 1 But by the way, I love, okay, so what happens is he's off the bat. I'm laughing my ass off.
I'm not even saying anything. People are like gapping this debate.
And then Leopold comes back.

Speaker 1 And the guy, somebody's like, oh, Leopold, we're having this debate about whether you meant

Speaker 1 you want to buy the galaxy or you you want to buy the other thing and leopold assumes they must mean not the friday playing the galaxy versus the actual galaxy but do you want to buy the property rights of the galaxy or actually just send out the probes right now

Speaker 1 exactly exactly

Speaker 1 oh my gosh all right back to china

Speaker 1 um

Speaker 1 there's a whole bunch of things i could ask about uh that plan about whether you're going to get credible promised uh yeah you will get some part of the galaxies where they have to care about that.

Speaker 2 AIs help you enforce stuff. Okay, sure.

Speaker 1 We'll leave that aside. That's a different rabbit hole.
The thing I want to ask is.

Speaker 2 But it has to be the thing we need. The only way this is possible is if we lock it down.
I see. If we don't lock it down, we are in this feverish struggle.

Speaker 2 Greatest peril mankind will have ever seen.

Speaker 1 So, but given the fact that during this period, instead of just taking their chances and they don't really understand how this AI governance scheme is going to work, where they're going to check, whether we actually get the galaxies.

Speaker 1 The data centers, they can't be built underground. They have to be built above ground.
Taiwan is right off the coast of us. They need the chips from there.

Speaker 1 Why aren't we just going to invade? Listen, we don't want, like, worst case scenario is they win the super intelligence, which they're on track to do anyways.

Speaker 1 Wouldn't this instigate them to either invade Taiwan or blow up the data center in Arizona or something like that?

Speaker 2 Yeah, I mean, look, I mean, you talked about the data center one, and then, you know, you probably have to like threaten nuclear retaliation to protect that. They might also just blow it up.

Speaker 2 There's also maybe ways they can do it without sort of attribution, right? Like you put.

Speaker 2 Yeah, I mean, this is, I mean, this is part, we'll talk about this later, but, you know, I think, look, I think we need to be working on the Stuxnet for the Chinese project. But the

Speaker 2 Taiwan, I mean, Taiwan, the Taiwan thing, the, you know,

Speaker 2 I talk about, you know, AGI by, you know, 27 or whatever.

Speaker 2 Do you know about the like terrible 20s?

Speaker 1 No.

Speaker 2 Okay, well, I mean, sort of in the sort of Taiwan watcher circles, people often talk about like the late 2020s as the maximum period of risk for Taiwan because sort of like, you know, military modernization cycles and basically extreme fiscal tightening on the military budget in the United States over the last decade or two

Speaker 2 has meant that sort of we're in this kind of like you know trough in in in the late 20s of like you know basically overall naval capacity and you know that's sort of when china is saying they want to be ready so it's already kind of like it's kind of pitching you know there's some sort of like you know parallel timeline there um yeah look it looks appealing to invade taiwan i mean maybe not because they you know basically remote cutoff of the chips um

Speaker 2 and so then it doesn't mean they get the chips but it just means they um

Speaker 2 you know, it's just, it's, you know, the machines are deactivated.

Speaker 2 But look, I mean, imagine if during the Cold War, you know, all of the world's uranium deposits had been in Berlin, you know, and Berlin was already, I mean, almost multiple times, it was caused nuclear war.

Speaker 2 So

Speaker 1 God help us all. Well, the Groves had a plan after the war.
Yeah. That the plan was that America would go around the world and

Speaker 1 getting the rights to every single uranium deposit because they didn't realize how much uranium there was in the world. And they thought this was the thing that was feasible.

Speaker 1 Not realizing, of course, that there's like huge deposits in the Soviet Union itself. Right, right.

Speaker 1 Okay. East Germany, too.

Speaker 2 There's always a lot of East German workers who kind of got screwed. Oh, interesting.
Got cancer.

Speaker 1 Okay, so the framing we've been talking about that we've been assuming, and I'm not sure I buy yet, is that the United States, this is our leverage, this is our data center, China is the competitor.

Speaker 1 Right now, obviously, that's not the way things are progressing. Private companies control these AIs.
They're deploying them. It's a market-based thing.

Speaker 1 Why will it be the case that it's like the United States has this leverage or is doing this thing versus China is doing this thing?

Speaker 2 Yeah, I mean, look, look, on the project, you know, I mean, there's sort of descriptive and prescriptive claims or sort of normative positive claims.

Speaker 2 I think the main thing I'm trying to say is, you know, you know, look, we're at these SF parties or whatever, and I think people talk about AGI, and they're always just talking about the private AI labs.

Speaker 2 And I think I just really want to challenge that assumption.

Speaker 2 It just seems like, seems pretty likely to me, you know, as we've talked about, for reasons we've talked about, that look, like the national security state is going to get involved.

Speaker 2 And, you know, I think there's a lot of ways this could look like, right? Is it like nationalization? Is it a public-private partnership? Is it a kind of defense contractor-like relationship?

Speaker 2 Is it a sort of government project that soaks up all the people?

Speaker 2 And so there's a spectrum there.

Speaker 2 But I think people are just vastly underrating the chances of this more or less looking like a government project.

Speaker 2 And look, I mean, look, if,

Speaker 2 you know, it's sort of like, you know, do you think, do you think like we all have literal, like, you know, when we have like literal super intelligence on our cluster, right?

Speaker 2 And it's like, you know, you have 100 billion, they're like, sorry, you have a billion like super intelligent scientists that they can like hack everything.

Speaker 2 They can like stock snit the Chinese data centers. You know, they're starting to build the robo armies.

Speaker 2 You know, you like, you really think they'll be like a private company and the government would be like, oh my God, what is going on? You know, like, yeah.

Speaker 1 Suppose there's no China.

Speaker 1 Suppose there's people like Iran, North Korea, who theoretically at some point will go to do superintelligence, but they're not on our heels or don't have the ability to be on our heels.

Speaker 1 In that world, are you advocating for the national project or do you prefer the private path forward?

Speaker 2 Yeah, so I mean, two responses to this. One is, I mean, you still have like Russia, you still have these other countries.

Speaker 2 You know, you've got to have Russia-proof security.

Speaker 2 It's like

Speaker 2 you can't just have Russia steal all your stuff.

Speaker 2 And maybe their clusters aren't going to be as big, but like they're still going to be able to make the crazy bioweapons and the, you know, the mosquito-sized drone swarm, you know, and so on. And

Speaker 2 so, I mean, I think, I think the security component is just actually a pretty large component of the project in the sense of like, I currently do not see another way where we don't kind of like instantly proliferate this to everybody.

Speaker 2 And so, yeah, so I think it's sort of like you still have to deal with Russia, you know, Iran, North Korea.

Speaker 2 And, you know, like, you know, Saudi and Iran are going to be trying to get it because they want to screw each other. And, you know, Pakistan and India because they want to screw each other.

Speaker 2 There's like this enormous destabilization still.

Speaker 2 That said, look, I agree with you. If, if, you know, if, you know, if by somehow things had shaken out differently, and like, you know, AGI would have been in 2005,

Speaker 2 you know sort of like unparalleled you know American hegemony

Speaker 2 I think there would have been more scope for

Speaker 1 less government involvement but again you know as we were talking about earlier I think that would have been sort of this like very unique moment in history and I think basically at you know almost all other moments in history there would have been this sort of great power competitor so okay so let's get into this debate so I my position here is if you look at the people who are involved in the Manhattan Project itself yeah many of them regretted their participation as you said.

Speaker 1 Now, we can infer from that that we should sort of start off with a cautious approach to the nationalized ASI project.

Speaker 1 Then you might say, well, listen, obviously

Speaker 2 did they regret their participation because of the project or because of the technology itself? I think people will regret it, but I think

Speaker 2 it's about the nature of the technology and it's not about the project.

Speaker 1 I think they also probably had a sense that different decisions would have been made if it wasn't some concerted effort that everybody had agreed to participate in.

Speaker 1 That if it wasn't in the context of this, we need to race to beat Germany and Japan, you might not develop. So, that's the technology part, but also, like, you wouldn't actually like

Speaker 2 them. You know, it's like the sort of destructive potential, the sort of you know, military potential.

Speaker 2 It's not, it's not because of the project, it is because of the technology, and that will unfold regardless. You know,

Speaker 1 I think this uh underrates the power of modeling.

Speaker 2 You imagine you go through like the 20th century in like you know, a decade. Uh, you know, it's just the sort of the sort of yes, great technology.

Speaker 1 Actually,

Speaker 1 let me just actually run to that example. Suppose you actually, there was some reason that the 20th century would be run through in one decade.
Do you think the cause of that should have been,

Speaker 1 then like the technologies that happened through the 20th century shouldn't have been privatized? That it should have been a more sort of concerted

Speaker 1 government-led project?

Speaker 2 You know, look,

Speaker 2 there is a history of just dual-use technologies, right? And so I think AI, in some sense, is going to be dual-use in the same way. And so there's going to be lots of civilian uses of it, it, right?

Speaker 2 Like nuclear energy itself, right? It was like, you know, there's the government project developed the military angle of it.

Speaker 2 And then, you know, it was like, you know, then the government worked with private companies. There was a sort of like real flourishing of nuclear energy until the environmentalists stopped it.

Speaker 2 You know,

Speaker 2 planes, right? Like Boeing, or actually, you know, the Manhattan Project wasn't the biggest defense RD project during World War II. It was the B-29 bomber, right?

Speaker 2 Because they needed the bomber that had long enough range to reach Japan

Speaker 2 to destroy their cities.

Speaker 2 And then, you know, Boeing made, so Boeing made that. Boeing made the B-47, made the B-52, you know, the plane the U.S.
military uses today. And then they use that technology later on to,

Speaker 2 you know, build the 707 and sort of the.

Speaker 1 But what does later on mean in this context? Because in the other, like, I get what it means after a war to privatize.

Speaker 1 But if you have the government has ASI, let me just let me back up and explain my concern. Yeah.
So you have the only institution in our society which has a monopoly on violence.

Speaker 1 And then we're going to

Speaker 1 give it some, in a way that's not broadly deployed, access to the ASI. The counterfactual, and this maybe sounds silly, but listen, we're going to go through higher and higher levels of intelligence.

Speaker 1 Private companies will be required by regulation to increase their security.

Speaker 1 But they'll still be private companies and they'll deploy this and they're going to release the AGI.

Speaker 1 Now, McDonald's and JP Morgan and some random startups are now more effective organizations because they have a bunch of AGI workers.

Speaker 1 And it'll be sort of like the Industrial Revolution in the sense that the benefits were widely diffused. If you don't end up in a situation like that, then the,

Speaker 1 I mean, even backing up, like, what is it we're trying to, why do we want to win against China? We want to win against China because we don't want a top-down authoritarian system to win.

Speaker 1 Now, if the way to beat that is that the most important technology that humanity will have has to be controlled by a top-down government, like, what was the point? Like, why do

Speaker 1 you, so let's like run our cards with privatization. That's the way we get to the classic liberal market-based system we want for the ESIs.
Yeah.

Speaker 2 All right. So a lot of talk about here.
Yeah. I think, yeah, maybe I'll start a bit about like actually looking at what the private world would look like.

Speaker 2 And I think this is part of where the sort of there's no alternative comes from.

Speaker 2 And then let's look at, look at like what the government project looks like, what checks and balances look like, and so on.

Speaker 1 All right, private world.

Speaker 2 I mean, first of all, okay, so right, like a lot of people right now talk about open source.

Speaker 2 And I think there's this sort of misconception that like AGI development is going to be like, oh, it's going to be some like beautiful decentralized thing.

Speaker 2 And, you know, like, you know, some giddy community of coders who gets to like, you know, collaborate on it that's not how it's going to look like right you know it's you know the hundred billion dollar trillion dollar cluster it's not going to be that many people that have it the algorithms you know it's like right now open source is kind of good because people just use the stuff that was published and so they basically you know the algorithms were published or you know as mistral they just kind of like leave deep mind and you know take all the secrets with them and they just kind of replicate it um

Speaker 2 but that's not going to continue being the case and so you know the sort of like open source will turn i mean also people say stuff like you know 1026 flops it'll be in my phone or you know it's no it won't you know it's like more's law is really slow i mean AI chips are getting better, but like, you know, the $100 billion computer will not cost

Speaker 2 $1,000

Speaker 2 within your lifetime or whatever, aside from AI. So it's going to be, it's going to be like two or three, you know, big players

Speaker 2 on the private world.

Speaker 2 And so look, a few things. So first of all,

Speaker 2 you know, you talk about the sort of like, you know, enormous power that sort of super intelligence will have and that the government will have.

Speaker 2 I think it's pretty plausible that the alternative world is that like one one AI company has that power, right? And especially if we're talking about lead, you know, it's like, what?

Speaker 2 I don't know, Open AI has a six-month lead. And then, you know, so then you're not talking, you're talking about basically, you know, the most powerful weapon ever.

Speaker 2 And it's, you know, you're kind of making this like radical bet on like a private company CEO as the benevolent dictator.

Speaker 1 No, no, not necessarily like any other thing that's privatized. We don't count on them being benevolent.
We just

Speaker 1 look, think of, for example, somebody who manufactures industrial fertilizer. Yes.
Right.

Speaker 1 This, the person with this factory, if they went back to an ancient civilization, they could like blow up Rome. They could probably blow up Washington, D.C.

Speaker 1 And I think in their series, you talk about Tyler Cowen's phrase of muddling through.

Speaker 1 And I think even with privatization, people sort of underrate that there are actually a lot of private actors who have the ability to like, there's a lot of people who control the water supply or whatever.

Speaker 1 And we can count on cooperation and market-based incentives to basically keep a balance of power. Sure.
I get that things are proceeding really fast. Yes.

Speaker 1 But we have a lot of historical evidence that this is the thing that that works best.

Speaker 2 So look,

Speaker 2 I mean, what do we do with nukes, right?

Speaker 2 The way we keep the sort of nukes in check is not like, you know, a sort of beefed up Second Amendment where like each state has their own like little nuclear arsenal and like, you know, Dario and Sam have their own little nuclear arsenal.

Speaker 2 No, no, it's like it's institutions, it's constitutions, it's laws, it's it's it's courts.

Speaker 2 And so anyway, I don't actually, I'm not sure that this, you know, I'm not sure that this sort of balance of power analogy holds.

Speaker 2 In fact, you know, sort of the government having the biggest guns was sort of like an enormous civilizational achievement, right? Like Landfrieden in the sort of Holy Roman Empire, right?

Speaker 2 You know, if somebody from the town over kind of committed a crime on you, you know, you know, you didn't kind of start a sort of a, you know, a big battle between the two towns.

Speaker 2 No, you take it to a court of the Holy Roman Empire and they would decide. And it's a big achievement.

Speaker 2 Now, the thing about, you know, the industrial fertilizer, I think the key difference is kind of speed and offense defense balance issues, right? So it's like 20th century and 10 years in a few years.

Speaker 2 That is an incredibly scary period.

Speaker 2 And it is incredibly scary because it's, you know, you're going through just this sort of enormous array of destructive technology and the sort of like enormous amount of like, you know, basically military advance.

Speaker 2 I mean, you would have gone from, you know, kind of like, you know, you know, bayonets and horses to kind of like tank armies and fighter jets in like a couple of years.

Speaker 2 And then from, you know, like, you know, and then to like, you know, nukes and, you know, ICBMs and stuff, you know, just like in a matter of years. And so

Speaker 2 it is sort of that speed that creates, I think basically the way I think about it is there's going to be this initial just incredibly volatile, incredibly dangerous period.

Speaker 2 And somehow we have to make it through that. And that's going to be incredibly challenging.

Speaker 2 That's where you need the kind of government project. If you can make it through that, then you kind of go to like, you know, now we can, now, you know, the situation has been stabilized.

Speaker 2 You know, we don't face this imminent national security threat.

Speaker 2 You know, it's like, yes, there were kind of WMDs that came along the way, but either we've managed to kind of like have a sort of stable offense-defense balance, right?

Speaker 2 Like, I think bioweapons initially are a huge issue, right? Like, an attacker can just create like a thousand different synthetic, you know, viruses and spread them.

Speaker 2 And it's like going to be really hard for you to kind of like make a defense against each.

Speaker 2 But maybe at some point, you figure out the kind of like, you know, universal defense against every possible virus. And then you're in a stable situation again on the offense-defense balance.

Speaker 2 Or you do the thing you do with planes, where it's there's like, you know, there's certain capabilities that the private sector isn't allowed to have, and you've like figured out what's going on, restrict those, and then you can kind of like let you know, you can let the sort of civilian civilian uses.

Speaker 1 So I'm skeptical of this because,

Speaker 2 well, there's sorry, I mean, the other important thing is, so I talked about the sort of, you know, maybe it's like, it's, it's a, you know, it's, you know, it's one company with all this power.

Speaker 2 And I think it is like, I think it is unprecedented because it's like the industrial fertilizer guy cannot overthrow the U.S. government.

Speaker 2 I think it is quite plausible that like the AI company with super intelligence can overthrow the U.S.

Speaker 1 But there'll be multiple AI companies, right? And I buy that one of them could be.

Speaker 2 So it's not obvious that it'll be multiple. I think it's, again, if there's like a six-month lead, maybe, maybe there's two or three.

Speaker 1 I agree.

Speaker 2 But if there's two or three, then what you have is just like a crazy race between these two or three companies. You know, it's like, you know, whatever.

Speaker 2 Demis and Sam, they're just like, I don't want to let the other one win. And they're both developing their nuclear arsenals and the robot.

Speaker 2 It's just like, also, like, come on, the government is not going to let these people, you know, are they going to let, like, you know, is Dario going to be order, the one developing the kind of like, you know, you know, super hacking Stuxnet and like deploying against the Chinese data center?

Speaker 2 The other issue, though, is it won't just, if it's two or three, it won't just be two or three.

Speaker 2 It'll be two or three, and it'll be China and Russia and North Korea because the private in the private lab world, there is no way they will have security that is good enough.

Speaker 1 I think we're also assuming that somehow, if you nationalize it, like the security, just especially in the world where

Speaker 1 this stuff is priced in by the CCP, that now you've like got it nailed down. And I'm not sure why we would expect that to be the case.

Speaker 1 But on the government's the only one who does this stuff, so if it's not Sam or Dario, who's we don't trust them to be benevolent dictator or whatever,

Speaker 1 so but here we're counting on

Speaker 1 if it's because you can cause a coup, the same capabilities are going to be true of the government project, right?

Speaker 1 And so the modal president in 2020, 2025, but Donald Trump will be the person that you don't trust Sam or Dario to have these capabilities.

Speaker 1 And why, okay, I agree that like I'm worried if Sam or Dario have a one-year lead on ASI in that world, then I'm like concerned about this being privatized.

Speaker 1 But in that exact same world, I'm very concerned about Donald Trump having the capability.

Speaker 1 And potentially, if we're living in a world where the takeoff is slower than you anticipate, in that world, I'm like, very much, I want the private company.

Speaker 1 So like in no part of this matrix, this is obviously true that the government-led project is better than the private project.

Speaker 2 Let's talk about the government project a little bit and checks and balances.

Speaker 2 In some sense, I think my argument is a sort of Berkeley argument, which is like American checks and balances have held for, you know, over 200 years and through crazy technological revolutions.

Speaker 2 You know, the U.S. military could kill like every civilian in the United States.

Speaker 1 But if you're going to make that argument, the private public balance of power has held for hundreds of years.

Speaker 2 But yeah, why has it held? Because the government has the biggest guns. And as never before has a single CEO or a random nonprofit board had the ability to launch nukes.

Speaker 2 And so again, it's like, you know, what is the track record of the government checks and balances versus the track record of the private company checks and balances?

Speaker 2 Well, the iLab, you know, like first stress test, you know, went really badly. You know, that didn't really work, you know?

Speaker 2 I mean, even worse in the sort of private company world.

Speaker 2 So it's both like, it is not just the two, it is, it is like the two private companies and the CCP, and they just like instantly have all the shit.

Speaker 2 And then it's, you know, they probably won't have good enough internal controls.

Speaker 2 So it's like, it's not just like the random CEO, but it's like, you know, rogue employees that can kind of like use these super intelligences to do whatever they want.

Speaker 1 And this won't be true of the government? Like the rogue employees won't exist on the project?

Speaker 2 Well, the government actually like, you know, has decades of experience and like actually really cares about this stuff.

Speaker 2 I mean, it's like they deal, they deal with nukes, they deal with really powerful technology. And it's, you know, this is like, this is the stuff that the national security state cares about.

Speaker 2 You know, again, to the government, let's talk about the government checks and balances a little bit. So, you know, what are what are checks and balances in the government world?

Speaker 2 First of all, I think it's actually quite important that you have some amount of international coalition. And I talked about these sort of two tiers before.

Speaker 2 Basically, I think the inner tier is a sort of modeled on the Quebec Agreement, right? This was like Churchill and Roosevelt. They kind of agreed secretly,

Speaker 2 we're going to pull our efforts on nukes, but we're not going to use them against each other, and we're not going to use them against anyone else with their consent. And I think basically, look,

Speaker 2 bring in the UK, they have DeepMind, bring in the kind of like Southeast Asian states who have the chip supply chain, bring in some more kind of like NATO, close democratic allies for talent and industrial resources.

Speaker 2 And you have this sort of like, you know, so you have, you have those checks and balances in terms of like more international countries at the table.

Speaker 2 Sorry, somewhat separately, but then you have the sort of second tier of coalitions, which is the sort of Adams for Peace thing, where you go to a bunch of countries, including the UAE, and you're like, look, we're going to basically like, you know, there's a deal similar to like the NPT stuff where it's like, you're not allowed to like do the crazy military stuff, but we're going to share the civilian applications.

Speaker 2 We're in fact going to help you

Speaker 2 and share the benefits and, you know, sort of kind of like this new sort of post-superintelligence world order. All right, U.S.
checks and balances, right?

Speaker 2 So obviously Congress is going to have to be involved, right? Appropriate trillions of dollars. I think probably ideally you have Congress needs to kind of like confirm whoever's running this.

Speaker 2 So you have Congress, you have like different factions of the government, you have the courts. I expect the First Amendment to continue being really important.

Speaker 2 And maybe that, I think that sounds kind of crazy to people, but I actually think, again, I think these are like institutions that have withheld the test of time in a really sort of powerful way.

Speaker 2 You know, eventually, you know, this is why, honestly, alignment is important.

Speaker 2 The AIs, you program the AIs to follow the Constitution. And it's like, you know, why does the military work? It's like generals, you know,

Speaker 2 are not allowed to follow unlawful orders. They're not allowed to follow unconstitutional orders.
You have the same thing for the AIs.

Speaker 1 So, what's wrong with this argument? Where you say, listen, maybe you have a point in the world where we have extremely fast takeoff. It's like one year from AGI to ASI.

Speaker 1 And then you have the like

Speaker 2 10 years after of ASI where you have this like extraordinary.

Speaker 1 Maybe you have a point. We don't know.

Speaker 1 You have these arguments. We'll get into the weeds on them about why that's a more likely world, but maybe that's not the world we live in.

Speaker 1 And in the other world, I'm like very on the side of making sure that these things are privately held. Now, why when you nationalize?

Speaker 1 So when you nationalize, that's a one-way function. You can't go back.
Why not wait until we have more evidence on which of those worlds we live in? Why?

Speaker 1 And I think like rushing on the nationalization might be a bad idea while we're not sure. And okay, I'll just respond to that first.

Speaker 2 I mean, I don't expect us to nationalize tomorrow. If anything, I expect it to be kind of with COVID where it's like kind of too late.

Speaker 2 Like ideally, you nationalize it early enough to like actually lock stuff down. It'll probably be kind of chaotic.

Speaker 2 And like, you know, you're going to be trying to do this crash program to lock stuff down. And it'll be kind of late.
It'll be kind of clear what's happening.

Speaker 2 We're not going to nationalize when it's not clear what's happening.

Speaker 1 I think the whole battle, the whole historically institutions have held up well. First of all, they've actually almost broken a bunch of times.

Speaker 1 It didn't break the first round.

Speaker 1 This is similar to the argument that some people who are, say, that we shouldn't be that concerned about nuclear war, say, where it's like, listen, we have the nuke for 80 years and we've been fine so far.

Speaker 1 So the risk must be low. And then the answer to that is no.
Actually, it is a really high risk.

Speaker 1 And the reason we've avoided it is like people have gone through a lot of effort to make sure that this thing doesn't happen.

Speaker 1 I don't think that giving government ASI without knowing what that implies is going through a lot of effort.

Speaker 1 And I think the base rate, like you can talk about America, I think America is very exceptional, not just in terms of dictatorship, but in terms of every other country in history has had a complete drawdown of wealth because of war, revolution, and something.

Speaker 1 America is very unique in not having that. And the historical base rate, we're talking about Greek power competition.

Speaker 1 I think that has a really big, that's something we haven't been thinking about the last 80 years, but is really big.

Speaker 1 Dictatorship is also something that is just the default state of mankind.

Speaker 1 And I think relying on institutions, which in an ASI world, like there's a, it's fundamentally right now, if the government tried to overthrow, there's a, it's much harder if you don't have the ASI, right?

Speaker 1 Like there's people who have

Speaker 1 AR-415s and there's like things that make it harder.

Speaker 1 No, I think it actually be pretty hard. The reason is Vietnam and Afghanistan are pretty hard.

Speaker 2 It's a whole country.

Speaker 1 Yeah, yeah, I agree. But like, I'm

Speaker 2 similar with ASI.

Speaker 1 Yeah, I think it's just like easier if you have what you were talking about.

Speaker 2 With constitutions, there are legal restraints, there are courts, there are checks and balances. The crazy bet is the bet which you're like private companies see you.

Speaker 1 The same thing, by the way, isn't the same thing true of nukes where we have these institutional agreements about non-proliferation and whatever.

Speaker 1 And we're still very concerned about that being broken and somebody getting nukes and like you should stay up that night working about it.

Speaker 2 It's a precarious situation. But ASI is going to be a really precarious situation as well.
And like given, given how precarious nukes are, we've done pretty well.

Speaker 1 And so what does privatization in this world even mean? I mean, I think the other thing is like what happens after?

Speaker 2 I mean, the other thing, you know, because we're talking about like whether the government project is good or not. And it's like, I have very mixed feelings about this as well.

Speaker 2 Again, I think my primary argument is like,

Speaker 2 you know, if you're at the point where this thing has like

Speaker 2 vastly superhuman hacking capabilities, if you're at the point where this thing can develop, you know, bioweapons, you know, like in crazy bioweapons, ones that are like targeted, you know, can kill everybody but the Han Chinese or that, you know,

Speaker 2 would wipe out entire countries, where you're talking about like building robo armies, you're talking about kind of like drone swarms that are, you know, again, the mosquito-sized drones that could take it out, you know.

Speaker 2 The United States national security state is going to be intimately involved with this.

Speaker 2 And this will, you know, the labs, whether, you know, and I think, again, the government, a lot of what I think is the government project looks like, it is basically a joint venture between like, you know, the cloud providers, between some of the labs and the government.

Speaker 2 And so I think there is no world in which the government isn't intimately involved in this like crazy period.

Speaker 2 The very least, basically, you know, like the intelligence agencies need to be running security for these labs.

Speaker 2 So they're already kind of like, they're controlling everything, they're controlling access to everything.

Speaker 2 Then they're going to be like, probably, again, if we're in this really volatile international situation, like a lot of the initial applications, it'll suck. It's not what I want to use ASI for.

Speaker 2 Will be like trying to somehow stabilize this crazy situation.

Speaker 2 Somehow, we need to prevent proliferation of some crazy new WMDs and like the undermining of mutually assured destruction to kind of like, you know, North Korea and Russia and China.

Speaker 2 And so

Speaker 2 I think, you know, I basically think your world,

Speaker 2 you know, I think there's much more spectrum than you're acknowledging here. And I think basically the world in which it's private labs is like extremely heavy government involvement.

Speaker 2 And really what we're debating is like, you know, what form of government project, but it is going to look much more like, you know, the national security state than anything it does look like, like a startup as it is right now.

Speaker 2 And I think the, yeah.

Speaker 1 look i think something like that makes sense i would be if it's like the manhattan project then i'm very worried where it's like this is part of the u.s military um

Speaker 1 where if it's more like listen you got to talk to jake sullivan before you like run the next training line like lockheed martin skuncourt's part of the u.s military it's like they call the shots yeah i don't think that's great i think that's i think that's bad i think it would be bad if that happened with a like what is it what is the scenario what is why is it raised what is the what is the alternative okay so it's a it's closer to my end of the spectrum where yeah, you do have to talk to JX Sullivan before you can launch the next training cluster.

Speaker 1 But there's many companies who are still going for it.

Speaker 1 And the government will be intimately involved in the security.

Speaker 1 But three different companies are.

Speaker 2 I'm watching the Stuxnet attack.

Speaker 1 Yeah.

Speaker 1 Launching. Okay.

Speaker 1 So they're launching. Darius deactivating the Chinese data centers.
I think this is similar to the story you could tell about there's a lot of companies, like literally big tech right now.

Speaker 1 I think Sasha, if you wanted to, he probably could get his engineers, like, what are the zero days in Windows and the companies

Speaker 1 and like, well, how do we get infiltrate the president's computer so that like we can be shut down? No, no, no. But like right now, I'm saying Sasha could do that, right?

Speaker 1 Because he knows how to shut down. What do you mean?

Speaker 2 Government wouldn't let them do that.

Speaker 1 Yeah, I think there's a story you could tell where like they could pull off a coup or whatever. But like, I think there's like multiple AI companies.

Speaker 1 Okay,

Speaker 2 come on.

Speaker 1 Okay, fine, fine, fine. I agree.
I'm just saying like something closer to so what's wrong with the scenario where

Speaker 1 you have the government is there's like multiple companies going for it,

Speaker 1 but the AI is still broadly deployed and alignment works in the sense that you can make sure that it's not you the system level prompt is like you can't help people make bioweapons or something.

Speaker 1 But these are still broadly deployed.

Speaker 2 So that I mean, I expect AIs to be broadly deployed.

Speaker 2 I mean, first of all, even if it's a government project, yeah, I mean, look, I think, first of all, like, I think the metas of the world, you know, open sourcing their eyes, you know, that are two years behind or whatever, yeah, super valuable role.

Speaker 2 They're going to like, you know, and so there's going to be some question of like either the offense-defense balance is fine.

Speaker 2 And so like, even if they open source two-year-old AIs, it's fine, or it's like there's some restrictions on the most extreme dual-use capabilities, like, you know, you don't let private companies sell kind of crazy weapons.

Speaker 2 And that's great. And that will help with the diffusion.
And, you know, you know, after the government project, you know, there's going to be this initial tense period. Hopefully that's stabilized.

Speaker 2 And then, look, yeah, like Boeing, they're going to go out and they're going to like make

Speaker 2 do all the flourishing civilian applications. And, you know, like nuclear energy, you know, it'll like all the civilian applications will have their day.

Speaker 2 I think part of my argument here is that how does that proceed, right?

Speaker 1 Because in the other world, there's existing stocks of capital that are worth a lot of the time.

Speaker 2 Yeah, the clusters, there'll still be Google clusters.

Speaker 1 And so Google, because they got the contract from the government, they'll be the ones that control the ASI. But like, why are they trading with anybody else?

Speaker 1 When is this a random startup?

Speaker 2 It'll be the same companies that would be doing it anyway. But in this world, they're just contracting with the government or like their DPA'd for all their compute goes to the government.

Speaker 1 But

Speaker 1 it's very natural. So

Speaker 1 after you get the ASI and we're building the robot armies and building fusion reactors or whatever,

Speaker 1 that's only the government will get to build robot armies. Yeah, now I'm worried.
Or like the fusion reactors and stuff. That's what we do with me.
It's the same situation we have today.

Speaker 1 Because if you already have the robot armies and everything, the existing society doesn't have some leverage where it makes sense for the government to.

Speaker 1 Yeah, they gave me the sense that there's like they have a lot of capital that the government wants and there's other things. Why was Boeing privatized?

Speaker 2 It's the biggest guns. Government has the biggest guns.
And the way we regulate it is institutions, constitutions, legal restraints.

Speaker 1 Okay, so tell me what privatization looks like in the ASI world afterwards. Afterwards.

Speaker 2 Like the Boeing example, right? It's like you have this government.

Speaker 1 Who gets it? Like,

Speaker 1 Google, Microsoft, and who are they selling it to? They already have the robot factory. And then like, why are they selling it to us? Like, they already have the, they don't need like our,

Speaker 1 this is chum change in the ASI world because we didn't get like the

Speaker 1 ASI broadly deployed throughout this takeoff. So we don't have the robot.
Like, we don't have like the fusion reactors and whatever advanced decades of advanced science that you're talking about.

Speaker 1 So like, just what are they trading with us for?

Speaker 2 Trading with whom for?

Speaker 1 Everybody who was not part of the project.

Speaker 2 They got the technology that's decades ahead yeah i mean look that's a whole nother issue of like well how does like economic distribution work or whatever i don't know that'll be rough yeah i think i'm just saying

Speaker 2 i don't basically i'm kind of like i don't see the alternative the alternative is you like overturn a 500 year civilizational achievement of landfluiden you

Speaker 2 basically instantly leak the stuff to the ccp and either you like barely scrape out ahead um and but you're in this feverish struggle you're like proliferating crazy wmds it's this like enormously dangerous situation enormously dangerous on alignment because you're in this kind of like crazy race at the end and you don't have the ability to like take six months to get alignment right.

Speaker 2 The alternative is

Speaker 2 you know alternative is like you aren't actually bundling your efforts to kind of like win the race against the authoritarian powers.

Speaker 2 You know, yeah, and so

Speaker 2 you know

Speaker 2 I don't like it. You know, I wish, I wish the thing we use the ASI for is to like, you know, cure the diseases and do all the good in the world, but it is my prediction that sort of like

Speaker 2 by the in the end game, game

Speaker 2 What will be at stake will not just be kind of cool products, but what will be at stake is like whether liberal democracy survives like whether the CCP survives like what the world order for the next century will be and when that is at stake forces will be activated that are sort of way beyond what we're talking about now and like you know in in the sort of like crazy race at the end like the sort of national security implications will be the most important you know sort of like you know world war ii it's like yeah you know nuclear energy had its day, but in the initial kind of period, when

Speaker 2 this technology was first discovered, you had to stabilize the situation, you had to get nukes, you had to do it right.

Speaker 2 And then the civilian applications had their day.

Speaker 1 I think a closer analogy to what this is, because nuclear, I agree that nuclear energy is a thing that happens later on, and it's like dual use in that way, but it's something that happened like literally a decade after nuclear weapons were developed.

Speaker 1 Yeah, because whereas with AI, like the immediately all the applications are unlocked. And it's closer to literally, I mean, this is an analogy people explicitly make in the context of AGI: AGI:

Speaker 1 assume your society had 100 million more John Von Neumans.

Speaker 1 And I don't think, like, if that was literally what happened, if tomorrow you just have 100 million more of them, the approach would have been, well, some of them will convert to ISIS, and we need to be really careful about that.

Speaker 1 And then, like, oh, you know, like, what if a bunch of them are born in China? And then we like, if we get to nationalize the John Von Neumans, I'm like, no, I think it'll be generally a good thing.

Speaker 1 And I'd be concerned about one power getting like all the John Von Neumans.

Speaker 2 I mean, I think the issue is the sort of like bottling up in the sort of intensely short period of time, like this enormous sort of like, you know,

Speaker 2 unfolding of technological progress of an industrial explosion. I mean, I think we do worry about the 100 million John Monnians.
And it's like rise of China.

Speaker 2 Why are we worried about the rise of China? Because it's like 100 billion people and they're able to do a lot of industry and do a lot of technology.

Speaker 2 And but it's just like, you know, the rise of China times like, you know, 100, because it's not just 1 billion people. It's like a billion super intelligent, crazy, you know, crazy things.

Speaker 2 And in like, you know, a very short period.

Speaker 1 Let's talk practically because if the the goal is we need to beat China, part of that is protecting.

Speaker 2 I mean, that's one of the goals, right?

Speaker 1 Yeah, I agree. One of the goals is to beat China.

Speaker 2 And also just manage this incredibly crazy, scary period. Right.

Speaker 1 So part of that is making sure we're not leaking algorithmic secrets to them. Yep.
Part of that is. The trillion-dollar cluster.

Speaker 2 I mean, building the trillion-dollar cluster.

Speaker 1 That's right, right? Yeah, but like

Speaker 1 your whole point that Microsoft can release corporate bonds that are.

Speaker 2 I think Microsoft can do the hundreds of billions of dollars cluster. I think the trillion-dollar cluster is closer to a national effort.

Speaker 1 I thought your earlier point was that American capital markets are DE bending. They're good.

Speaker 2 They're pretty good. I mean,

Speaker 2 I think it's possible it's private. It's possible.

Speaker 2 But it's going to be like, you know.

Speaker 1 By the way, at this point, we have AGI that's rapidly accelerating productivity.

Speaker 2 I think the trillion-dollar cluster is going to be planned before, before the AGI.

Speaker 2 I think it's sort of like you get the AGI on the 10-gigawatt cluster.

Speaker 2 Maybe you have one more year where you're kind of doing some final unhobbling to fully unlock it. Then you have the intelligence explosion.

Speaker 2 And meanwhile, the trillion-dollar cluster is almost finished. And then you do your super intelligence on your trillion-dollar cluster, or you run it on your trillion-dollar cluster.

Speaker 2 And by the way you have not just your trillion dollar cluster but like you know hundreds of millions of gpus on inference clusters everywhere and this isn't result like i i i think private

Speaker 1 historians i think private companies have the capital and can raise capital thing you will need the government force to do it fast you know

Speaker 1 i was just about to ask like wouldn't it be the like we know private companies are on track to be able to do this and be china if they're unhindered by yeah um climate pledges or whatever well that's part of what i'm saying so i i if that's the case and if it be so if it really matters that we be china yeah there's all kind of gonna be all kinds of practical difficulties of like, will the AI researchers actually join the AI effort?

Speaker 1 If they do, there's going to be three different teams at least who are currently doing pre-training on different

Speaker 1 companies. Now, who decides at some point, you're going to have

Speaker 1 YOLO the hyperparameters of the trillion-dollar cluster.

Speaker 1 Who decides that? Just like merging extremely complicated research and development processes across very different organizations.

Speaker 1 This is somehow supposed to speed up America against the Chinese? Why don't we just let Brain and DeepMind merge?

Speaker 2 And it was like a little messy.

Speaker 1 It was pretty messy. And it was also the same company and also much earlier on in the process.

Speaker 2 Pretty similar, right? Same code, different code bases and lots of different infrastructure and different teams.

Speaker 2 And it was like, you know, it wasn't, it wasn't like, it wasn't the smoothest of all processes, but you know, DeepMind is doing, I think, very well.

Speaker 1 I mean, look, you give the example of COVID and the COVID example was like, listen, we woke up to it. Maybe it was late, but then we had deployed all this money.

Speaker 1 And COVID response to government was a cluster fuck over. And like, the only part of it that was worked is, I agree, Warp Speed was like enabled by the government.

Speaker 1 It was literally just giving the permission that you can actually do it.

Speaker 2 It was also taking, making like the biggest advanced market commitments or whatever.

Speaker 1 But I agree, but it was like fundamentally a private sector-led effort. Yeah.
That was the only part of COVID that worked.

Speaker 2 I mean, I think, I think, again, I think the project will look closer to Operation Warp Speed.

Speaker 2 And it's not even, I mean, I think, I think you'll have all the companies involved in the government project. I'm not that sold that merging is that difficult.
You know, you have one, okay.

Speaker 2 You select one code base and, you know, you, you run free training on like GPUs with, you know, one code base, and then you do the sort of second RL step on the other code base with TPUs.

Speaker 2 I don't know. I think it's fine.

Speaker 2 I mean, to the topic of like, will people sign up for it? They wouldn't sign up for it today. I think this would be kind of crazy to people.

Speaker 2 But also, you know, I mean, this is part of the like secrets thing, you know, people gather at parties or whatever. You know, you know this.

Speaker 2 You know, I don't think anyone has really gotten up in front of these people and been like, look, you know, the thing you're building is.

Speaker 2 the most important thing for like the national security of the United States for like whether you know like you know the free world will have another another century ahead of it.

Speaker 2 Like, this is this thing you're doing is really important, like, for your country, for democracy.

Speaker 2 And, you know, don't talk about the secrets. And it's not just about, you know, oh, deep mind or whatever.
It's about, it's about, you know, these really important things.

Speaker 2 And so, you know, I don't know, like, again, we're talking about the Manhattan Project, right? This stuff was really contentious initially.

Speaker 2 But, you know, at some point, it was like clear that this stuff was coming. It was clear that there was like sort of a real sort of like exigency on the military and national security front.
And,

Speaker 2 you know, I think a lot of people will come around. On the like whether it'll be competent.
I agree. I mean, this is again where it's like a lot of the stuff is more like predictive in the sense.

Speaker 2 I think this is like reasonably likely. And I think not enough people are thinking about it.
You know, like a lot of people think about like AI lab politics or whatever.

Speaker 2 But like nobody has a plan for the project.

Speaker 1 You know, it's like, you know, like should they think you're pessimistic about it? And like, we don't have a plan for it. We need to do it very soon because AGI is upon us.
Yeah.

Speaker 1 Then fuck the only capable, competent technical institutions capable of making AI right now are private companies. And those are going to play that leading role.

Speaker 2 It'll be a sort of a partnership, basically.

Speaker 2 But the other thing is like, you know, again, we talked about World War II and American unpreparedness, the beginning of World War II is complete, you know, complete shambles, right?

Speaker 2 And so there is a sort of like very competitive, I think America has a very deep bench of just like incredibly competent managerial talent. You know, I think that, you know, there's a lot of

Speaker 2 really dedicated people. And, you know, I think basically a sort of operational warp speed, public-private partnership, something like that, you know, is sort of what I imagine it would look like.

Speaker 1 Yeah, I mean, the recruiting the talent is an interesting question because the same sort of thing where

Speaker 1 initially for the Manhattan Project, you had to convince people we've got to beat the Nazis and you got to get on board. I think a lot of them maybe regretted how much they accelerated the bomb.
And

Speaker 1 I think this is generally a thing of war

Speaker 1 where...

Speaker 2 I mean, I think they're also wrong to regret it, but.

Speaker 1 Yeah,

Speaker 1 why?

Speaker 2 What's the reason for regretting it?

Speaker 1 I think there's a world in which you don't have

Speaker 1 the way in which nuclear weapons were were developed after the war was pretty explosive because there was a precedent that you actually can use nuclear weapons.

Speaker 1 Then, because of the race that was set up, you immediately go to the H-bomb.

Speaker 2 I mean, I think my view is, again, this is related to the view on AI and maybe some of our disagreement is like, that was inevitable.

Speaker 2 Like, of course, like, you know, there's this, you know, world war. And then obviously there was the, you know, the Cold War right after.

Speaker 2 Of course, like, you know, the military and technology angle of this would be like, you know, pursued with ferocious intensity.

Speaker 2 And I don't really think there's a world in which that doesn't happen where it's like ah we're all not going to build nukes and also just like nukes went really well i think that could have gone terribly right you know like in you know uh again i mean this sort of i think this is like not physically possible with nukes the sort of pocket nukes for everybody but i think sort of like wmds that are sort of proliferated and democratized and like all the countries have it like the U.S.

Speaker 2 leading on nukes and then sort of like building this new world order that was kind of US-led or at least sort of like a few great powers and a non-proliferation regime for nukes, a partnership and a deal that's like, look, no military sort of application of nuclear technology, but we're going to help you with the civilian technology.

Speaker 2 We're going to enforce safety norms on the rest of the world. That worked.
It worked. And it could have gone so much worse.

Speaker 1 Okay, so we're zooming out.

Speaker 2 And Russia and Nagasaki, you know, they were, I mean, this is, I mean, I say this a bit in the piece, but it's like, actually, the A-bomb, you know, like the A bomb and Hiroshima and Nagasaki was just like, you know, the firebombing, yeah.

Speaker 2 Firebombing.

Speaker 2 I think the thing that really changed the game was like the super, you know, the

Speaker 2 H-bombs and ICBMs. And then I think that's really when it took it to like a whole new level.

Speaker 1 I think part of me thinks when you say we

Speaker 1 will tell the people that for the free world to survive, we need to pursue this project. It sounds similar to World War II is

Speaker 1 so World War II is a sad story, obviously, in the fact that it happened, but also like the victory is sad in the sense that

Speaker 1 Britain goes in to protect Poland.

Speaker 1 And at the end, the USSR, which is,

Speaker 1 you know, as your family knows,

Speaker 1 is incredibly brutal, ends up occupying half of Europe.

Speaker 1 And

Speaker 1 part of like,

Speaker 1 we're protecting the free world, that's why we got to rush the AI.

Speaker 1 And like, if we end up with the American AI Leviathan, I think there's a world where we look back on this, where it has the same sort of twisted irony that Britain going into World War II had about trying to protect Poland.

Speaker 2 Look, I mean, I think there's going to be a lot of unfortunate things that happen. I'm just like, I'm just hoping we make it through.

Speaker 2 I mean, to the point of it's like, I really don't think the pitch will only be the sort of like, you know, the race. I think the race will be sort of a backdrop to it.

Speaker 2 I think the sort of general, like, look, it's important that democracy shape this technology. We can't just like leak this stuff to, you know, North Korea is going to be important.

Speaker 2 I think also for the just safety, including alignment, including the sort of like creation of new WMDs.

Speaker 2 I'm not currently sold there's another path, right?

Speaker 2 So it's like, if you just have the breakneck race, both internationally, because you're just instantly leaking all the stuff, including the weights, and just, you know, the commercial race, you know, Demis and Dario and Sam, you know, just just kind of like, they all want to be first.

Speaker 2 I think that's incredibly rough for safety. And then you say, okay, safety regulation.
But, you know, it's sort of like the safety regulation that people talk about.

Speaker 2 It's like, oh, well, NIST, and they take years and they figure out what the expert consensus is. And then they're going to be able to do that.

Speaker 1 Isn't that what's going to happen to the project as well? But

Speaker 2 I think the sort of alignment angle during the intelligence explosion,

Speaker 2 it's not a process of like years of bureaucracy and then you can kind of write some standards. I think it looks much more like basically a war and like you have a fog of war.

Speaker 2 It's like, look, it's like, is it safe to do the next oom? You know, know, and it's like, ah, you know, like, you know, we're like three ooms into the intelligence explosion.

Speaker 2 We don't really understand what's going on anymore. Um, you know, the,

Speaker 2 you know, like a bunch of our like generalization scaling curves are like kind of looking not great.

Speaker 2 You know, some of our like automated AI researchers that are doing alignment are saying it's fine, but we don't quite trust them.

Speaker 2 In this test, you know, the like the AI started doing naughty things and ah, but then we like hammered it out and then it was fine. And like, ah, should we, should we go ahead?

Speaker 2 Should we take, you know, another six months? Also, by the way, you know, like China just stole the weights or we, you know, they're about to deploy the Roma army. What do we do?

Speaker 2 I think it's this, I think it is this crazy situation.

Speaker 2 And

Speaker 2 basically, you're relying much more on kind of like a sane chain of command than you are on sort of some deliberative regulatory scheme. I wish you were able to do the deliberative regulatory scheme.

Speaker 2 And this is the thing about the private companies, too. I don't think

Speaker 2 they all claim they're going to do safety, but

Speaker 2 I think it's really rough when you're in the commercial race and they're startups, you know, and startups, startups are startups. You know, I think they're not fit to handle WMDs.

Speaker 1 Yeah, I'm coming closer to your position,

Speaker 1 but part of me also,

Speaker 1 so with the responsible scaling policies, I was told by people who are advancing that the way to think about this, because they know I'm like a libertarian type of person. Yeah, yeah, yeah.

Speaker 1 And the way they approached me about it was

Speaker 1 that

Speaker 1 fundamentally, this is a way to protect market-based development of AGI in the sense that if you didn't have this at all, then you would have the sort of misuse and then you would have to be nationalized.

Speaker 1 And the RSPs are a way to make sure that through this deployment, you can still have a market-based order, but then there's these safeguards that make sure that things don't go off the rails.

Speaker 1 And I wonder if

Speaker 1 it seems like your story seems self-consistent, but it does feel, I know this was never your position, so I'm not like, I'm not looping you into this, but

Speaker 1 a sort of Martin Bailey almost in the sense of well, look, here's what I think about RSP type stuff or sort of safety regulation that's happening now.

Speaker 2 I think they're important for helping us figure out what world we're in and like flashing the warning signs when we're coast, right? And so

Speaker 2 the story we've been telling is sort of like, you know, sort of what I think the modal version of this decade is. But it's like, I think there's lots of ways it could be wrong.

Speaker 2 I really, you know, we should talk about the data wall more. I think there's like, again, I think there's a world where the stuff stagnates, right? There's a world where we don't have AGI.

Speaker 2 And so I basically, you know, the RSP thing is like preserving the optionality, let's see how the stuff goes. But like, we need to be prepared.

Speaker 2 Like, if the red lights start flashing, if we're getting the automated eye researcher, then it's like, then it's crunch time, and then it's time to go.

Speaker 1 I think, okay, I can be on the same page on that, that we should have a very, very strong prior on pursuiting in a market-based way, unless you're right about what the explosion looks like, the intelligence explosion.

Speaker 1 And so, like,

Speaker 1 don't move yet, but in that world where like really it does seem like Alec Radford can be automated and that is the only bottleneck to getting TSI. Okay, I think we can leave it at that.

Speaker 1 Yeah,

Speaker 1 I'm some somewhat of the way there. Okay, okay.

Speaker 2 Yeah, I hope it goes well.

Speaker 1 It's gonna be

Speaker 1 very stressful.

Speaker 2 And again, right now is the chill time.

Speaker 2 Enjoy your vacation while it lasts.

Speaker 1 It's like funny to look on over.

Speaker 1 I'm just like, this is San Francisco. Yeah, yeah, yeah.

Speaker 2 And open the eyes right there, you know, Anthropics there.

Speaker 2 I mean, again, this is kind of like, you know, it's like you guys have this enormous power over how it's, how it's going to go for the next couple of years. And that power is depreciating.

Speaker 1 Yeah.

Speaker 1 Who is you guys?

Speaker 2 Like, you know, people at labs.

Speaker 1 Yeah, yeah, yeah.

Speaker 2 But it is a sort of crazy world. And you're talking about like, you know, I feel like you talk about like, oh, maybe they'll nationalize too soon.
It's like, you know, almost nobody like.

Speaker 2 really like feels it, sees what's happening.

Speaker 2 And it's, it's, I think this is the thing that I find stressful about all this stuff is like, look, maybe I'm wrong, but like if I'm right, we're in this crazy situation where there's like, you know, like a few hundred guys that are like paying attention.

Speaker 2 Um,

Speaker 2 and um, it's it's daunting.

Speaker 1 I went to Washington a few months ago, yeah, and I was talking to some people who were doing AI policy stuff there, yeah, and I was asking them how likely they think nationalization is, yeah, and they said, Oh, you know, like

Speaker 1 it's really hard to nationalize stuff, it's been a long time since we've done it.

Speaker 1 There's these very specific procedural constraints on what kinds of things can be nationalized, and then I was asked, well, like ASI, so that means means because there's constraints on the Defense Production Act or whatever, that won't be nationalized.

Speaker 1 The Supreme Court would overturn that.

Speaker 1 And they were like, yeah, I guess that would be nationalized.

Speaker 1 That's the short summary of my post or my view on the project.

Speaker 1 Okay, so

Speaker 1 before we go further on the AI stuff, let's just back up.

Speaker 1 We began the conversation. I think people will be confused.
You graduated from Valedictorian of Columbia when you were 19. So you got to college when you were 15.
Right. And

Speaker 1 so you were in Germany. Then you got to college at 15.
Yeah.

Speaker 1 How the fuck did that happen?

Speaker 2 I really wanted out of Germany.

Speaker 2 I, you know, I went to kind of a German public school. It was not a good environment for me.

Speaker 1 And, you know, I mean, in what sense? It's just like no peers that are.

Speaker 2 Yeah, look, I mean, it wasn't, yeah, it was, you know, there's, I mean, there's also just a sense in which sort of like, there's this particular sort of German cultural sense.

Speaker 2 I think in the US, you know, there's all these like amazing high schools and like sort of an appreciation of excellence.

Speaker 2 And in Germany, there's really this sort of like Paul Poppy syndrome almost, right? Where it's, you know, you're the curious kid in class and you want to learn more.

Speaker 2 Instead of the teacher being like, ah, that's great. They're like, they kind of resent you for it and they're like trying to crush you.
And

Speaker 2 there's also like, there's no kind of like elite universities for undergraduate, which is kind of crazy.

Speaker 2 So, you know, the sort of, you know, there's sort of like basically like the meritocracy was kind of crushed in Germany at some point.

Speaker 2 Also, I mean, there's this sort of incredible sense of, you know, complacency,

Speaker 2 you know, across the board. I mean, one of the things that always puzzles me is like, you know, even just going to a U.S.
college was this kind of like radical act.

Speaker 2 And like, you know, it doesn't seem radical to anyone here because it's like, ah, this is obviously the thing you do.

Speaker 2 And you can go to Columbia, you go to Columbia, but it's, you know, it's very unusual. And

Speaker 2 it's wild to me because it's like, you know, this is where stuff is happening. You can get so much of a better education.
And, you know, like America is where, you know, it's

Speaker 2 where all the stuff is. And people don't do it.
And, and so, um,

Speaker 2 yeah, anyway, so I, you know, I don't know, I skipped a few grades. And, and, you know, I think at the time it seemed very normal to me to kind of like go to college and make you come to America.

Speaker 2 I think,

Speaker 2 you know, now my, one of my sisters is now like turning 15, you know, and so then I, you know, and I look at her and I'm like, now I understand how my mother's plan.

Speaker 2 And then so you get to college, you're like presumably the only year old yeah yeah as it was just like normal for you to be a 15 year old like what was the initial year normal at the time you know i didn't yeah so you're it's like now i understand why my mother's working it you know i think you know i worked i worked on my parents for a while you know eventually i was you know i persuaded them no but yeah it felt felt very normal at the time and it was great it was also great because i you know i actually really like college right um and in some sense it sort of came at the right time for me um where you know i um

Speaker 2 I mean, I, you know, for example, I really appreciated the sort of like liberal arts education and, you know, like the core curriculum and reading sort of core works of political philosophy and literature.

Speaker 2 And you did what econ and I mean, my majors were math and statistics and economics.

Speaker 2 But, you know, Columbia has a sort of pretty heavy core curriculum and liberal arts education. And honestly, like, you know, I shouldn't have done all the majors.

Speaker 2 I should have just, I mean, the best courses were sort of the courses where it's like, there's some amazing professor and it's some history class. And it's,

Speaker 2 I mean, that's, that's honestly the thing I would recommend people spend their time on in college.

Speaker 1 Was there one professor or class that stood out that way?

Speaker 2 I mean, a few. There's like a class by Richard Betts

Speaker 2 on war, peace, and strategy.

Speaker 2 Adam too is obviously fantastic

Speaker 2 and has written very riveting books.

Speaker 2 You should have him on the podcast, by the way.

Speaker 1 I tried. I tried.

Speaker 1 I think you tried for me. Yeah, you got to get him on the pod, man.

Speaker 2 Oh, it'd be so good.

Speaker 1 Okay, so then in a couple of years,

Speaker 1 we were talking to Tyler Cowen recently, and he said that when the way

Speaker 1 he first encountered you was you wrote this paper on economic growth and existential risk. And he said,

Speaker 1 when I found, read it, I couldn't believe that a 17-year-old had written it. I thought if this was an MIT dissertation, I'd be impressed.

Speaker 1 So you were like,

Speaker 1 how did you go from,

Speaker 1 I guess you were in junior then,

Speaker 1 you're writing, you know, pretty novel economic papers.

Speaker 1 Where did you get interested in this kind of thing? And what was the process to get into that?

Speaker 2 I don't know. I just, you know, I get interested in things.
In some sense, it's sort of like it feels very natural to me. It's like I get excited about a thing.
I read about it. I immerse myself.

Speaker 2 I think I can learn information very quickly and understand it.

Speaker 2 I mean, I think to the paper, I mean, I think one actual,

Speaker 2 at least for the way I work, I feel like sort of moments of peak productivity matter much more than sort of average productivity.

Speaker 2 I think there's some jobs, you know, like COO or something, you know, like average productivity really matters. But I think there's sort of a,

Speaker 1 I often feel like I have periods of like, you know, there's some, there's a couple months where there's sort of nefferlescence and i'm like you know and the other times i'm sort of computing stuff in the background and at some point you know like writing the series this is also kind of similar and it's just like you you write it and and it's it's like it's really flowing and um that's sort of what ends up mattering i think even for ceos it might be the case that the peak productivity is very important there's i don't know one of our following chattima's rules one of our friends in a group chat has uh pointed out how many famous ceos and founders have been bipolar manic right right right yeah which is very much much the peak, um, like the call option on your productivity is the most important thing, and you get it by just increasing the volatility through bipolar.

Speaker 1 Um, okay, so that's interesting. And so, you get interested in economics first.
First of all, why economics? Like, you could read about anything at this move.

Speaker 1 Like, you, if you wanted, you know, you could, you kind of got a slow start on ML, right?

Speaker 1 You could have

Speaker 1 done all these years on Econ. There's an alternative world where you're like on the Super Alignment team at 17 instead of 21 or whatever it was.

Speaker 1 Oh, no.

Speaker 2 I mean, in some sense, I'm still doing economics, right? You know, what is what is straight lines on a graph?

Speaker 2 I'm looking at the log, log plots and like figuring out what the trends are and like thinking about the feedback loops and equilibrium, orange control dynamics.

Speaker 2 And, you know, it's, I think it is a sort of a way of thinking that I find very useful.

Speaker 2 And,

Speaker 2 you know, like what, you know, Dario and Ilya seeing scaling early, in some sense, that is a sort of very economic way of thinking. And also the sort of physics, kind of like empirical physics.

Speaker 2 You know, a lot of them are physicists.

Speaker 2 I think the economists usually can't code well enough and that's their issue but i think it's that sort of way of thinking um i mean the other thing is you know i i thought they were sort of um

Speaker 2 you know i thought of a lot of the sort of like core ideas of economics i thought were just beautiful um

Speaker 2 and um you know in some sense i feel like i was a little duped you know where it's like actually econ academia is kind of decadent now you know i think that you know for example the paper i wrote you know it's sort of i think the takeaway you know it's a long paper it's 100 pages of math or whatever i think the core takeaway i can you know kind of give the core intuition for in like you you know, 30 seconds, and it makes sense.

Speaker 2 And it's, and it's like, you don't actually need the math.

Speaker 2 I think that's the sort of the best pieces of economics are like that, where you do the work, but you do the work to kind of uncover insights that weren't obvious to you before.

Speaker 2 Once you've done the work, it's like some sort of like mechanism falls out of it that like makes a lot of crisp intuitive sense that like explains some facts about the world that you can then use in arguments.

Speaker 2 And I think, you know, I think you know, like a lot of Econ 101 like this, and it's great. A lot of Econ in the, you know, in the 50s and the 60s, you know,

Speaker 2 was like this. And, you know, Chad Jones' papers are often like this.
I really like Chad Jones' papers for this. You know, I think,

Speaker 2 you know, why did I ultimately not pursue econ academia was

Speaker 2 a number of reasons. One of them was Tyler Cowan.

Speaker 2 You know, he kind of took me aside and he was kind of like, look, I think you're one of the top young economists I've ever met, but also you should probably not go to grad school.

Speaker 1 Oh, interesting. Really? I didn't realize that.

Speaker 2 Well, yeah, and it was good because he kind of introduced me to the, you know, I don't know, like the Twitter weirdos.

Speaker 2 And I think the takeaway from that was kind of um you know gotta move out west one more time did wait tyler introduced you to the twitter weirdos a little bit yeah or just kind of like the sort of brought you like the 60-year-old

Speaker 2 economist to introduce you to the twitter yeah well you know i i had been i had so i went from germany you know completely you know on the periphery to kind of like you know in a u.s elite institution and sort of got got some vibe of like sort of you know

Speaker 2 beritocratic elite you know u.s society and then sort of yeah basically this sort of like there was a sort of trajectory then to being like look i you know to find the true american spirit i I got to come out here.

Speaker 2 But anyway, the other reason I didn't become an economist was because, or at least econ academia, was because I think sort of econ academia has become a bit decadent.

Speaker 2 And maybe it's just ideas getting harder to find, and maybe it's sort of things, you know, and the sort of beautiful, simple things have been discovered.

Speaker 2 But, you know, like, what are econ papers these days? You know, it's like, you know, it's like

Speaker 2 200 pages of like empirical analyses on what happened when you know like Wisconsin bought 100,000 more textbooks on like educational outcomes. And I'm really happy that work happens.

Speaker 2 I think it's important work, but I think it is not in covering these sort of like fundamental insights and sort of mechanisms in society.

Speaker 2 Or, you know, it's like even the theory work is kind of like, here's a really complicated model, and the model spits out, you know, if the Fed does X, you know, then Y happens, and you have no idea what that why that happened because it's like gazillion parameters and they're all calibrated in some way, and it's some computer simulation, and you have no idea about the validity, you know.

Speaker 2 Yeah, so I think, I think the sort of you know, the most important insights are the ones where you have to do a lot of work to get them, but then there's sort of this crisp intuition.

Speaker 1 Yeah, yeah, the P versus N P of sure, yeah, yeah, yeah. Um,

Speaker 1 That's really interesting. So just going back to your time in college,

Speaker 1 you say that peak productivity kind of explains this paper and things, but the valedictorian, that's getting straight A's or whatever, is very much

Speaker 1 an average productivity phenomenon, right? So

Speaker 2 there's one award for the highest GPA, which I won, but the valedictorian is like among the people which have the highest GPA and then like selected by faculty.

Speaker 1 Okay, yeah. So it's just not, but it's not just peak productivity.

Speaker 2 It's just, it's, it's, it's just, it's, I generally just love this stuff. You know, I just, I was curious and I thought it was really interesting and I love learning about it.

Speaker 2 And, and I loved kind of like, it made sense to me. And, you know, it was, it was very natural.
And so, you know, I think I'm, you know, I'm not,

Speaker 2 you know, I think one of my faults is I'm not that good at eating class or whatever. And I think there's some people who are very good at it.

Speaker 2 I think the sort of like the sort of moments of pre-productivity come when I, you know, I'm just really excited and engaged and

Speaker 2 love it. And, you know, I, I,

Speaker 2 you know, if you take the right courses, you know, that's what you got in college. Yeah.

Speaker 1 It's the Bruce Banner code in Avengers.

Speaker 1 I'm always angry.

Speaker 1 I'm always excited. I'm always curious.
That's why I'm always deep activity.

Speaker 1 So it's interesting, by the way, when you were in college, I was also in college.

Speaker 1 I think you were, despite being a year younger than me, I think a year ahead in college than me, or at least maybe two years ahead.

Speaker 1 And we met around this time. Yeah, yeah, yeah.
We also met, I think, through the Tyler Cowan universe. Yeah, yeah.
And it's very insane how small the world is. Yeah.
I think I, did I reach out to you?

Speaker 1 I must have. Yes.
About

Speaker 1 when I had a couple of videos and they had a couple hundred views or something. Yeah.

Speaker 2 It's a small world. Yeah.
I mean, this is the crazy thing about the AI world, right?

Speaker 2 It's kind of like it's the same few people at the convesta parties and they're the ones, you know, running the models at DeepMind and, you know, open AI and anthropic. And, and,

Speaker 2 you know, I mean, I think some other friends of ours have mentioned this who are now later in their career and very successful That, you know, they actually met all the people who are also kind of very successful in Silicon Valley now, like, you know, when they're, when they're in their, you know, when they're before their 20s or when they're early 20s.

Speaker 2 I mean, look, I actually think, you know, and why is it a small world?

Speaker 2 I mean, I think one of the things is some amount of like, you know, some sort of agency.

Speaker 2 And I think in a funny way,

Speaker 2 this is a thing I sort of took away from the sort of Germany experience, where it was, I mean, look,

Speaker 2 it was crushing. I really didn't like it.
And it was like, it was such an unusual move to kind of skip grades. It was such an unusual move to come to the United States.

Speaker 2 And, you know, a lot of these things I did were kind of unusual moves. And,

Speaker 2 you know, there's some amount where like

Speaker 2 just like just trying to do it and then it was fine and it worked.

Speaker 2 That kind of reinforced like, you know, you don't, you don't just have to kind of conform to what the Overton window is.

Speaker 2 You can just kind of like try to do the thing, the thing that seems right to you.

Speaker 2 And like, you know, most people can be wrong. And I don't know, things like that.
And I think that was kind of a, you know, a valuable kind of like early experience that was sort of formative.

Speaker 1 Okay. So after college, what did you do?

Speaker 2 I did econ research for a little bit, you know, in Oxford and stuff. And then I worked at Future Fund.

Speaker 1 Yeah. Okay.
So, and so tell me about it. Yeah.

Speaker 2 Future Fund was a, you know, it was a foundation that was, you know, funded by Sam Bankman Fried. I mean, we were our own thing, you know, we were based in the Bay.

Speaker 2 You know, at the time, this was in sort of early 22,

Speaker 2 it was, it was this just like incredibly exciting opportunity, right?

Speaker 2 It was basically like a startup, you know, foundation, which is like, you know, it doesn't come along that often that, you know, we thought would be able to give away billions of dollars.

Speaker 2 You know, thought would be able to kind of like, you know, remake how philanthropy is done, you know, from first principles.

Speaker 2 Thought would be able to have, you know, this like great impact. You know, we, the causes we focused on were, you know, biosecurity, you know, AI,

Speaker 2 you know, finding exceptional talent and putting them to work on hard problems.

Speaker 2 And, you know, like a lot of the stuff we did, I was really excited about, you know, like academics who would, you know, usually take six months would send us emails like, ah, you know, this is great.

Speaker 2 This is so quick and easy, you know, and straightforward.

Speaker 2 You know, in general, I feel like I've often found that with like, you know, a little bit of encouragement, a little bit of sort of empowerment, kind of like removing excuses, making the process easy, you know, you can kind of like get people to do great things.

Speaker 1 I think on the Fear Future Fund, that I think is

Speaker 1 context for people who might not realize. Yeah.
Not only were you guys planning on deploying billions of dollars, but it was a team of four people. Yeah, yeah, yeah.

Speaker 1 So you at 18 are on a team of four people that is in charge of deploying billions of dollars. Yeah.

Speaker 2 I mean, just, I mean, yeah, and Future Fund, you know, the

Speaker 2 yeah, I mean, the, yeah, so that was, that was sort of the heyday, right?

Speaker 2 I mean, then obviously, you know, when, when in sort of, you know, November of 22,

Speaker 2 you know, it was kind of revealed that Sam was this, you know, giant fraud. Um, and from one day to the next, you know, the whole thing collapsed.

Speaker 2 That was just really tough. I mean, you know, obviously, yeah, it was devastating.
It was devastating, obviously, for the people at their money and FTX,

Speaker 2 you know,

Speaker 2 closer to home, you know, all the, you know, all these grantees, you know, we wanted to help them and we thought they were doing amazing projects.

Speaker 2 And so, but instead of helping them, we ended up saddling them with like a giant problem.

Speaker 2 You know, personally, it was, you know, it was a startup, right? And so I, you know, I had worked 70-hour weeks every week for basically a year on this to kind of build this up.

Speaker 2 You know, we're a tiny team.

Speaker 2 And then from one day to the next, it was all gone. And not just gone, it was associated with this giant fraud.

Speaker 2 And so, you know, that was incredibly tough.

Speaker 1 Yeah.

Speaker 1 And then were there any signs early on that SBF was?

Speaker 2 Yeah, like, obviously, I didn't know he was a fraud. And the whole, you know,

Speaker 1 I would have never worked if I did, you know?

Speaker 2 And, you know, we weren't, you know, we were a separate thing. We weren't working with the business.

Speaker 2 I mean, I think, I do think there were some takeaways for me. I think one takeaway was,

Speaker 2 you know, I think there's a,

Speaker 2 I had this tendency.

Speaker 2 I think people in general have this tendency to kind of like, you know, give successful CEOs a pass on their behavior because you know they're successful CEOs and that's how they are and that's just successful CEO things and

Speaker 2 you know

Speaker 2 I didn't know Sam Mankman Fried was a fraud but I knew SBF and I knew he was extremely risk-taking right I knew he he was narcissistic

Speaker 2 he didn't tolerate disagreement well you know sort of by the end he and I just like didn't get along well and sort of I think the reason for that was like there's some biosecurity grants he really liked because they were kind of cool and flashy and at some some point, I'd kind of run the numbers and it didn't really seem that cost-effective.

Speaker 2 And I pointed that out, and he was pretty unhappy about that.

Speaker 2 And so I knew his character.

Speaker 2 And I think, you know, I feel like one takeaway for me was,

Speaker 2 you know, like, I think it's really worth paying attention to people's character, and including like people you work for and successful CEOs.

Speaker 2 And, you know, that can save you a lot of pain down the line.

Speaker 1 Okay, so after that, FTX implodes and you're out.

Speaker 1 And then

Speaker 1 you got into,

Speaker 1 you went to OpenAI. The Super Alignment team had just started.
I think you were like part of the initial team. And so

Speaker 1 what was the original idea? What was compelling about that for you to join?

Speaker 2 Yeah, totally.

Speaker 2 So, I mean, what was the goal of the Super Alignment team?

Speaker 2 You know, the

Speaker 2 alignment team at OpenAI, you know, at, you know, other labs sort of like several years ago kind of had done sort of basic research and they developed RLHF, reinforcement learning from human feedback.

Speaker 2 And that was sort of a, you know, ended up being a really successful technique for controlling sort of current generation of AI models.

Speaker 2 What we were trying to do was basically kind of be the basic research vet to figure out what is the successor to RLHF.

Speaker 2 And the reason we needed that is, you know, basically, you know, RLHF probably won't scale to superhuman systems.

Speaker 2 RLHF relies on sort of human raiders who kind of thumbs up, thumbs down, you know, like the model said something. It looks fine, it looks good to me.

Speaker 2 At some point, you know, the superhuman models, the super intelligence, it's going to write, you know, a million lines of crazy complex code. You don't know at all what's going on anymore.

Speaker 2 And so how do you kind of steer and control these systems? How do you add side constraints?

Speaker 2 The reason I joined was

Speaker 2 I thought this was an important problem. And I thought it was just a really solvable problem, right? I thought this was basically,

Speaker 2 I think there's a, I still do, I mean, even more so do. I think there's a lot of just really promising sort of ML research on alignment on sort of aligning superhuman systems.

Speaker 2 And maybe we should talk about that a bit more later.

Speaker 1 But

Speaker 2 so, and then it was so solvable.

Speaker 1 You solved it in the year. It's all over.

Speaker 2 But anyway, so look, OpenAI, I wanted to do this like really ambitious effort on alignment. And, you know, Elliot was backing it.
And, you know, I liked a lot of the people there.

Speaker 2 And so I was, you know, I was really excited. And I was kind of like, you know, I think there was a lot of people

Speaker 2 sort of on alignment, there's always a lot of people kind of making hay about it. And, you know,

Speaker 2 I appreciate people highlighting the importance of the problem. And I was just really into like, let's just try to solve it.
And let's do the ambitious effort.

Speaker 2 You know, let's do the operational warp speed for solving alignment. And it seemed like an amazing opportunity to do so.

Speaker 1 Okay. And now basically the team doesn't exist.
I think the head of it has left.

Speaker 1 Both heads of it have left, Jan and Ilya. That's been the news of the last week.

Speaker 1 What happened? Why did the thing break down?

Speaker 2 I think OpenAI sort of decided to take things in a somewhat different direction.

Speaker 1 Meaning what?

Speaker 1 I mean, that super alignment isn't the best way to frame the...

Speaker 2 No, I mean, look, obviously, sort of after the November board events, you know, there were personnel changes. I think Ilya leaving was just incredibly tragic for OpenAI.

Speaker 2 And,

Speaker 2 you know, I think some amount of reprioritization, I think some amount of, you know, I mean, there's been some reporting on the super alignment compute commitment.

Speaker 2 You know, there's this 20% compute commitment as part of, you know, how a lot of people were recruited. You know, it's like, we're going to do this ambitious effort in alignment.
And,

Speaker 2 you know.

Speaker 2 some amount of, you know, not keeping that and deciding to go in a different direction.

Speaker 1 Okay, so now Jan has left, Ilya has left.

Speaker 1 So this team itself has dissolved, but you were the sort of first person who left or was forced to leave. You were the information reported that you were fired for leaking.

Speaker 1 What happened? Was this accurate?

Speaker 2 Yeah.

Speaker 1 Look,

Speaker 2 why don't I tell you what they claim I leaked and you can tell me what you think. Yeah, so OpenAI did claim to employees that I was fired for leaking.

Speaker 2 And, you know, I and others have sort of pushed them to say what the leak is. And so here's their response in full.

Speaker 2 You know, sometime last year, I had written a sort of brainstorming document on preparedness, on safety and security measures we need in the future on the path to AGI.

Speaker 2 And I had shared that with three external researchers for feedback. So that's it.
That's the leak.

Speaker 2 You know, I think for context, it was totally normal at OpenAI at the time to share sort of safety ideas with external researchers for feedback.

Speaker 2 You know, it happened all the time.

Speaker 2 You know, the doc was sort of my ideas. You know, before I shared it, I reviewed it for anything sensitive.

Speaker 2 the internal version had a reference to a future cluster but I redacted that for the external copy

Speaker 2 you know there's a link in there to some to some slides of mine internal slides but you know that was a dead link to the external people I shared it with you know the slides weren't shared with them and so look obviously I

Speaker 2 pressed them to sort of tell me what is the confidential information in this document and what they came back with was a line in the doc about planning for AGI by 27-28 and that setting timelines for preparedness.

Speaker 2 You know, I wrote this doc, you know, a couple months after the super alignment announcement. We had put out, you know, this sort of four-year planning horizon.

Speaker 2 I didn't think that planning horizon was sensitive. You know, it's the sort of thing Sam says publicly all the time.

Speaker 2 I think sort of John said it

Speaker 1 on my phone.

Speaker 2 Anyway, so that's it.

Speaker 1 That's it. So that seems pretty thin for

Speaker 1 if the cause was leaking, that seems pretty thin. Was there anything else to it?

Speaker 2 Yeah, I mean, so that was, that was the leaking claim. I mean, I can say a bit more about sort of what happened in the filing.

Speaker 2 So, one thing was

Speaker 2 last year, I had written a memo, internal memo, about opening eye security. I thought it was egregiously insufficient.

Speaker 2 I thought it wasn't sufficient to protect the theft of model weights or key algorithmic secrets from foreign actors.

Speaker 2 So, I wrote this memo. I shared it with a few colleagues, a couple members of leadership, who sort of mostly said it was helpful.

Speaker 2 But then, a couple of weeks later, a sort of major security incident occurred.

Speaker 2 And that prompted me to share the memo with a couple members of the board.

Speaker 2 And so after I did that, you know, days later, it was made very clear to me that leadership was very unhappy with me having shared this memo with the board.

Speaker 2 You know, apparently the board had hassled leadership about security.

Speaker 2 And then I got sort of an official HR warning for this memo, you know, for sharing it with the board.

Speaker 2 The HR person told me it was racist to worry about CCP espionage.

Speaker 2 And they said it was sort of unconstructive.

Speaker 2 And, you know, look, I think I probably wasn't at my most diplomatic. You know, I definitely could have been more politically savvy,

Speaker 2 but I thought it was a really, really important issue. And, you know, the security incident had made me really worried.

Speaker 2 Anyway, and so I guess the reason I bring this up is when I was fired, it was sort of made very explicit that the security memo is a major reason for my being fired.

Speaker 2 You know, I think it was something like, you know, the reason that this is a firing and not a warning is because of the security memo.

Speaker 1 You sharing it with the board.

Speaker 2 The warning I'd gotten for the security memo.

Speaker 2 Anyway, and I mean, some other, you know, what might also be helpful context is the sort of questions they asked me when they fired me. So, you know, this was a bit over a month ago.

Speaker 2 I was pulled aside for a chat with a lawyer, you know, that quickly turned very adversarial.

Speaker 2 And, you know, the questions were all about my views on AI progress, on AGI,

Speaker 2 on the level of security appropriate for AGI, on, you know, whether government should be involved in AGI, on

Speaker 2 whether I and Super Alignment were loyal to the company,

Speaker 2 on what I was up to during the OpenAI board events, things like that. And then they chatted to a couple of my colleagues, and then they came back and told me I was fired.

Speaker 2 And they'd gone through all of my digital artifacts from the

Speaker 2 time at OpenAI, messages, docs. And that's when they found the leak.

Speaker 2 Yeah. And so, anyway, so the main claim they made was this leaking allegation.
You know, that's what they told employees.

Speaker 2 They, you know, the security memo.

Speaker 2 There's a couple other allegations they threw in.

Speaker 2 One thing they said was that I was unforthcoming during the investigation because I didn't initially remember who I had shared the doc with, the sort of preparedness brainstorming doc, only that I had sort of spoken to some external researchers about these ideas.

Speaker 2 And, you know, look, the doc was over six months old. You know, I'd spent a day on it.

Speaker 2 You know, it was a Google Doc. I shared with my open AI email.
It wasn't a screenshot or anything I was trying to hide. It simply didn't stick because it was such a non-issue.

Speaker 2 And then they also claimed that I was engaging on policy in a way that they didn't like.

Speaker 2 And so, what they cited there was that I had spoken to a couple external researchers, somebody got a think tank about my view that AGI would become a government project, as we discussed.

Speaker 2 In fact, I was speaking to lots of sort of people in the field about that at the time. I thought it was a really important thing to think about.

Speaker 2 Anyway, and so they found, you know, they found a DM that I'd written to like a friendly colleague, you know, five or six months ago where I relayed this and they cited that.

Speaker 2 And I had thought it was well within OpenAI norms to kind of talk about high-level issues on the future of AGI with external people in the field. So anyway, so that's what they alleged.

Speaker 2 That's what happened.

Speaker 2 I've spoken to kind of a few dozen former colleagues about this since. I think the sort of universal reaction is kind of like, that's insane.

Speaker 2 I was sort of surprised as well.

Speaker 2 I had been promoted just a few months before.

Speaker 2 I think Ilya's comment for the promotion case at the time was something like, Leopold's amazing. We're lucky to have him.

Speaker 2 But look, I think the thing I understand, and I think in some sense is reasonable, is like, you know, I think I ruffled some feathers. And I think I was probably kind of annoying at times.

Speaker 2 It's like I...

Speaker 2 security stuff and I kind of like repeatedly raised that and maybe not always in the most diplomatic way.

Speaker 2 I didn't sign the employee letter during the board events despite pressure to do so.

Speaker 1 And you were, what, one of like eight people or something, I'd bet.

Speaker 2 I guess the, I think the sort of two senior most people who didn't sign were Andre and January since left.

Speaker 2 And, you know, I mean, on the letter, by the way, I, by the time on sort of Monday morning when that letter was going around, I think probably it was appropriate for the board to resign.

Speaker 2 I think they'd kind of like lost too much credibility and trust with the employees.

Speaker 2 But I thought the letter had a bunch of issues. I mean, I think one of them was it just didn't call for an independent board.

Speaker 2 I think it's sort of like basics of corporate governance to have an independent board. Anyway, you know, it's other things.
You know, I,

Speaker 2 in sort of other discussions, I pressed leadership for sort of OpenAI to abide by its public commitments.

Speaker 2 You know, I raised a bunch of tough questions about whether it was consistent with the OpenAI mission and consistent with the national interest to sort of partner with authoritarian dictatorships to build the core infrastructure for AGI.

Speaker 2 So, you know, look, you know, it's a free country, right?

Speaker 2 That's what I love about this country. You know, we talked about it.

Speaker 2 And so they have no obligation to keep me on staff.

Speaker 2 And, you know, I think in some sense, I think it would have been perfectly reasonable for them to come to me and say, look, you know, we're taking the company in a different direction.

Speaker 2 You know, we disagree with your point of view.

Speaker 2 You know, we don't trust you enough to sort of tow the company line anymore. And,

Speaker 2 you know, thank you so much for your work at OpenAI, but I think it's time to part ways. I think that would have made sense.

Speaker 2 I think, you know, we did start sort of materially diverging on sort of views on important issues. I'd come in very excited and aligned with OpenAI, but that sort of changed over time.
And

Speaker 2 look, I think there would have been a very amicable way to part ways. And I think it's a bit of a shame that it sort of this is the way it went down.

Speaker 2 You know, all that being said, I think, you know, I really want to emphasize

Speaker 2 there's just a lot of really incredible people at OpenAI, and it was an incredible privilege to work with them. And, you know, overall, I'm just extremely grateful for my time there.

Speaker 1 When you left, now that there's now there's been reporting about

Speaker 1 an NDA that former employees have to sign in order to have access to their vested equity, did you sign such NDA?

Speaker 2 No.

Speaker 2 My situation was a little different in that it was sort of, I was basically right before my CLEF.

Speaker 2 But then they still offered me the equity,

Speaker 2 but I didn't want to sign the non-disparagement. Freedom is priceless.
And

Speaker 1 how much was the equity?

Speaker 2 close to a million dollars.

Speaker 1 So it was definitely a thing you and others were aware of, that this is like a choice that OpenAI is explicitly offering you. Yeah.

Speaker 1 And presumably the person on OpenAI's staff knew that we're offering them equity, but they had to sign this NDA that has these conditions that you can't, for example, give the kind of statements about your thoughts on AGI and OpenAI that you're giving on this podcast right now.

Speaker 2 Look, I don't know what the whole situation is. I certainly think sort of vested equity is pretty rough if you're conditioning that onto an NDA.

Speaker 2 It might be a somewhat different situation if it's a sort of severance agreement.

Speaker 1 Right. But an OpenAI employee who had signed it presumably could not give the podcast that you're giving today.

Speaker 2 Quite plausibly not.

Speaker 2 Yeah, I don't know.

Speaker 1 Okay, so analyzing the situation here, I guess if you were to.

Speaker 1 Yeah, the board thing is really tough because if you were trying to defend them, you would say, well,

Speaker 1 listen, you were just kind of going outside the regular chain of command and maybe there's a point there although the way in which the person from hr thinks that you have an adversarial relationship with or you're the you're supposed to have an adversarial relationship with the board where to to give the board some information which is relevant to

Speaker 1 whether open ai is fulfilling its mission and whether it can do that in a better way is part of the leak as if the board is that is supposed to ensure that open ai is following its mission is some sort of external actor that seems pretty.

Speaker 2 I mean, I think, I think, I mean, to be clear, the leak allegation was just that sort of document I chose for feedback.

Speaker 2 This is just sort of a separate thing that they cited, and they said, I wouldn't have been fired if not for the security memo.

Speaker 2 They said you wouldn't have been fired if they said the reason this is a firing and not a warning is because of the warning you had gotten for the security memo.

Speaker 1 Oh.

Speaker 1 Before you left, the incidents with the board happened

Speaker 1 where Sam was fired and then rehired as CEO, and now he's on the board.

Speaker 1 Now,

Speaker 1 Ilya and Jan, who are the heads of the Super Alignment team, team, and Ilya, who is a co-founder of OpenAI, obviously the most significant in terms of stature member of OpenAI from a research perspective, they've left.

Speaker 1 Seems like, especially with regards to Super Alignment stuff and just generally with OpenAI, a lot of this sort of personnel drama has happened over the last few months. What's going on?

Speaker 2 Yeah, there's a lot of drama.

Speaker 2 Yeah, so why is there so much drama?

Speaker 2 You know, I think there would be a lot less drama if all OpenAI claimed to be be was sort of building ChatGPT or building business software.

Speaker 2 I think where a lot of the drama comes from is, you know, OpenAI really believes they're building AGI, right? And it's not just, you know, a claim they make for marketing purposes, you know,

Speaker 2 whatever. You know, there's this report that Sam is raising, you know, $7 trillion for chips.
And it's like, that stuff only makes sense if you really believe in AGI.

Speaker 2 And so I think what gets people sometimes is sort of the cognitive dissonance between sort of really believing in AGI, but then sort of not taking some of the other implications seriously.

Speaker 2 You know, this is going to be incredibly powerful technology, both for good and for bad. And that implicates really important issues like the national security issues we spoke about.

Speaker 2 Like, you know, are you protecting the secrets from the CCP? Like, you know, does America control the core AGI infrastructure?

Speaker 2 Or does it, you know, a Middle Eastern dictator control the core AGI infrastructure?

Speaker 2 And then, I mean, I think the thing that

Speaker 2 you know really gets people is the sort of tendency to kind of then make commitments.

Speaker 2 And sort of like, you know, they say they take these issues really seriously, they make big commitments on them, but then sort of frequently don't follow through. Right.

Speaker 2 So, you know, again, as mentioned, there was this commitment around super alignment compute, you know, sort of 20% of compute for this long-term safety research effort.

Speaker 2 And I think, you know, you and I could have a totally reasonable debate about what is the appropriate level of compute for super alignment.

Speaker 2 But that's not really the issue. The issue is that this commitment was made and it was used to recruit people and, you know, it was very public.

Speaker 2 And it was made because, you know, there's a recognition that there would always be something more urgent than a long-term safety research effort, you know, like some new product or whatever.

Speaker 2 And but then, in fact, they just, you know, really didn't keep the commitment. And so, you know, there was always something more urgent than long-term safety research.

Speaker 2 I mean, I think another example of this is, you know, when I raised these issues about security, you know,

Speaker 2 they would tell me, you know, security is our number one priority.

Speaker 2 But then, you know, invariably, when it came time to sort of invest serious resources, when it came time to make trade-offs, to sort of take some pretty basic measures,

Speaker 2 security would not be prioritized.

Speaker 2 And so, yeah, I think it's the cognitive dissonance, and I think it's the sort of unreliability that causes a bunch of the drama.

Speaker 1 So, let's zoom out

Speaker 1 to talk about a big part of the story and also a big motivation of the way in which it must proceed with regards to geopolitics and everything is that once you have the AGI, pretty soon after you proceed to ASI because superintelligence, because you have these AGIs which can function as researchers into further AI progress and within a matter of years maybe less you go to something that is like super intelligence and at the high and then from there then you do up in according to your story do all this research and development into robotics and pocket nukes and whatever other crazy shit yeah but at a high level

Speaker 1 okay but there's I'm skeptical of this story for many reasons Yes. At a high level,

Speaker 1 it's not clear to me this input-output model of research is how things actually happen in research. We can look at economy-wide, right?

Speaker 1 Patrick Collis and others have made this point that compared to 100 years ago, we have 100x more researchers in the world. It's not like progress is happening 100x faster.

Speaker 1 So it's clearly not the case that you can just pump in more population into research and you get higher research on the other end.

Speaker 1 I don't know why it would be be different for the AI resources themselves.

Speaker 2 Okay, great. So this is getting into some good stuff.

Speaker 2 I have a classic disagreement I have with Patrick and others. Okay, so, you know, obviously inputs matter, right?

Speaker 2 So it's like the United States produces a lot more scientific and technological progress than, you know, Liechtenstein, right? Or Switzerland.

Speaker 2 And even if I made, you know, Patrick Collison dictator of like Liechtenstein or Switzerland, and Patrick Collison was able to implement his like, you know, utopia of ideal institutions, keeping the talent pool fixed.

Speaker 2 He's not able to like do some crazy high school immigration thing or like, you know, whatever, some like crazy genetic breeding scheme or whatever he wants to do.

Speaker 2 Keeping the talent pool fixed, but amazing institutions.

Speaker 2 I claim that still, even if you made Patrick Collison dictator of Switzerland, maybe you get some factor, but Switzerland is not going to be able to outcompete the United States in scientific and technological powers.

Speaker 2 Obviously, magnitudes matter.

Speaker 1 Okay, now this is. No, I actually, I'm not sure I agree with this.

Speaker 1 There's been many examples in history where you have small groups of people who are part of like Bell Labs or Skunk Works or something. There's a couple hundred researchers.
Open AI, right?

Speaker 1 A couple hundred researchers.

Speaker 2 They do highly selected, though, right? You know, it's like, it's like saying, you know,

Speaker 1 that's part of why Patrick Hollison is his dictator is going to do a good job of this.

Speaker 2 Well, yes, if he can highly select all the best AI researchers in the world, he might only need a few hundred. But if you can, you know, that's, that's the talent pool.

Speaker 2 It's like you have the, you know, 300 best AI researchers in the world.

Speaker 1 But there's, there has been, it's not the case that from 100 years to now, there haven't been. Population has increased massively.

Speaker 1 A lot of the world, in fact, you would expect the density of talent to have increased in the sense that malnutrition and other kinds of debility, poverty, whatever, that have debilitated past talent at the same sort of level is no longer deliberate

Speaker 2 to the 100x point, right? So, I don't know if it's 100x, I think it's easy to inflate these things, probably at least 10x.

Speaker 2 Um, and so people are sometimes like, ah, you know, like, you know, come on, ideas haven't gotten that much harder to find. You know, why would you have needed this 10x increase in research effort?

Speaker 2 Um, whereas to me, I think this is an extremely natural story. And why is it a natural story? It's a straight line on a log-log plot.

Speaker 2 This is sort of a you know, deep learning researcher's dream, right? What is this log-log plot? On the x-axis, you have log cumulative research effort.

Speaker 2 On the y-axis, you have log GDP or ooms of algorithmic progress or

Speaker 2 log transistors per square inch or in the sort of experience curve for solar, kind of like whatever the log of the price for a gigawatt of solar.

Speaker 2 And it's extremely natural for that to be a straight line. This is sort of a classic, yeah, it's a classic.
And it's basically the first thing is very easy.

Speaker 2 Then basically, you have to have log increments of cumulative research effort to find the next thing.

Speaker 2 And so, you know, in some sense, I think this is a a natural story.

Speaker 2 Now, one objection kind of people then make is like, oh, you know, isn't it suspicious, right?

Speaker 2 That like ideas, you know, well, we increased research effort 10x and ideas also just got 10x harder to find. And so it perfectly, you know, equilibriates.

Speaker 2 And to there, I say, you know, it's just, it's an equilibrium. It's an adjusted equilibrium, right? So it's like, you know.

Speaker 2 Isn't it a coincidence that supply equals demand, you know, and the market clears, right? And that's, and the same thing here, right?

Speaker 2 So it's, you know, ideas getting, how much ideas have gotten harder to find is a function of how much progress you've made.

Speaker 2 And then, you know, what the overall growth rate has been is a function of how much ideas have gotten harder to find in ratio to how much you've been able to like increase research effort.

Speaker 2 What is the sort of growth of log cumulative research effort? So in some sense, I think the story is sort of like fairly natural. And you see this, and you see this not just economy-wide.

Speaker 2 You see it in kind of experience curves for all sorts of individual technologies.

Speaker 2 So I think there's some process like this. I think it's totally plausible that, you know, institutions have gotten worse by some factor.

Speaker 2 Obviously, there's some sort of exponent of diminishing returns on more people, right? So serial time is better than just parallelizing.

Speaker 2 But still, I think it's like clearly inputs matter.

Speaker 1 Yeah, I agree.

Speaker 1 But if the coefficient of

Speaker 1 how fast they diminish as you grow the input is high enough, then

Speaker 1 in the abstract, the fact that inputs matter isn't that relevant. Okay, so I mean, we're talking at a very high level, but just like take it down to the actual concrete thing here.

Speaker 1 OpenAI has a staff of at most low hundreds who are directly involved in in the algorithmic progress in future models.

Speaker 1 If it was really the case that you could just arbitrarily scale this number and you could have much faster algorithmic progress, and that would result in much higher, much better AIs for OpenAI, basically, then it's not clear why OpenAI doesn't just go out and hire every single person with 150 IQ, of which there are hundreds of thousands in the world.

Speaker 1 And

Speaker 1 my story there is there's transaction costs to managing all these people that don't just go away if you have a bunch of AIs, that

Speaker 1 these tasks aren't easy to parallelize.

Speaker 1 And I think you, I'm not sure how you would explain the fact of like, why doesn't OpenAI go on a recruiting binge of every single

Speaker 1 genius in the world?

Speaker 2 All right, great. So let's talk about the OpenAI example and let's talk about the automated AI researchers.

Speaker 2 So, I mean, in the OpenAI case, I mean, just, you know, just kind of like look at the inflation of like AI researcher salaries over the last year.

Speaker 2 I mean, I think like, I don't know, I don't know what it is, you know, 4x, 5x. It's kind of crazy.
So they're clearly really trying to recruit the best AI researchers in the world.

Speaker 2 And, you know, I don't know, it's, they do find the best AI researchers in the world.

Speaker 2 And I think my response to your thing is like, you know, almost all of these 150 IQ people, you know, if you just hired them tomorrow, they wouldn't be good AI researchers.

Speaker 2 They wouldn't be in Alec Radford.

Speaker 1 But they're willing to make investments that take your step out of the fourth.

Speaker 1 The data centers they're buying right now will come online in 2026 or something.

Speaker 1 Why wouldn't they be able to make every 150 IQ person, some of them won't work out, some of them won't have the traits we like. But some of them by 2026 will be amazing AI researchers.

Speaker 1 Why aren't they making that bet?

Speaker 2 Yeah. And so sometimes this happens, right? Like smart physicists have been really good at AI research.
You know, it's like all the anthropic co-founders.

Speaker 1 But like if you talk to, I had Dario on the podcast, like they have this very careful policy of like, we're not going to just hire arbitrarily. We're going to be extremely selective.
Yep.

Speaker 2 Training is not as easily scalable, right? So training is very hard. You know, if you just hired, you know, 100,000 people, it's like,

Speaker 2 I mean, you couldn't train them all. It'd be really hard to train them all.
You know, you wouldn't be doing any AI research.

Speaker 2 Like, you know, there's, there's huge costs to bringing on a new person, training them. This is very different with AIs, right?

Speaker 2 And I think this is, it's really important to talk about the sort of like advantages the AIs will have. So it's like, you know, training, right? It's like, what does it take to be an Alec Radford?

Speaker 2 You know, we need to be in a really good engineer, right? The AIs, they're going to be an amazing engineer. They're going to be amazing at coding.
You can just train them to do that.

Speaker 2 They need to have, you know, not just be a good engineer, but have really good research intuitions and like really understand deep learning.

Speaker 2 And this is stuff that, you know, Alec Radford or somebody like him has acquired over years of research, over just like being deeply immersed in deep learning, having tried lots of things himself and failed.

Speaker 2 The AIs, you know, they're going to be able to read every research paper I've written, every experiment ever run at the lab, you know, like gain the intuitions from all of this.

Speaker 2 They're going to be able to learn in parallel from all of each other's experiment, you know, experiences.

Speaker 2 You know, I don't know what else. You know, it's like, what does it take to be an Alec Radford? Well, there's a, there's a sort of cultural acclimination aspect of it, right?

Speaker 2 You know, if you hire somebody new, you know, there's like politicking, maybe they don't fit in. Well, in the AI case, you just make replicas, right? There's a like motivation aspect for it, right?

Speaker 2 So it's like, you know, Alec, you know, if I could just like duplicate Alec Radford, and before I run every experiment, I haven't spent like, you know, a decade's worth of human human time like double checking the code and thinking really careful carefully about it.

Speaker 2 I mean, first of all, I don't have that many Alec Radfords and he wouldn't care and he would not be motivated. But you know, the AIs, it can just be like, look, I have 100 million of you guys.

Speaker 2 I'm just going to put you on just like really making sure this code is correct. There are no bugs.
This experiment is thought through. Every hyperparameter is correct.

Speaker 2 Final thing I'll say is, you know, the 100 million human equivalent AI researchers, that is just a way to visualize it. So that doesn't mean you're going to have literally 100 million copies.

Speaker 2 So there's trade-offs you can make between serial speed and in parallel. So you might make the trade-off is: look, we're going to run them at 10x, 100x serial speed.

Speaker 2 It's going to result in fewer tokens overall because of sort of inherent trade-offs.

Speaker 2 But then we have, I don't know what the numbers would be, but then we have 100,000 of them running at 100x human speed and thinking, and you know, and there's other things you can do on coordination, you know, they can kind of like share latent space, it tends to each other's context.

Speaker 2 There's basically this huge range of possibilities of things you can do. The 100 million thing is more.

Speaker 2 I mean, another illustration of this is, you know, if you kind of, I run the math in my series, and it's basically, you know, 27, 28, you have this automated AI researcher.

Speaker 2 You're going to be able to generate an entire internet's worth of tokens every single day. So there's clearly sort of a huge amount of like intellectual work that you can do.

Speaker 1 I think the analogous thing there is today we generate more patents in a year than during the actual physics revolution in the early 20th century.

Speaker 1 They were generating across like half a century or something. And are you making more physics progress in a year today than we merely? So yeah, you're going to generate all these tokens.

Speaker 1 Are you generating as much codified knowledge as humanity has been able to generate in the

Speaker 1 initial creation of the internet?

Speaker 2 Internet tokens are usually final output, right?

Speaker 2 A lot of these tokens, if we talk, we talked about the unhobbling, right? And I think I think of a kind of like, you know, a GPDN token as sort of like one token of my internal monologue.

Speaker 2 And so that's how I do this math on human equivalence. You know, it's like 100 tokens a minute, and then, you know, humans working for X hours.
And, you know,

Speaker 2 what is the equivalent there?

Speaker 1 I think this goes back to something we were talking about earlier, where why haven't we seen the huge revenues from

Speaker 1 people often ask this question that if you took GPT-4 back 10 years and you showed people this, that they think this is going to automate, this has already automated half the jobs.

Speaker 1 And so there's a sort of a modus ponens, modus totens here, where part of the explanation is like, oh, it's like just on the verge, you need to do these unhobblings. And part of that is probably true.

Speaker 1 But there is another lesson to learn there, which is that just looking at face value at a set of abilities,

Speaker 1 there's probably more sort of hobblings that you don't realize that are hidden behind the scenes. And I think the same will be true of the AGIs that you have running as AI researchers.

Speaker 1 I think a lot of things. I basically agree, right?

Speaker 2 I think my story here is like, you know,

Speaker 2 I talk about, I think there's going to be some long tail, right?

Speaker 2 And so part, you know, maybe it's like, you know, 26, 27, you have like the proto-automated engineer, and it's really good at engineering. He doesn't have the research intuition yet.

Speaker 2 You don't quite know how to put them to work.

Speaker 2 But, you know, the sort of even the underlying pace of AI progress is already so fast, right?

Speaker 2 In three years from not being able to do any kind of like math at all to now crushing, crushing these math competitions.

Speaker 2 And so you have the initial thing in like 26, 27, maybe the sort of auto, it's an automated research engineer. It speeds you up by 2x.
You go through a lot more progress in that year.

Speaker 2 By the end of the year, you figured out like the remaining kind of unhobblings, you've got a smarter model.

Speaker 2 And maybe then that thing, or maybe it's two years, and that thing, just like that thing really can do automate 100%.

Speaker 2 And again, they don't need to be doing everything. They don't need to be making coffee.
They don't need to like, you know, maybe there's a bunch of

Speaker 2 tacit knowledge in a bunch of other fields. But

Speaker 2 AI researchers at AI labs really know the job of an AI researcher. And it's in some sense, it's a sort of, there's lots of clear metrics.
It's all virtual. There's code.

Speaker 2 It's things you can kind of develop and train for.

Speaker 1 So, I mean, another thing is, how do you actually manage a million AI researchers? Humans, the sort of comparative ability we have

Speaker 1 that we've been especially trained for is like working in teams. And despite this fact, we have...
For thousands of years, we've been learning about how we work together in groups.

Speaker 1 And despite this, management is a cluster fuck, right? It's like most companies are badly managed.

Speaker 1 It's really hard to do this stuff.

Speaker 1 For AIs,

Speaker 1 the sort of like, we talk about AGI, but it'll be some bespoke set of abilities,

Speaker 1 some of which will be higher than human, some of which will be at human level.

Speaker 1 And so it'll be some bundle, and we'll need to figure out how to

Speaker 1 put these bundles together with their human overseers, with the equipment and everything.

Speaker 1 And the idea that as soon as you get the bundle, you'll figure out how to get millions, like just shove millions of them together and manage them. I'm just very skeptical of.

Speaker 1 Like any other revolution or

Speaker 1 technological revolution in history has been very piecemeal,

Speaker 1 much more piecemeal than you would expect on paper if you just thought about what is the industrial revolution.

Speaker 1 Well, we dig up coal, that powers the steam engines, you use the steam engines to run these railroads, that helps us get more coal out.

Speaker 1 And there's sort of like factorial story you can tell where in like

Speaker 1 six hours, you can be pumping thousands of times more coal. But in real life, it takes centuries often, right? In fact, the electrification, there's this famous study about how to

Speaker 1 initially to electrify factories,

Speaker 1 it was decades after electricity to change from the pulleys and water wheel-based system that we had for steam engines to one that works with more spread out electrical motors and and everything.

Speaker 1 I think this will be the same kind of thing. It might take like decades to actually get millions of AI resources to work together.
Okay, great.

Speaker 2 This is great. Okay, so a few responses to that.
First of all, I mean, I totally agree with the kind of like real-world bottlenecks type of thing.

Speaker 2 I think this is sort of, you know, I think it's easy to underrate.

Speaker 2 You know, basically what we're doing is we're removing the labor constraint.

Speaker 2 We automate labor and we like kind of explode technology, but you know, there's still lots of other bottlenecks in the world.

Speaker 2 And so I think this is part of why the story is that kind of like starts pretty narrow at the thing where you don't have these bottlenecks.

Speaker 2 And then only over time, as we let it, it kind of expands to sort of broader areas.

Speaker 2 This is part of why I think it's like initially this sort of AI research explosion, right? It's like AI research doesn't run into these real-world bottlenecks.

Speaker 2 It doesn't require you to like plow a field or dig up coal. It's just, you're just doing AI research.
The other thing, you know, the other thing about.

Speaker 1 I love how it's like in your model, AI research.

Speaker 1 It's not complicated

Speaker 1 about flipping a burger. It's just AI research.

Speaker 2 I mean, this is because people make these arguments of like, oh, you know, AGI won't do anything because it can't flip a burger.

Speaker 1 Like, yeah, it won't be able to flip a burger, but it's going to be able to do algorithmic progress, you know, and, and then, and then, and then when it does algorithmic progress, it'll figure out how to flip a burger, you know, and then we'll have a burger flip a robot.

Speaker 2 You know, look, the, the, the, sorry, the other thing is about, you know, again, these, the sort of quantities are lower bound, right?

Speaker 2 So it's like, this is just like, we can definitely run 100 million of these.

Speaker 2 Probably what will happen is one of the first things we're going to try to figure out is how to like, again, run, like, you know, translate quantity into quality, right?

Speaker 2 And so it's like, even at the baseline rate of progress, you're like quickly getting smarter and smarter systems, right?

Speaker 2 If we said it was like, you know, four years between the preschooler and the high schooler, right? So I think pretty quickly, there's probably some simple algorithmic changes you find.

Speaker 2 If instead of one Alec Radford, you have 100, you don't even need 100 million.

Speaker 2 And then you get even smarter systems. And now these systems are, you know, they're capable of sort of creative, complicated behavior you don't understand.

Speaker 2 Maybe there's some way to use all this test time compute in a more unified way rather than all these parallel copies.

Speaker 2 And, you know, so they won't just be quantitatively superhuman. They'll pretty quickly become kind of qualitatively superhuman.

Speaker 2 You know, it's sort of like, look, like, you know, you're a high school student, you're like trying to wrap yourself, wrap your mind around kind of standard physics, and then there's some like super smart professor who is like, quantum physics all makes sense to him, and you're just like, what is going on?

Speaker 2 And sort of, I think pretty quickly you kind of enter that regime, just given even the underlying pace of AI progress, but even more quickly than that, because you have the sort of accelerative force of now this automated AI research.

Speaker 1 I agree that over time, you would,

Speaker 1 I'm not denying that ASI is the thing that's possible.

Speaker 1 I'm just like, you know, how is this happening in a a year? Like you've, okay, first of all.

Speaker 2 So I think the story is sort of like, basically, I think it's a little bit more continuous.

Speaker 2 I think already, you know, like I talked about, you know, 25, 26, you're basically going to have models as good as a college graduate.

Speaker 2 And, you know, I don't, I don't know where the unhobbling is going to be, but I think it's plausible that even then you have kind of the proto-automated engineer.

Speaker 2 So there's, I think there is a bit of like a smear, kind of an AGI smear or whatever, where it's like, there's sort of unhobblings that you're missing.

Speaker 2 There's kind of like ways of connecting them you're missing. There's like some level of intelligence you're missing.

Speaker 2 But then at some point, you are going to get the thing that is like 100% automated, Alec Radford. And once you have that, you know, things really take off, I think.

Speaker 1 Yeah. Okay.
So let's go back to the unhobblings. Yeah.
Is there, we're going to get a bunch of models by the end of the year.

Speaker 1 Is there something, suppose we didn't get some capacity by the end of the year? Yeah.

Speaker 1 Is there some such capacity which lacking would suggest that EI progress is going to take longer than you are projecting?

Speaker 2 Yeah. I mean, I think there's, there's two kind of key things.
There's the unhobbling and there's the data wall, right? I think we should talk about the data wall for a moment.

Speaker 2 I think the data wall is,

Speaker 2 you know, even though kind of like all this stuff has been about, you know, crazy AI progress, I think the data wall is actually sort of underrated.

Speaker 2 I think there's like a real scenario where we just stagnate.

Speaker 2 You know, because we've been running this tailwind of just like, it's really easy to bootstrap and you just do unsupervised learning next token prediction.

Speaker 2 It learns these amazing world models, like, bam, you know, great model. And you just got to buy some more compute, you know, do some simple efficiency changes.

Speaker 2 You know, and again, like so much of deep learning, all these like big gains on efficiency have been like pretty dumb things, right? Like, you know, you add a normalization layer, you know,

Speaker 2 you know, you fix the scaling laws, you know, and these already have been huge things, let alone kind of like obvious ways in which these models aren't good yet. Anyway, so data wall, big deal.

Speaker 2 Um, you know, I don't know, some like put some numbers on this, you know, some like you do common crawl, you know, online is like, you know, 30 trillion tokens, llama 3 trained on 15 trillion tokens.

Speaker 2 So you're basically already using all the data. And then, you know, you can get somewhat further by repeating it.

Speaker 2 So there's an academic paper by, you know, Boaz Barak and some others that does scaling laws for this. And they're basically like, yeah, you can repeat it sometime.

Speaker 2 After 16 times of repetition, it's just like returns basically go to zero. You're just completely screwed.
And so, I don't know, say you can get another 10x on data from repetition.

Speaker 2 Say like Lama 3, and GP4, yeah, Lama 3 is already kind of like at the limit of all the data.

Speaker 2 Maybe you can get 10x more by repeating data.

Speaker 2 I don't know. Maybe that's like at most 100x better model than GPT-4, which is like, you know, 100x effective compute from GP4 is not that much.

Speaker 2 If you do half an order of magnitude a year of compute, half an order of magnitude a year of algorithmic progress, that's kind of like two years from GP4.

Speaker 2 So GP4 finished pre-training in 22, you know, 24.

Speaker 2 So I think one thing that really matters, I think we won't quite know by the end of the year, but you know, 25, 26, are we cracking the data wall?

Speaker 1 Okay, so suppose we had three orders of magnitude less data in common crawl on the internet than we just happen to have now. And for decades,

Speaker 1 the internet, other things, we've been rapidly increasing the stock of data that humanity has.

Speaker 1 Is it your view that for contingent reasons, we just happen to have enough data to train models that are just powerful enough at 4.5 level where they can kick off the self-play RL loop?

Speaker 1 Or is it just that

Speaker 1 if it had been three oomes higher, then progress would have been slightly faster.

Speaker 1 In that world, we would have been looking back at like, oh, how hard it would have been to kick off the RL explosion with just 4.5, but we would have figured it out.

Speaker 1 And then so in this world, we would have gotten to to GPT-3 and then we'd have to kick off some sort of RL explosion.

Speaker 1 But we would have still figured it out.

Speaker 1 We didn't just like gluck out on the amount of data we happen to have in the world.

Speaker 2 I mean, three ooms is pretty rough, right? Like three ooms, if less data means like six ooms, smaller, six ooms, less compute model and chill scaling laws.

Speaker 2 You know, that's, I mean, it's basically like capping out of like GPT-2, but I think that would be really rough. I think you do make an interesting point about the contingency.

Speaker 2 You know, I guess earlier we were talking about the sort of like, when in the sort of human trajectory are you able to learn from yourself?

Speaker 2 And so, you know, if we go with that analogy again like if you'd only gotten the preschooler model it can't learn from itself you know if you'd only gotten the elementary schooler model can't learn from itself and you know maybe gp4 you know smart high schooler is really where it starts ideally you have a somewhat better model you know then it really is able to kind of like learn from itself um or learn by itself so i yeah i think there's an interesting i think i mean i think maybe one um less data i would be like more iffy but maybe still doable Yeah, I think I would feel chiller if we had, you know, like one or two.

Speaker 1 It would be an interesting exercise to get probability distributions of HEI contingent done yeah across like

Speaker 1 data yeah okay i the thing that makes me skeptical of this story yeah is that the things it totally makes sense why pre-training works so well yeah these other things they're stories of in principle why they ought to work like humans can learn this way and so on yes and maybe they're true but i worry that a lot of this case is based on sort of first principles with evaluation of how learning happens that fundamentally we don't understand how humans learn and maybe there's some key thing we're missing.

Speaker 1 Yeah. On the sort of sample efficiency, yeah, humans actually, maybe there's

Speaker 1 you say, well, the fact that these things are way less sample efficient in terms of learning than humans are suggests that there's a lot of room for improvement. Yeah.

Speaker 1 Another perspective is that we are just on the wrong path altogether, right? That's why they're so sample inefficient when it comes to pre-training. Yeah.
So

Speaker 1 yeah, I mean, I'm just like,

Speaker 1 there's a lot of like

Speaker 1 first principles arguments stacked on top of each other where you get these unhopplings and then you get to HII.

Speaker 1 Then you, because of these reasons why you can stack all these things on top of each other, you get to ASI. And I'm worried that there's too many steps of this sort of first principles thinking.

Speaker 2 I mean, we'll see, right?

Speaker 2 I mean,

Speaker 2 on the sort of sample efficiency thing, again,

Speaker 2 sort of first principles, but I think, again, there's this clear sort of missing middle. And so, you know, and sort of like, you know,

Speaker 2 People hadn't been trying. Now people are really trying, you know, and so it's sort of, you know, I think often again in deep learning, something like the obvious thing works.

Speaker 2 And there's a lot of details to get right. So it might take some time, but it's now where people are really trying.
So I think we get a lot of signal in the next couple of years.

Speaker 2 You know,

Speaker 2 on unhobbling, I mean, what is the signal on unhobbling that I think would be interesting?

Speaker 2 I think the question is basically, like, are you making progress on this test time compute thing?

Speaker 2 Is this thing able to think longer horizon than just a couple hundred tokens? Right. That was unlocked by chain of thought.

Speaker 1 And on that point in particular,

Speaker 1 the many people who have longer timelines have come on the podcast have made the point that the way to train this long horizon RL, it's not, I mean,

Speaker 1 earlier talking about like, well, they can think for five minutes, but not for longer.

Speaker 1 But it's not because they can't physically output an hour's worth of tokens. It's just really, at least from what I understand, what they say.

Speaker 2 Right. Like even like Gemini has like a million contacts, and the million of contacts is actually great for consumption.
And it solves one important spot hobbling, which is the sort of onboarding.

Speaker 2 problem, right? Which is, you know,

Speaker 2 a new coworker, you know, in your first five minutes, like a new smart high school intern, first five minutes, not useful at all. A month in, you know, much more useful, right?

Speaker 2 Because they've like looked at the monorepo and understand how the code works, and they've read your internal docs. And so being able to put that in context, great, solves this onboarding problem.

Speaker 2 Yeah, but they're not good at sort of the production of a million tokens yet.

Speaker 1 Yeah. Right.
But on the production of a million tokens,

Speaker 1 there's no public evidence that there's some easy loss function where you can GPT4 has gotten a lot better since.

Speaker 2 It's actually, so the GPv4 gains since launch, I think, are a huge indicator that there's like, you know, so you talked about this with John Simon on the podcast.

Speaker 2 John said this was mostly post-training gains. Right.
You know, if you look at the sort of LM sys scores,

Speaker 2 you know, it's like 100 ELO or something. It's like a bigger gap than between Cloud3 Opus and Cloud3 Haiku.
And the price difference between those is 60x.

Speaker 1 But it's not more originic. It's like better in the same chat about math, right?

Speaker 2 Like, you know, it went from like, you know, 40%.

Speaker 1 But the crux is like whether it'll be able to.

Speaker 2 No, but I think it indicates that clearly there's stuff to be done on Hobbling.

Speaker 2 I think, yeah, yeah i think the i think the interesting question is like this time a year from now you know is there a model that is able to think for like you know a few thousand tokens coherently cohesively agentically um and i think probably there's you know again this is i'd feel better if we had an um or two more data because it's like the scaling just gives you this sort of like tailwind right um where like for example tools right tools i think you know talking to people who try to make things work with tools you know actually sort of gpv4 is really when tools start to work and it's like you can kind of make them work with gp3.5 but it's just really tough um And so it's just like having GP4, you can kind of help it learn tools in a much easier way.

Speaker 2 And so just a bit more tailwind from scaling.

Speaker 2 And then, and then, yeah. And does, does, I don't know if it'll work, but it's a key question.

Speaker 1 Okay, I think it was a good place to sort of close that part where we know what the crux is and what the progress,

Speaker 1 what evidence for that would look like. On the AGI to superintelligence,

Speaker 1 maybe it's the case that the gains are really easy right now and you can just sort of let loose an Alec Ratford, give him a compute budget, and he comes out the other end with something that is an additive

Speaker 1 piece, like change this part of the code. This is a compute multiplier, changes the part.

Speaker 1 What other parts of the world,

Speaker 1 like, maybe here's an interesting way to ask this: how many other domains in the world are like this, where you think you could get the equivalent of

Speaker 1 in one year, you just throw enough intelligence across multiple instances,

Speaker 1 and you just come out the other end with something that is remarkably decades, centuries ahead.

Speaker 1 Like you start off with no flight, and then you're the Wright brothers, a million instances of GPD-6, and you come out the other end with Starlink.

Speaker 1 Like, is that your model of how things work?

Speaker 2 I think you're exaggerating the timelines a little bit, but I think a decade's worth of progress in a year or something. I think that's a reasonable prompt.

Speaker 2 So I think this is where, you know, basically the sort of automate AI research comes in because it gives you this enormous headwind on all the other stuff, right?

Speaker 2 So, it's like you know, you automate AI research with your sort of automated Alec Radfords, you come out the other end, you've done another five booms, you have a thing that is like vastly smarter.

Speaker 2 Not only is it vastly smarter, you like, you know, you've been able to make it good at everything else, right? You're like, you're solving robotics. The robots are important, right?

Speaker 2 Because, like, for a lot of other things, you do actually need to try things in the physical world.

Speaker 2 I mean, I don't know, maybe you can do a lot in simulation. Those are the really quick worlds.

Speaker 2 I don't know if you saw the like last NVIDIA GTC, you know, it was all about the like digital twins and just like having all your manufacturing processes and simulation.

Speaker 2 Like, I don't know, like, again, if you have these like, you know, super intelligent, like, cognitive workers, like, can they just make simulations of everything, you know, kind of alpha-float style and then, and then, you know, make a lot of progress in simulation possible?

Speaker 2 But I also just think you're going to get the robots.

Speaker 2 Again, I agree about like there are a lot of real world bottlenecks, right? And so, you know, I don't know.

Speaker 2 It's quite possible that we're going to have, you know, crazy drone swarms, but also, you know, like lawyers and doctors still need to be humans because of like, you know, regulation.

Speaker 2 But, you know, I think, you know, you kind of start narrowly, you broaden, and then the worlds in which you kind of let them loose, which again, because of, I think, these competitive pressures, we will have to let them loose in some degree on, you know, various national security applications.

Speaker 2 I think like quite rapid progress is possible. The other thing, though, is it's sort of, you know, basically in the sort of explosion after, there's kind of two components.

Speaker 2 There's the A, right, in the production function, the like growth of technology. And that's massively accelerated by you.

Speaker 2 Now you have a billion super intelligent scientists and engineers and technicians, you know, superbly competent at everything. You also just automated labor, right?

Speaker 2 And so it's like, even without the whole technological explosion thing, you have this industrial explosion, at least if you let them loose, which is like now you can just build, you know, you can cover Nevada and like, you know, you start with one robot factory as producing more robots.

Speaker 2 And basically, this like just a cumulative process because you've taken labor out of the equation.

Speaker 1 Yeah,

Speaker 1 that's super interesting. Yeah.
Although, when you increase the K or the L without increasing the A,

Speaker 1 you can look at the Soviet Union or China where they rapidly increase inputs.

Speaker 1 And that does have the effect of being geopolitically game-changing, where

Speaker 1 it is remarkable. Like you go to Shanghai over a course of decades.

Speaker 2 You can look at these crazy cities in a decade. Right, right.

Speaker 1 And that's the closest thing to like people talk about 30% growth rates or whatever.

Speaker 1 Even tigers, 10%.

Speaker 2 It's totally possible.

Speaker 1 But without productivity gains, it's not like the Industrial Revolution, where

Speaker 1 from the perspective of you're looking at the system from the outside, your goods have gotten cheaper or they can manufacture more things, but you know, it's not like the next century is coming at you.

Speaker 2 Yeah, it's both. It's both.
So it's both that are important. The other thing I'll say is like all of this stuff, I think the magnitudes are really, really important, right?

Speaker 2 So, you know, we talked about a 10x of research effort or maybe 10, 30x over a decade.

Speaker 2 You know, even without any kind of like self-improvement type loop, you know, we talk the sort of even in the sort of GP4 to AGI story, we're talking about an order of magnitude of effective compute increase a year, right?

Speaker 2 Half an order of magnitude of compute, compute half an order of magnitude of algorithmic progress that sort of translates into effective compute and so um you're doing a 10x a year right basically on your labor force right so it's like it's a radically different world if you're doing a 10x or 30x in a century versus a 10x a year on your labor force so the magnitudes really matter they also really matter on the sort of intelligence explosion right so like just the automated ai research part so you know one story you could tell there is like well ideas get harder to find right algorithmic progress is going to get harder yeah right now you have the easy wins but in like four or five years there'll be fewer easy wins And so the sort of automated AI researchers are just going to be what's necessary to just keep it going, right?

Speaker 2 Because it's gotten harder. But that's sort of, it's like a really weird knife-edge assumption in economics, where you assume it's just.

Speaker 1 But isn't that the equilibrium story you were just telling with why the economy as a whole has 2% economic growth? Because you just proceed on the equilibrium.

Speaker 1 I guess you were saying by the time it gets to the equilibrium.

Speaker 2 The equilibrium here is it's like way faster. At least, you know, and it's at least, and it depends on the sort of exponents, but it's basically it's the increase.

Speaker 2 Like, suppose you need to like 10x effective research effort in AI research in the last, you know, four or five years to keep the pace of progress where you're not just getting a 10x you're getting you know a million X or 100,000 X those just the magnitudes really matter and the magnitudes just basically you know once one way to think about this is that you have kind of two exponentials you have your sort of like normal economy that's growing at you know two percent a year and you have your like ai economy and that's going at like 10x a year and it's starting out really small but sort of eventually it's gonna it's just it's it's it's it's way faster and eventually it's gonna overtake right and even if you have you you can almost sort of just do the simple revenue extrapolation right if you think your ai economy you know that has some growth rate I mean, it's a very simplistic way and so on, but there's, there's this sort of 10x a year process, and that will eventually kind of like, you're going to transition the sort of whole economy from, as it broadens, from the sort of, you know, 2% a year to the sort of much faster growing process.

Speaker 2 And I don't know, I think that's very like

Speaker 2 consistent with historical

Speaker 2 stories. There's this sort of like,

Speaker 2 you know, there's this sort of long-run hyperbolic trend, you know, it manifested in the sort of like sort of change in growth mode in the Austrian revolution, but there's just this long-run hyperbolic trend.

Speaker 2 And now you have this sort of, now you have that another sort of change in growth mode.

Speaker 1 Yeah, yeah. I mean, that was one of the questions I asked Tyler when I had him on the podcast

Speaker 1 is that you do go from the fact that after 1776, you go from a regime of negligible economic growth 2%

Speaker 1 is really interesting. It shows that, I mean, from the perspective of somebody in the Middle Ages or before,

Speaker 1 2% is equivalent to the sort of 10%.

Speaker 1 I guess you're projecting even higher for the AI economy.

Speaker 2 Yeah, I mean, it depends. I think, again, and it's all of this stuff, you know, I have a lot of uncertainty, right?

Speaker 2 So a lot of the time, I'm trying to kind of tell the modal story because I think it's important to be kind of concrete and visceral about it. And I, you know,

Speaker 2 I have a lot of uncertainty basically over how the 2030s play out. And basically, the thing I know is it's going to be fucking crazy.

Speaker 2 But, but, you know, exactly what, you know, where the bottlenecks are and so on, I think that'll be kind of like.

Speaker 1 So

Speaker 1 let's talk through the numbers here. You hundreds of millions of AI researchers.

Speaker 1 So right now, GPD 4.0 turbo is like 15 bucks for a million tokens outputted and a human thinks 150 tokens a minute or something.

Speaker 1 And if you do the math on that, I think it's for an hour's worth of human output. It's like 10 cents or something.

Speaker 1 Now,

Speaker 1 cheaper than a human worker.

Speaker 2 Cheaper than a human worker. Oh, yeah.

Speaker 1 But I can't do the job yet. That's right.
That's right. But by the time you're talking about models that are trained on the 10 gigawatt cluster,

Speaker 1 then you have

Speaker 1 something that is four orders of magnitude more expensive, yeah, inference, three orders of magnitude, something like that. So that's like $100 an hour of labor.

Speaker 1 And now you're having hundreds of millions of such laborers.

Speaker 1 Is there enough compute to do with the model that is a thousand times bigger, this kind of labor?

Speaker 2 Great, okay, great question. So I actually don't think inference costs for sort of frontier models are necessarily going to go up that much.
So, I mean, one historical data point is.

Speaker 1 But isn't the test time sort of thing that it will go up even higher?

Speaker 2 I mean, we're just doing per token, right? And then I'm just saying, you you know, if suppose each model token was the same as sort of a human token thing at 100 tokens a minute.

Speaker 2 So it's like, yeah, it'll use more, but the sort of, if you just, the token calculations is already pricing that in.

Speaker 2 The question is like per token pricing, right?

Speaker 2 And so like GPT-3 when it launched was like actually more expensive than GPT-4 now.

Speaker 2 And so over just like, you know, fast increases in capability gains, inference costs has remained constant.

Speaker 2 That's sort of wild. I think it's worth appreciating.
And I think it gestures at sort of an underlying pace of algorithmic progress.

Speaker 2 I think there's a sort of like more theoretically grounded way to why inference costs would stay constant. And it's the following story, right? So on Chichelle's scaling laws,

Speaker 2 half of the additional compute you allocate to bigger models, and half of it you allocate to more data, right?

Speaker 2 But also, if we go with the sort of basic story of half an order of a year more compute and half an order of magnitude a year of algorithmic progress, you're also kind of like, you're saving half an order of magnitude a year.

Speaker 2 And so that kind of would exactly compensate for making the model bigger. The caveat on that is, you know, obviously not all training efficiencies are also inference efficiencies.

Speaker 2 You know, a bunch of the time they are separately, you can find inference efficiencies.

Speaker 2 So I don't know, given this historical trend, given the sort of like, you know, baseline sort of theoretical reason,

Speaker 2 you know,

Speaker 2 I don't know, I, I think it's not a crazy baseline assumption that actually these models, the frontier models are not necessarily going to get more expensive per token.

Speaker 1 Oh, really? Yeah. Like, okay, that's.

Speaker 2 That's wild. We'll see.

Speaker 1 We'll see.

Speaker 2 I mean, the other thing, you know, maybe they get, you know, even if they get like 10x more expensive, then, you know, you have 10 10 million instead of 100 million.

Speaker 2 So it's like, it's not really, you know, like.

Speaker 1 But, okay, so part of the Intelli Explosion is that each of them has to run experiments

Speaker 1 that are

Speaker 1 GBT4 sized.

Speaker 1 And the result, so that takes up a bunch of compute. Yes.
Then you have to consolidate the results of the experiments and what is a synthesized project.

Speaker 2 I mean, you have a much bigger influence street anyway than you're training. Sure.

Speaker 2 But I think the experiment compute is a constraint.

Speaker 1 Yeah. Okay.

Speaker 1 Going back to maybe a sort of bigger fundamental thing we're talking about here.

Speaker 1 We're projecting

Speaker 1 in a series you say we should denominate the probability of getting to AGI in terms of orders of magnitude of effective compute.

Speaker 1 Effective here, accounting for the fact that there's a compute quote-unquote compute multiplier if you have better algorithms.

Speaker 1 And I'm not sure.

Speaker 1 that it makes sense to be confident that this is a sensible way to project progress. It might be, but I'm just like, like, I have a lot of uncertainty about it.

Speaker 1 It seems similar to somebody trying to project when we're going to get to the moon. And they're like looking at the Apollo program in the 50s or something.
And they're like,

Speaker 1 we have some amount of effective jet fuel. And if we get more efficient engines, then we have more effective jet fuel.

Speaker 1 And so we're going to like the probability of getting to the moon based on the amount of effective jet fuel we have.

Speaker 1 And I don't deny that jet fuel is important to launch rockets, but that seems like an odd way to denominate when you're going to get to the moon. Yeah.

Speaker 2 Yeah. So, I mean, I think these cases are pretty different.
I don't know.

Speaker 2 I didn't, I don't think there is a sort of clear, I don't know how rocket science works, but I didn't, I didn't get the impression that there's some clear scaling behavior with like, you know, the amount of jet fuel.

Speaker 2 I think the,

Speaker 2 I think in AI, you know, I mean, first of all, the scaling laws, you know, they've just held. Right.

Speaker 2 And so if you, a friend of mine pointed this out, and I think it's a great point, if you kind of concatenate both the sort of original Kaplan scaling laws paper that I think went from 10 to the negative nine to 10 petaflop days, and then concatenate additional compute to from there to kind of GPT-4, you assume some algorithmic progress.

Speaker 2 It's like the scaling laws have held

Speaker 2 probably over 15 ooms. I know it's rough calculator, probably maybe even more.
They've held for a lot of ooms.

Speaker 1 They held for the specific loss function, which they're training on, which is training the next token. Whereas

Speaker 1 the progress you are forecasting, which we required for further progress. Yes.
in capabilities. Yeah.

Speaker 1 Specifically, we know that scaling can't work because of the data wall. And so there's some new thing that has to happen.
And I'm not sure whether

Speaker 1 you can extrapolate that same scaling curve to tell us whether these hobblings will also be, like, is this not on the same graph?

Speaker 2 The hobblings are just a separate thing.

Speaker 2 So this is, this is sort of like, you know, it's, yeah. So, I mean, a few few things here, right? Okay.
So

Speaker 2 on the effect of compute scaling, the, you know, in some sense, I think it's like people center the scaling laws because they're easy to explain and the sort of like, why, why is scaling matter?

Speaker 2 The scaling laws like came way after people, at least, you know, like Dario Ilya realized realized that scaling mattered.

Speaker 2 And I think, you know, I think that almost more important than the sort of loss curve is just like, just in general, you make, you know, there's this great quote from Dario on your, on your, on your, on your podcast, it's just like, you know, Ilya was like, you know, the models, they just want to learn.

Speaker 2 You know, you make them bigger, they learn more. And that just applied just across domains, generally, you know, all the capabilities.
And so, and you can look at this in benchmarks.

Speaker 2 Again, like you say, headwind, data wall. And I'm sort of bracketing that and talking about that separately.
The other thing is on hoblinks, right?

Speaker 2 If you just put them on the effective compute graph, these on-hoblings would be kind of huge, right? So, like, I think I- What does it even mean?

Speaker 1 Like, what is on the y-axis here?

Speaker 2 Um, like, say, MLPR on this benchmark or whatever, right?

Speaker 2 And so, you know, like, you know, we mentioned the sort of the LM sys differences, you know, RHF, you know, again, as good as 100x more chain of thought, right?

Speaker 2 Chain of just going from this prompting change, a simple algorithmic change can be like 10x effective compute increases on like math benchmarks.

Speaker 2 I think this is like, you know, I think this is useful to illustrate that on-hoblings are large, but I think they're like, I kind of think of them as like slightly separate things.

Speaker 2 And kind of the way I think about it is that like at a per token level, I think GPU4 is not that far away from like a token of my internal monologue, right?

Speaker 2 Even like 3.5 to 4 took us kind of from like the bottom of the human range to the top of the human range on like a lot of, you know, on a lot of, you know, kind of like high school tests.

Speaker 2 And so it's like a few more 3.5 to 4 jumps. per token basis, like per token intelligence.
And then you've got to unlock the test time.

Speaker 2 You've got to solve the onboarding problem, make it use a computer.

Speaker 2 And then you're getting real close.

Speaker 1 I'm reminded of.

Speaker 2 And again, the story might be wrong, but I think it is strikingly plausible. I agree.

Speaker 1 And so

Speaker 2 I think actually, I mean, the other thing I'll say is: like, you know, I say this 2027 timeline. I think it's unlikely, but I do think there's worlds that are like AGI next year.

Speaker 2 And that's basically if the test time compute overhang is really easy to crack.

Speaker 2 If it's really easy to crack, then you do like four rooms of test time compute, you know, from a few hundred tokens to a few million tokens, you know, quickly.

Speaker 2 And then, you know, again, maybe it's, maybe it only only takes one or two, 3.5 to 4 jumps per token.

Speaker 2 It's like one or two of those jumps per token, plus uses test time compute, and you basically have the proto-automated engineer.

Speaker 1 So I'm reminded of

Speaker 1 Steven Pinker releases his book on,

Speaker 1 what is it, the Better Angels of Our Nature. And it's like a couple years ago or something.
And he says, the secular decline in violence and war and everything.

Speaker 1 And you can just like plot the line from the end of World War II. And in fact, before World War II, then these are just aberrations, whatever.
And basically, as soon as it happened, Ukraine, Gaza,

Speaker 1 everything is like, so

Speaker 1 ASI and crazy DWND.

Speaker 1 I think this is the sort of thing that happens in history where you see a straight line and you're like, oh, my gosh.

Speaker 1 And then just like, as soon as you make that prediction, who is that famous author?

Speaker 2 But you know, my, so yeah, just, you know, again, people are predicting deep learning will hit a wall every year.

Speaker 1 Maybe one year they're right, but it's like gone a long way and it hasn't hit a wall. And we don't have that much more to go.

Speaker 2 And, and, you know, so, yeah.

Speaker 1 Okay, so I think this is a sort of plausible story. And let's just run with it and see what it implies.
Yeah.

Speaker 1 So we were talking in your series, you talk about alignment from the perspective of

Speaker 1 this is not about some doomer scheme to get the point zero and personal probability distribution where things don't go off the rails.

Speaker 1 It's more about just controlling the systems, making sure they do what we intend them to do. If that's the case, and we're going to be in this sort of geopolitical conflict with China,

Speaker 1 And part of that will involve, and what we're worried about, is them making the CCP bots that go out and take the red flag of Mao across the galaxies or something.

Speaker 1 Then

Speaker 1 shouldn't we be worried about alignment as something that, if in the wrong hands, this is the thing that enables brainwashing,

Speaker 1 sort of dictatorial control?

Speaker 1 This seems like a worrying thing.

Speaker 1 This should be part of the sort of algorithmic secrets we keep hidden, right? How to align these models, because that's also something the CCP can use to control their models.

Speaker 2 I mean, I think in the world where you get the Democratic Coalition, yeah. I mean, also just alignment is often dual use, right? Like RHF, you know, it's like alignment team developed.
It was great.

Speaker 2 You know, it was a big win for alignment, but it's also, you know, obviously makes these models useful. Right.

Speaker 2 But yeah, so yeah, alignment enables the CCP bots.

Speaker 2 Alignment also is what you need to get the, you know, get the sort of, you know, whatever, USAIs to like follow the Constitution and like disobey unlawful orders and, you know, like respect separation of powers and checks and balances.

Speaker 2 And so, yeah, you need alignment for whatever you want to do. It's just, it's the sort of underlying technique.

Speaker 1 Tell me what you make of this take. I'm going to stream with this a little bit.

Speaker 1 So

Speaker 1 fundamentally, there's many different ways the future could go.

Speaker 1 There's one path in which the Eleazar type crazy AIs with the nanobots take the future and determine everything to Grey Goo or paperclips.

Speaker 1 And the more you solve alignment, the more that path of the decision tree is circumscribed.

Speaker 1 And then, so the more you solve alignment, the more it is just different humans and the visions they have. And of course, we know from history that things don't turn out the way you expect.

Speaker 1 So it's not like you can decide the future, but it will continue to be a very different thing. It's part of the beauty of it, right? Yeah.
You want the

Speaker 1 error correction. Exactly.

Speaker 1 But from the perspective of anybody who's looking at the system, it will be like, I can control where this thing is going to end up.

Speaker 1 And so the more you solve alignment and the more you circumscribe the different futures that are the results of AI will,

Speaker 1 the more that accentuates the conflict between humans and their visions of the future. And so, in the world where alignment is solved,

Speaker 1 and the world in which alignment is solved is the one, is the world in which you have the most sort of human conflict over where to take AI.

Speaker 2 Yeah, I mean, by removing the worlds in which the AIs take over, then like, you know, the remaining worlds are the ones where it's like the humans decide what happens.

Speaker 2 And then, as we talked about, there's a whole lot of, yeah, a whole lot of worlds and how that could go.

Speaker 1 And I worry, so when you think about alignment and it's just controlling these things,

Speaker 1 just

Speaker 1 think a little forward and there's worlds in which hopefully, you know, human descendants or some version of things in the future merge with super intelligences and they have the rules of their own, but they're in some sort of law and market-based order.

Speaker 1 I worry about if you have things that are conscious and should be treated with rights,

Speaker 1 if you read about what alignment schemes actually are, and then you read these books about what actually happened during the Cultural Revolution, what happened when Stalin took took over Russia?

Speaker 1 And you have

Speaker 1 very strong monitoring from different instances where one, everybody's tasked with watching each other.

Speaker 1 You have brainwashing, you have red teaming, where you have the spy stuff you were talking about, where you try to convince somebody you're on like a defector and you see if they defect with you.

Speaker 1 And if they do, then you realize they're an enemy. And then you take.
And listen, maybe I'm stretching the analogy too far. Yeah.

Speaker 1 But the way, like, the ease with which these alignment techniques actually map on to something you could have read about during like Mao's cultural revolution is a little bit troubling.

Speaker 2 Yeah, I mean, look, I think sentient AI is a whole other topic. I don't know if we want to talk about it.
I agree that like it's going to be very important how we treat them.

Speaker 2 You know, in terms of like what you're actually programming these systems to do, again, it's like alignment is just, it's a technical. It's a technical problem, a technical solution.

Speaker 2 It enables the CCP bots. I mean, in some sense, I think the,

Speaker 2 you know,

Speaker 2 I almost feel like the sort of model and also about talking about checks and balances is sort of, you know, like the Federal Reserve or Supreme Court justices.

Speaker 2 And there's a funny way in which they're kind of this like very dedicated order, you know, Supreme Court justices. And it's amazing, they're actually quite high quality, yeah, right?

Speaker 2 And they're like, you know, really smart people, they really believe in the Constitution, they love the Constitution, they believe in the principles, they have, you know, these, these, these wonderful, um, you know, back, you know, and yeah, they have different persuasions, but they have sort of, I think, very sincere kind of debates about what is the meaning of the Constitution, you know, what is the best actuation of these principles.

Speaker 2 Um, you know, I guess that's good, you know, by the way, recommendation, sort of SCOTUS oral arguments is like the best podcast.

Speaker 2 You know, when I run out of high-quality content, I'm going to try to, I mean, I think there's going to be a process of like figuring out what the Constitution should be.

Speaker 2 I think, you know, this Constitution has worked for a long time. You start with that.
Maybe eventually things change enough that you want edits to that.

Speaker 2 But anyway, you want them to like, you know, for example, for the checks and balances, they like... They really love the Constitution and they believe in it and they take it really seriously.

Speaker 2 And like, look, at some point, yeah, you are going to have like AI police and AI military.

Speaker 2 But I think sort of like, you know, being able to ensure that they like, you know, believe in it in the way that like a Supreme Court justice does or like in the way that like a Federal Reserve

Speaker 2 official takes their job really seriously. Yeah.

Speaker 1 Yeah. And I guess a big open question is whether if you do the project or something like the project.

Speaker 2 I'm sorry, the other important thing is like a bunch of different factions need their own AIs. Right.

Speaker 2 And so it's, I'm really important that like each political party gets to like have their own, you know, and like whatever crazy, you might totally disagree with their values, but it's like, it's really important that they get to like have their own kind of like super intelligence.

Speaker 2 And again, I think it's that these sort of like classical liberal processes play out, including like different people of different persuasions and so on.

Speaker 2 And I don't know, the AI advisors might not make them, you know, wise. They might not follow the advice or whatever, but I think it's important.

Speaker 1 Okay, so speaking of alignment, you seem pretty optimistic. So let's run through the source of the optimism.
Yeah. I think

Speaker 1 you laid out different worlds in which we could get AI. There's one that you think is low probability of next year where GPT-4 plus scaffolding plus unhopplings gets you to AGI.

Speaker 2 Not GPT-4, you know, like

Speaker 1 yeah, but yeah, yeah, yeah, yeah, yeah.

Speaker 1 And there's ones where it takes much longer. There's ones where it's something that's a couple years ago.

Speaker 2 Yeah.

Speaker 1 So

Speaker 1 GPD-4 seems pretty aligned in the sense that I don't expect it to go off the rails. Maybe with scaffolding, things might change.

Speaker 1 Yeah, exactly. So,

Speaker 1 and maybe you keep turning, there's cranks, you keep going up, and one of the cranks gets you to ASI. Yeah.

Speaker 1 Is there any point at which the sharp left turn happens?

Speaker 2 Is it when you start, is it the case that you think plausibly when they act more like agents this is a thing to worry about yeah is there anything qualitatively that you expect to change with regards to the alignment perspective yeah at these cranks so i don't know if i believe in this concept of sharp left turn but i do think there's basically i think there's important qualitative changes that happen between now and kind of like somewhat superhuman systems kind of like early on the intelligence explosion and then important qualitative changes that happen from like early in intelligence explosion to kind of like true super intelligence and all its power and might and um let's talk about both of those and so okay so the first part of the problem is one you you know, we're going to have to solve ourselves, right?

Speaker 2 We have to have to align the like initial AIs and the intelligent explosion, you know, the sort of automated AI Bradford.

Speaker 2 I think there's kind of like, I mean, two important things that change from GPT-4, right? So one of them is,

Speaker 2 you know, if you believe the story on like, you know, synthetic data or L or self-play to get past the data wall, and if you believe this on Hobbling Story, you know, at the end, you're going to have things, you know, they're agents, right?

Speaker 2 Including

Speaker 2 they do long-term plans, right? They have long, long, you know, they're somehow they're able to act over long horizons, right? But you need that, right?

Speaker 2 That's the sort of prerequisite to be able to do the sort of automated AI research.

Speaker 2 And so, you know, I think there's basically, you know, I basically think sort of pre-training is sort of alignment neutral in the sense of like, it has all these representations.

Speaker 2 It has good representations that, you know, as representations of doing bad things, you know, but there's, there's, it's not like, you know, scheming against you or whatever.

Speaker 2 I think the sort of misalignment can arise once you're doing more kind of long horizon training, right?

Speaker 2 And so you're training, you know, again, too simplified example, but to kind of illustrate, you know, you're training in AI to make money.

Speaker 2 And, you know, if you're just doing that with reinforcement learning, you know, it's, you know, it might learn to commit fraud or lie or deceive or seek power simply because those are successful strategies in the real world.

Speaker 2 Right. So maybe, you know, RL is basically it explores.
Maybe it figures out like, oh, it tries to like hack and then it gets some money and that made more money.

Speaker 2 You know, and then if that's successful, if that gets reward, that's just reinforced.

Speaker 2 So basically, I think there's sort of more serious misalignments, kind of like misaligned long-term goals that could arise between now and, or that sort of necessarily have to be be able to arise if you're able to get long horizon system.

Speaker 2 That's one.

Speaker 2 What you want to do in that situation is you want to add side constraints, right? So you want to add, you know, don't lie, don't deceive, don't commit fraud.

Speaker 2 And so, how do you add those side constraints, right? The sort of basic idea you might have is like RHF, right?

Speaker 2 You're kind of like, yeah, it has this goal of like, you know, make money or whatever, but you're watching what it's doing.

Speaker 2 If it starts trying to like, you know, lie or deceive or fraud or whatever, or break the law, you're just kind of like, thumbs down, don't do that. You anti-reinforce that.

Speaker 2 The sort of critical issue that comes in is that these I systems are getting superhuman, right? And they're going to be able to do things that are too complex for humans to evaluate, right?

Speaker 2 So, again, even early on, you know, in the intelligence explosion, the automated AI researchers and engineers, you know, they might write millions, you know, billions, trillions of lines of complicated code.

Speaker 2 You know, they might be doing all sorts of stuff you just don't understand anymore.

Speaker 2 Um, and so, you know, in the million lines of code, you know, is it somewhere kind of like, you know, hacking, hacking, or like exfiltrating itself, or like, you know, trying to go for the nukes or whatever?

Speaker 2 You know, like, you don't know anymore, right? And so, this sort of like you know, thumbs up, thumbs down, pure pure LHF doesn't fully work anymore. Second part of the picture, and

Speaker 2 maybe we can talk more about this. First part of the picture, I think it's going to be like, there's a hard technical problem of what do you do sort of post-RLHF, but I think it's a solvable problem.

Speaker 2 And it's like, you know, there's various things I'm bullish on. I think there's like ways in which deep learning has shaped out favorably.

Speaker 2 The second part of the problem is you're going from your like initial systems, the intelligence explosion to like super intelligence. And it's like many ooms.

Speaker 2 It ends up being like, by the end of it, you have a thing that's vastly smarter than humans.

Speaker 2 I think the intelligence explosion is really scary from an alignment point of view, because basically, if you have this rapid intelligence explosion, you know, less than a year or two years or whatever, you're going, say, in the period of a year from systems where like, you know, failure would be bad, but it's not catastrophic to like, you know, saying a bad word, it's like, you know, it's, it's something goes awry to like, you know,

Speaker 2 failure is like, you know, it extraterrestriated itself. It starts hacking the military.
It can do really bad things.

Speaker 2 You're going less than a year from sort of a world in which like, you know, it's some descendant of current systems and you kind of understand it. And it's like, you know, has good properties.

Speaker 2 There's something that potentially has a very sort of alien and different architecture, right, after having gone through another decade of mal advances.

Speaker 2 I think one example there that's very salient to me is legible and faithful chain of thought, right?

Speaker 2 So, a lot of the time when we're talking about these things, we're talking about, you know, it has tokens of thinking and then it uses many tokens of thinking.

Speaker 2 And, you know, maybe we bootstrap ourselves by, you know, it's pre-trained, it learns to think in English, then we do something else on top so it can do the sort of longer chains of thought.

Speaker 2 And so, you know, it's very plausible to me that like for the the initial automated alignment researchers, you know, we don't need to do any complicated mechanistic interpretability.

Speaker 2 We can just like literally read what they're thinking, which is great. You know, it's like a huge advantage.

Speaker 1 However,

Speaker 2 it's very likely not the most efficient way to do it, right? There's like probably some way to have a recurrent architecture. It's all internal states.
There's a much more efficient way to do it.

Speaker 2 That's what you get by the end of the year.

Speaker 2 You know, you're going this year from like R L H F plus Plus some extension works to like, it's vastly superhuman.

Speaker 2 It's like, you know, it's, it's, it's, it's, it's to us like, you know, you know, an expert in the field might be to like an elementary school or middle schooler.

Speaker 2 And so, you know, I think it's this sort of incredibly sort of like hairy period for alignment. Um, thing you do have is you have the automated AI researchers, right?

Speaker 2 And so you can use the automated AI researchers to also do alignment.

Speaker 1 And so in this world,

Speaker 1 why are we optimistic that the project is being run by people who are thinking? I think, so here's, here's, here's, here's something to think about. Okay.
The

Speaker 1 Open AI

Speaker 1 starts off with people who are very explicitly thinking about exactly these kinds of things.

Speaker 2 Yes. Right? But are they still there?

Speaker 1 No, no, but so here, here's the thing. No, no, even the people who are there, even like the current leadership, it's like exactly these things.

Speaker 1 You can find them in interviews in their blog posts talking about. And what happens is when, as you were talking about, when some sort of trivial,

Speaker 1 and Jan talked about it, this is not just you, Jan talked about it in his tweet thread when there is some uh trade-off that has to be made with we need to do this flashy release this week and not next week because whatever google i o is the next week so we need to get it

Speaker 1 and then the trade-off is made in favor of um uh the the less the more careless decision uh-huh

Speaker 1 when we have the government or the national security advisor the military or whatever which is much less familiar with this kind of discourse isn't naturally thinking in this way about i'm worried the chain of thought is unfaithful?

Speaker 1 And how do we think about the features that are represented here? Why shouldn't we be optimistic that a project run by people like that will be thoughtful about these kinds of considerations?

Speaker 2 I mean, it might not be.

Speaker 2 You know, I agree. I think,

Speaker 2 okay, a few thoughts, right? First of all, I think the private world, even if they sort of nominally care, is extremely tough for alignment. A couple of reasons.

Speaker 2 One, you just have the race between the sort of commercial labs, right?

Speaker 2 And it's like, you don't have any headroom there to like be like, ah, actually, we're going to hold back for three months, like get this right.

Speaker 2 And, you know, we're going to dedicate 90% of our compute to automated alignment research instead of just like pushing the next zoom.

Speaker 2 The other thing, though, is like in the private world, you know, China has stolen your H, China has your secrets. They're right on your tails.
You're in this fever struggle.

Speaker 2 No room at all for maneuver.

Speaker 2 They're like, the way it's like absolutely essential to get alignment right and to get it during this intelligence explosion to get it right is you need to have that room to maneuver and you need to have that clear lead.

Speaker 2 And, you know, again, maybe you've made the deal or whatever, but

Speaker 2 I think you're in an incredibly tough space, tough spot if you don't have this clear lead.

Speaker 2 So I think the sort of private world is kind of rough there. Unlike whether people will take it seriously, you know, I don't know.

Speaker 2 I have some faith in sort of sort of normal mechanisms of a liberal society.

Speaker 2 If alignment is an issue, which we don't fully know yet, but sort of the science will develop. We're going to get better measurements of alignment, you know, and the case will be clear and obvious.

Speaker 2 I worry that there's, you know, I worry about worlds where evidence is ambiguous. And I think

Speaker 2 a lot of the most scary kind of intelligence explosion scenarios are worlds in which evidence is ambiguous.

Speaker 2 But again, it's sort of like, if evidence is ambiguous, then that's the worlds in which you really want the safety margins.

Speaker 2 And that's also the worlds in which kind of like running the intelligence explosion is sort of like, you know, running a war, right? It's like, ah, the evidence is ambiguous.

Speaker 2 We have to make these really tough trade-offs. And you like, you better have a really good chain of command for that.
And it's not just like, you know, YOLOing at, let's go. You know, it's cool.

Speaker 2 Yeah.

Speaker 1 Let's talk a little bit about Germany.

Speaker 1 We're making the analogy to World War II.

Speaker 1 And you made a really interesting point many hours ago at this point.

Speaker 1 We should saw it after, you know, after the marathon.

Speaker 1 The fact that throughout history, World War II is not unique, at least when you think in proportion to the size of the population.

Speaker 1 But these other sorts of catastrophes where a subsignificant portion of the population has been killed off. Yeah.

Speaker 1 After that, the nation recovers and they get back to their heights.

Speaker 1 So what's interesting after World War II is that Germany especially and maybe Europe as a whole, obviously they experienced fast economic growth in the direct aftermath because of catch-up growth.

Speaker 1 But subsequently, we just don't think of Germany as, we're not talking about Germany potentially launching an intelligence explosion and then they're going to get a seat at the AI table.

Speaker 1 We were talking about Iran and North Korea and Russia. We didn't talk about Germany, right? Well, because they're allies, but yeah, yeah, yeah.
But so, what happened?

Speaker 1 I mean, World War II and now it didn't like come back out of the Seven Years' War or something, right?

Speaker 2 Yeah, yeah, yeah. I mean, look, I'm generally very bearish on Germany.
I think in this context, I'm kind of like, you know, it's a little bit, you know, I think you're underrating a little bit.

Speaker 2 I think it's probably still one of the, you know, top five most important countries in the world.

Speaker 2 You know, I mean, Europe overall, you know, it still has, I mean, it's a GDP that's like close to the United States, the size of the GDP. You know,

Speaker 2 and there's things actually that Germany is kind of good at, right? Like state capacity, right? Like, you know, the, you know, the roads are good and they're clean and they're well maintained. And,

Speaker 2 you know, in some sense, the sort of a lot of this is the sort of flip side of things that I think are bad about Germany.

Speaker 2 So in the U.S., it's a little bit like there's a bit more of a sort of wild west feeling to the United States, right? And it includes the kind of like crazy bursts of creativity.

Speaker 2 It includes like, you know,

Speaker 2 political candidates that are sort of, you know, there's, there's a much broader spectrum and, you know, much, you know, like both an Obama and a Trump is somebody you just wouldn't see in the sort of much more confined kind of German political debate.

Speaker 2 You know, I wrote this blog post at some point, Europe's political stupor about this.

Speaker 2 But anyway, and so there's this sort of punctilious sort of rule following that is like good in terms of like, you know, keeping your kind of state capacity functioning,

Speaker 2 but that is also,

Speaker 2 you know, I think kind of, I think there's a sort of very constrained view of the world in some sense.

Speaker 2 You know, and that includes kind of, you know, I think after World War II, there's a real backlash against anything like elite, you know, and, you know, again, no, you know, no elite high schools or elite colleges.

Speaker 2 And sort of,

Speaker 1 why is that the logical?

Speaker 2 Excellence isn't cherished. You know, there's, yeah.

Speaker 1 Why is that the logical intellectual

Speaker 1 thing to rebel against if what if you're trying to overcorrect from the Nazis? Yeah. Was it because the Nazis were very much into elitism?

Speaker 1 What's I don't understand why that's a logical sort of counter reaction?

Speaker 2 I know, maybe it was sort of a counter reaction against the sort of like whole like Aryan race and the sort of that sort of thing.

Speaker 2 I mean, I also just think there was a certain amount in what amount, certain,

Speaker 2 I mean, look, look at sort of World War I, end of World War I versus end of World War II for Germany, right?

Speaker 2 And sort of, you know, a common narrative is that the peace of Versailles, you know, was too strict on Germany. But, you know, the peace imposed after World War II was like much more strict, right?

Speaker 2 It was a complete, you know, I mean, the whole country was destroyed.

Speaker 2 You know, it was, you know, in all the most of the major cities, you know, over half of the housing stock had been destroyed, right?

Speaker 2 Like, you know, in some birth cohorts, you know, like 40% 40% of the men had died.

Speaker 1 Half the population displaced.

Speaker 2 Oh, yeah. I mean, almost 20 million people are displaced, right? Huge, crazy, right?

Speaker 1 You know, like, and the borders are way smaller than the Versailles borders.

Speaker 2 Yeah, exactly. And, and, and, and sort of complete imposition of a new political system and, and, uh, you know, on both sides, you know, and

Speaker 2 yeah, so it was, um, but in some sense, that worked out better than the post-World War I piece, um, where then there was this kind of resurgence of German nationalism.

Speaker 2 And, um, you know, in some sense, a thing that has been a pattern. So it's sort of like it's unclear if you want to wake the sleeping beast.

Speaker 2 I do think that at this point, you know, it's gotten a bit too sleepy. Yeah.

Speaker 1 I do think it's an interesting point about we underrate the American political system. And I've been making the same correction myself.
Yeah.

Speaker 1 There's

Speaker 1 there was this book about Verdun by a Chinese economist called China's Worldview. And overall,

Speaker 1 I wasn't a big fan, but they made a really interesting point in there,

Speaker 1 which was

Speaker 1 the way in which candidates rise up through the Chinese hierarchy for politics, for administration.

Speaker 1 In some sense, that selects for you're not going to get some Marjorie Taylor Greene or somebody running some

Speaker 2 don't get that in Germany either.

Speaker 1 Right. But

Speaker 1 he explicitly made the point in the book that that also means we're never going to get a Henry Kissinger or Barack Obama into China.

Speaker 1 We're going to get like that by the time they end up in charge of the Politburo Bureau, there'll be like some 60-year-old bureaucrat who's never ruffled any feathers. Yeah, yeah, yeah.

Speaker 2 I mean, I think there's something really important about the sort of like very raucous political debate. And

Speaker 2 I mean, yeah, in general, kind of like, you know, there's the sense in which in America, you know, lots of people live in their kind of like own world.

Speaker 2 I mean, like, we live in this kind of bizarre little like bubble in San Francisco and people,

Speaker 2 you know, and, and, um,

Speaker 2 but I think that's important for the sort of evolution of ideas, like error correction, that sort of thing.

Speaker 2 You know, there's other ways in which the German system is more functional.

Speaker 1 Yeah, but it's interesting that there's some major mistakes, right?

Speaker 2 Like the sort of defense spending, right? And, you know, then, you know, Russia invades Ukraine and you're like, wow, what did we do?

Speaker 1 Right.

Speaker 1 No, that's a really good point, right? The main issues, there's everybody agrees, but exactly. Exactly.
Yeah.

Speaker 2 So consensus blob kind of thing. Right.

Speaker 2 And on the China point, you know, just having this experience of like reading German newspapers and I think how much, you know, how much more poorly I would understand the sort of German debate and sort of the sort of state of mind from just kind of afar,

Speaker 2 I worry a lot about, you know,

Speaker 2 or I think it is interesting just how how kind of impenetrable China is to me. It's like, it's a billion people, right? And like, you know, almost everything else is really globalized.

Speaker 2 You have a globalized internet. And I kind of have a sense of what's happening in the UK.

Speaker 2 You know, I probably, even if I didn't read German newspapers, I sort of would have a sense of what's happening in Germany. But I really don't feel like I have a sense of what, like.

Speaker 2 you know, what is the state of mind, what is the state of political debate, you know, of a sort of average Chinese person or like an average Chinese elite.

Speaker 2 And yeah, I think that I find that distance kind of worrying.

Speaker 2 And I, you know, you know, and there's, and, you know, there's some people who do this and they do really great work where they kind of go through the like party documents and the party speeches.

Speaker 2 And it seems to require a kind of a lot of interpretive ability where there's like very specific words and mandarin that like mean we'll have one connotation and not the other connotation.

Speaker 2 But yeah, I think it's sort of interesting given how globalized everything is. And like, I mean, now we have basically perfect translation machines and it's still so, so impenetrable.

Speaker 1 That's really interesting. I've been.

Speaker 1 I'm sort of ashamed almost that I haven't done this yet. Yeah.

Speaker 1 I think many months ago, when Alexey interviewed me on his YouTube channel, I said, I'm meaning to go to China to actually see for myself what's going on. And actually,

Speaker 1 I should. So, by the way, if anybody listening has a lot of context on China, if I went to China who could introduce me to people, please email me.

Speaker 2 You got to do some pods and you got to find some of the Chinese AI researchers, man.

Speaker 1 I know.

Speaker 1 I was thinking at some point. Again, this isn't.

Speaker 1 They cannot speak freely, but you know, I don't know if they can speak freely, but I was thinking of there's, so they have these papers, and on the paper, they will say who's a co-author. Yeah.

Speaker 1 It's funny because

Speaker 1 while I was thinking of just emailing, cold emailing everybody, like, here's my calendar, can you let's let's just talk. I just want to see what, what is the vibe?

Speaker 1 Even if they don't tell me anything, I'm just like, What kind of person is this? Are they? How westernized are they? Yeah.

Speaker 1 Um, but I, as I was saying this, I just remembered that, in fact, Byte Dance, on according to

Speaker 1 mutual friends we have at Google, they cold emailed every single person on the Gemini paper and said, if you come work for Byte Dance, we'll make you an LED engineer. You report directly to the CTO.

Speaker 1 And in fact, this actually, I'm going to go. That's how the secrets go over, right? Right.
No, I'm not sure if you're curious.

Speaker 1 I meant to ask this earlier, but suppose they hired what if there's only 100 or so people, or maybe less, who are working on the key algorithmic secrets.

Speaker 1 If they hired one such person, is all the off of gun that these labs have?

Speaker 2 If this person was intentional about it, they could get along. I mean, they couldn't get the sort of like, I mean, actually, it could probably just also exfiltrate the code.

Speaker 2 They could get a lot of the key ideas. Again, like, you know, up until recently, stuff was published, but, you know, they could get a lot of the key ideas if they tried.

Speaker 2 I think there's a lot lot of people who don't actually kind of like look around to see what the other teams are doing, but you know, I think you kind of can.

Speaker 2 But yeah, I mean, they could. It's scary.

Speaker 1 Right. I think the project makes more sense there where you can't just recruit a Manhattan project engineer and then just get it.

Speaker 2 I mean, and it's like, these are secrets that can be used for like probably every training around the future that'll be like, maybe are the key to the data wall that are like, they can't go on or they can go on, that are like, you know, they're going to be worth, you know, given sort of like the multipliers on compute, you know, hundreds of billions, trillions of dollars, you know, and all it takes is, you know, China to offer $100 million to somebody.

Speaker 2 And it's like, ah, it can work for us. Right.
And then, and then.

Speaker 2 Yeah. I mean, this is, I mean, yeah, I'm, I'm really uncertain on how sort of seriously China is taking AGI.

Speaker 2 Right now, you see, one anecdote that was related to me on the topic of anecdotes, the by another sort of like, uh, you know, kind of researcher in the field was at some point they were at a conference with somebody, a Chinese AI researcher, and he was talking to him, and he was like, I think it's really good that you're here.

Speaker 2 And like, you know, we got to have the international coordination and stuff.

Speaker 2 And apparently this guy said that I'm the kind of most senior-most person that they're going to let leave the country to come to things like this.

Speaker 1 Wait,

Speaker 1 what's the takeaway?

Speaker 2 As in, they're not letting really senior AI researchers leave the country.

Speaker 1 Interesting.

Speaker 2 Kind of classic, you know, Eastern Bloc move.

Speaker 1 Yeah.

Speaker 2 I don't know if this is true, but it's what I heard. That's interesting.

Speaker 1 So I thought the point you made earlier about

Speaker 1 being exposed to German newspapers and also to, because earlier you were interested in economics and law and national security, you have

Speaker 1 The variety in intellectual diet there has exposed you to thinking about the geopolitical question here and what is others talking about AI.

Speaker 1 I mean, this is the first episode I've done about this where we've talked about things like this, which is, now that I think about it, weird, given that this is an obvious thing in retrospect, I should have been thinking about.

Speaker 1 Anyways, so that's one thing we've been missing.

Speaker 1 What are you missing? And national security, you're thinking about, so you can't say national security.

Speaker 1 What like perspective are you probably underexposed to as a result? And China, I guess you mentioned.

Speaker 2 Yeah, so I think the China one is an important one.

Speaker 2 I mean, I think another one would be a sort of very Tyler-Cown-S take, which is like, you're not exposed to how, like, how will a normal person in America

Speaker 2 both like use AI, you know, probably not, you know, and that being kind of like bottlenecks to the fusion of these things.

Speaker 2 And, you know, I'm overrating the revenue because I'm kind of like, ah, you know, everyone on SF is adopting it.

Speaker 2 But, you know, kind of like, you know, Joe Schmo engineer at a company, you know, like, ah, will they, will they be able to integrate it?

Speaker 2 And also the reaction to it, right?

Speaker 2 You know, I mean, I think this was a question again hours ago where it was um um about like you know won't people kind of rebel against this yeah and they won't want to do the project i don't know maybe they will um yeah here's a political reaction that i didn't anticipate yeah

Speaker 1 so tucker carlson has recently ended the joe rogan episode i already told you about this but i'm just gonna tell the story again

Speaker 1 so

Speaker 1 Tucker Carlson is on Joe Rogan. Yeah.
And they start talking about World War II.

Speaker 1 And Tucker says, well, listen, I'm going to say something that my fellow conservatives won't like, but I think nuclear weapons are immoral.

Speaker 1 I think it was obviously immoral that we use them on Nagasaki and Hiroshima.

Speaker 1 And then he says,

Speaker 1 In fact, nuclear weapons are always immoral.

Speaker 1 Except when we would use them on data centers. In fact, it would be immoral not to use them on data centers.
Because look, these people in Silicon Valley, these fucking nerds are making AG

Speaker 1 super intelligence and they say that it could enslave humanity. We made machines to to serve humanity, not to enslave humanity.
And they're just going on and making these machines.

Speaker 1 And so we should of course be nuking the data centers.

Speaker 1 And that is definitely. That's a real horseshoe.
Real horseshoe. And that is definitely not a political reaction in 2024 I was expecting.

Speaker 2 I mean, who knows, man. It's going to be

Speaker 1 crazy. It's going to be crazy.
The thing we learned with COVID is that also the left-right

Speaker 1 reactions that you would anticipate just based on hunches.

Speaker 1 It completely flipped.

Speaker 1 Initially, like, kind of the right was like, you know, it's like so contingent.

Speaker 2 And then, and then, and then, and the left was like, this is racist. And then it flipped, you know, the left was really into the code.
Yeah. Yeah.

Speaker 2 And the whole thing also is just like so blunt and crude.

Speaker 2 And so, yeah, I think, I think, probably in general, you know, I think people are really under, you know, people like to make sort of complicated technocratic AI policy proposals.

Speaker 2 And I think, especially if things go kind of fairly rapidly on the path to AGI,

Speaker 2 you know, there might not actually be that much space for kind of like complicated, kind of like, you know, clever proposals. It might just be kind of much cruder reactions.
Yeah.

Speaker 1 Look, and then also when you mention the spies and national security getting involved and everything, and

Speaker 1 you can talk about that in the abstract, but now that we're living in San Francisco and we know many of the people who are doing the top AI research, it's also a little scary to think about.

Speaker 1 People I personally know and friends with.

Speaker 1 It's not unfeasible if they have secrets in their head that are worth $100 billion or something, kidnappings, assassination, sabotage.

Speaker 2 Oh, their family. Yeah, it's really bad.
I mean, I guess it's to the point on security. You know, like right now, it's just really foreign.

Speaker 2 But, you know, at some point, as it becomes really serious, it's, you know,

Speaker 2 you're going to want the security cards. Yeah.

Speaker 1 Yeah. Yeah.

Speaker 1 So presumably you have thought about the fact that people in China will be listening to this and will be reading your series. Yeah.

Speaker 1 And somehow you made the trade-off that

Speaker 1 it's better to let the whole world know.

Speaker 1 and also including China and make them up to AGI, which is part of the thing you're worried about is China making up to AGI,

Speaker 1 than to stay silent. I'm just curious, walk me through how you've thought about that trade-off.

Speaker 2 Yeah, I actually, look, I think this is a tough trade-off. I thought about this a bunch.

Speaker 2 I think people in the PRC will read this.

Speaker 2 I think, you know, I think there's some extent to which sort of cat is out of the bag. You know, this is like not, AGI being a thing people are thinking about very seriously is not new anymore.

Speaker 2 There's sort of, a lot of these takes are kind of old.

Speaker 2 Or, you know, I've had, I had, you know, similar views a year ago, might not have written it up a year ago, in part because I think this cat wasn't out of the bag enough.

Speaker 2 You know, I think the other thing is,

Speaker 2 I think to be able to manage this challenge, you know, I think much broader swaths of society will need to wake up, right?

Speaker 2 And if we're going to get the project, you know, we actually need sort of like, you know, a broad bipartisan understanding of the challenges facing us. And

Speaker 2 so, you know, I think it's a tough trade-off, but I think the sort of need to wake up people in the United States and the sort of Western world and the Democratic Coalition

Speaker 2 is ultimately imperative. And, you know, I think my hope is more people here will read it than the PRC.

Speaker 2 You know, and I think people sometimes underrate the importance of just kind of like writing it up, laying out the strategic picture.

Speaker 2 And, you know, I think you've done actually a great service to sort of mankind in some sense by, you know, with your podcast.

Speaker 2 And,

Speaker 1 you know, I think it's overall been good.

Speaker 1 Okay.

Speaker 2 So by the way, you know, on the topic of, you know, Germany,

Speaker 2 you know, we were talking at some point about kind of immigration stories. Right.
I feel like you have a kind of interesting story you haven't told. And I think you should tell.

Speaker 1 So

Speaker 1 a couple of years ago, I was in college and

Speaker 1 I was 20. Yeah.
I was about to turn 21. Yeah.

Speaker 2 I think it was. So

Speaker 2 you came from India when you were really young.

Speaker 1 Right. Yeah.

Speaker 1 So until I was eight or I was eight or nine, I lived in India, and then we moved around all over the place. But

Speaker 1 because of the backlog for Indians, the green card backlog.

Speaker 2 Yeah.

Speaker 1 We've been in the queue for like decades.

Speaker 2 Even though you came at eight, you're still on the H1B. Yeah.

Speaker 1 And when you're 21, you get kicked off the queue and you had to restart the process. I'm on my dad's, my dad's a doctor, and I'm on his H1B as it depended.
But when you're 21, you get kicked off.

Speaker 1 And so I'm 20, and it just kind of dawns on me that this is my situation. Yeah.
And And you completely screwed. Right.
And so I also had the experience that my dad,

Speaker 1 we've like moved all around the country. They have to prove that him as a doctor is like, you can't get native talent.
Yeah. So where's the startup? Yeah.

Speaker 1 You just like, so where can you not get native talent?

Speaker 2 And like even getting the H-1B for you would have been like a 20% lottery. So if you're lucky, you're in the state.

Speaker 1 And they had to prove that they can't get native talent, which means like for him, I'm like, we lived in North Dakota for three years, West Virginia for three years, Maryland, West Texas.

Speaker 1 And so it kind of dawned on me, this is my situation. As I turn 21, I'll I'll be like on this lottery.

Speaker 1 Even if I get the lottery, I'll be a fucking code monkey for the rest of my life because this thing isn't going to let up. Yeah.

Speaker 2 Can't do a startup.

Speaker 1 Exactly. And so at the same time, I had been reading

Speaker 1 for the last year, I've been super obsessed with Paul Graham essays.

Speaker 1 My plan at the time was to make a startup or something. I was super excited about that.
And it just occurred to me that I couldn't do this. Yeah.
That like this is just not in the cars for me. Yeah.

Speaker 1 And so I was kind of depressed about it. I remember I kind of just

Speaker 1 was in a daze through finals because I had like, it just occurred to me and I was really like

Speaker 1 anxious about it. Yeah.
And

Speaker 1 I remember thinking to myself at the time that if somehow

Speaker 1 I end up getting my green card before I turn 21,

Speaker 1 there's no fucking way I'm turning, becoming a code monkey. Yeah.
Because the thing that I've

Speaker 1 like this feeling of dread that I have is this realization that I'm just going to have to be a code monkey. And I realize that's my default path.

Speaker 1 If I, if I hadn't sort of made a proactive effort not to do that, I would have graduated college as a computer science student and I would have just done that.

Speaker 1 And that's the thing I was super scared about. Yeah.
So that was an important sort of

Speaker 1 realization for me. Anyway, so COVID happened because of that, since there weren't foreigners coming, the backlog cleared fast.
And by the skin of my teeth, like a few months before I turned 21,

Speaker 1 extremely contingent reasons.

Speaker 1 I ended up getting a green card. Yeah.
Because I got a green card. I could

Speaker 1 the whole podcast, right? Exactly. I graduated from college and I was like bumming around and I got, it was like, I got, I graduated a semester early.

Speaker 1 I'm going to like do this podcast, see what happens. And it was, it had an intended green card.
The best case scenario.

Speaker 2 It was an artifact, you know, and it only existed.

Speaker 2 Yeah. It's actually, because I think it's hard.
It's probably, it's, you know, what is the impact of like immigration reform? Right.

Speaker 2 Like, what is the impact of clearing, you know, like whatever, 50,000 green cards in the backlog? And you're such like an amazing example of like, you know, all of this is only possible.

Speaker 2 And it's, yeah, it's, I mean, it's just incredibly tragic that this is so dysfunctional.

Speaker 1 Yeah, yeah, yeah. No,

Speaker 1 yeah, it's insane.

Speaker 1 I'm glad you did it.

Speaker 2 I'm glad you kind of like, you know, tried the, you know, the, the, uh, the unusual path.

Speaker 1 Well, yeah, but I could only do it.

Speaker 1 Obviously, I was extremely fortunate that I got the green card. I was like,

Speaker 1 I had a little bit of saved up money and I got a small grant out of college thanks to the future fund to like do this for basically the equivalent of six months. And so it turned out really well.

Speaker 1 And then at each time, and I was like, oh, okay, podcast, come on. Like, I wasted a few months on this.
Let's now go do something real. Yeah.
Something big would happen. Yeah.

Speaker 1 I would, Jeff Bezos would, huh?

Speaker 2 I'm kept with her.

Speaker 1 Yeah, yeah, yeah. But there would always be just like the moment I'm about to quit the podcast, something like Jeff Bezos would say that something nice about me on Twitter.

Speaker 1 The alien episodes gets like a half a million views.

Speaker 1 And then now this is my career, but it was sort of very

Speaker 1 looking back on it, incredibly contingent that things worked out the right way. Yeah.

Speaker 2 I mean, look, if the AGI stuff goes down, down, you know, it'll be

Speaker 2 the most important kind of like, you know, source of, it'll be how maybe most of the people who kind of end up feeling AGI are short about it.

Speaker 1 Yeah, yeah, yeah. Also, very much, you're very linked with the story in many ways.
First,

Speaker 1 I got like a $20,000 grant from

Speaker 1 a future fund right out of college. And that sustained me for six months or

Speaker 1 however long it was.

Speaker 2 And without that, I wouldn't have kind of crazy.

Speaker 1 Yeah, 10 grand or what was it?

Speaker 2 No, it's tiny, but

Speaker 2 it goes to show kind of how far small grants can go. Yeah.

Speaker 1 Sort of the immersion ventures, too. Exactly.
The immersion ventures. And the,

Speaker 1 well, the last year I've been in San Francisco, we've just been

Speaker 1 in close contact the entire time and just bouncing ideas back and forth. I've learned just basically the alpha I have.

Speaker 1 I think people would be surprised by how much I got from you, Shulto, Trent, and a couple others.

Speaker 2 I mean, it's been an absolute pleasure. Yeah, likewise.

Speaker 1 Likewise. It's been super fun.
Yeah.

Speaker 1 Okay, so some random questions for you. Yeah.
If you could convert to Mormonism and you could really believe it, would you do it? Would you push the button?

Speaker 2 Well, okay, okay.

Speaker 2 Before I answer that question, one sort of observation about the Mormons. So actually, there's an article that actually made a big impact on me.

Speaker 2 I think it was by McKay Hoppins at some point, you know, in The Atlantic or whatever about the Mormons. And I think the thing he kind of

Speaker 2 you know, and I think he even like interviewed Mitt Romney in it and so on.

Speaker 2 And I think the thing I thought was really interesting in this article was was he kind of talked about how the experience of kind of growing up different, you know, growing up very unusual, especially if you grow up Mormon outside of Utah, you know, like the only person doesn't drink caffeine, you don't drink alcohol, you're kind of weird.

Speaker 2 How that kind of got people prepared for being willing to be kind of outside of the norm later on.

Speaker 2 And like, you know, Mitt Romney, you know, was willing to kind of take stands alone, you know, in his party because he believed, you know, what he believed is true. And

Speaker 2 I don't know, I mean, probably not to the same way, but I feel a little bit like this from kind of having grown up in Germany,

Speaker 2 you know, and really not having like this sort of German system and having been kind of an outsider or something.

Speaker 2 I think there's a certain amount in which kind of, yeah, growing up in an outsider gives you kind of unusual strength

Speaker 2 later on to be kind of like, you know, willing to say what you think.

Speaker 2 And anyway, so that is one thing I really appreciate about the Mormons, at least the ones that grew up outside of Utah. The other thing, you know, the fertility rates, they're good, they're important.

Speaker 1 But yeah, they're going down as well, right?

Speaker 2 This is the thing that really clinched the kind of fertility decline story.

Speaker 1 Even the Mormons.

Speaker 2 Yeah, even the Mormons, right?

Speaker 1 You're like, oh, this is like a sort of a good story. Mormons will replace replace everybody.

Speaker 2 I don't know if it's good, but it's like, at least, you know, at least come on.

Speaker 2 At least some people will retain high. But it's no, no, you know, even the Mormons.

Speaker 2 And sort of basically, once the sort of these religious subgroups have high fertility rates, once they kind of grow big enough, they become, you know, they're too close in contact with sort of normal society and become normalized.

Speaker 2 Mormon fertility rates drop from. I don't remember the exact numbers, maybe like four to two in the course of 10, 20 years.

Speaker 2 Anyway, so it's like, you know, now people point to the Amish or whatever, but I'm just like, it's probably just not scalable.

Speaker 2 And if you grow big enough, then there's just like, you know, the sort of like, you know, the sort of like overwhelming force of modernity kind of gets you. Yeah.

Speaker 2 No, if I could convert to Mormonism, look, I think there's something,

Speaker 2 I don't believe it, right? If I believed it, I obviously would convert to Mormonism, right?

Speaker 1 Because you got to, you got to, but you can choose a world in which you do believe it.

Speaker 2 I think there's something really valuable in kind of believing in something greater than yourself and believing and having a certain amount of faith.

Speaker 1 You do, right? That's your concept. Yeah, yeah.
And,

Speaker 2 you know,

Speaker 2 there's a,

Speaker 2 you know, feeling some sort of duty to the thing greater than yourself.

Speaker 2 And, you know, maybe my version of this is somewhat different. You know, I think I feel some sort of duty to, like, I feel like there's some sort of historical weight on like how this might play out.

Speaker 2 And I feel some sort of duty to like make that go well. I feel some sort of duty to, you know, our country, to the national security of the United States.
And,

Speaker 2 you know, I think that I think you can be a force for a lot of good.

Speaker 1 I,

Speaker 1 Going back to the opening eye thing, just

Speaker 1 the thing that's especially impressive about that is, look, there's people who at the company who have through

Speaker 1 years and decades of building up savings from working in tech, have probably tens of millions, liquid, more than that in terms of their equity. And the person,

Speaker 1 many people were concerned about the clusters and the Middle East and the secrets leaking to China and all these things.

Speaker 1 But the person who actually made a hassle about it, I didn't think hassling people is so underrated.

Speaker 1 I think that one person who made a hassle about it is the 22-year-old who has less than a year of the company, who doesn't have savings built up,

Speaker 1 who isn't like a solidified member of the,

Speaker 1 I think that's sort of like.

Speaker 2 And maybe it's me being naive and, you know, not having a, knowing how big companies work. And, you know, but look, there's a, you know, I think sometimes I'm a bit of a speech deontologist.

Speaker 2 You know, I kind of believe in saying what you think. Yeah.
Sometimes friends tell me I should be more of a speech consequentialist.

Speaker 1 No, I think

Speaker 1 I really think

Speaker 1 the amount of people who, when they have the opportunity to talk to the person, will just bring up the thing.

Speaker 1 I've been with you in multiple contexts, and I guess I shouldn't reveal who the person is or what the context was, but I've just been like very impressed that the dinner begins and by the end, somebody who has a major voice in how things go is seriously thinking about a worldview they would have found incredibly alien before the dinner or something.

Speaker 1 And I've been impressed with it like just like, just give them the spiel and hassle them.

Speaker 2 I mean look I just I think I think I feel this stuff pretty viscerally now.

Speaker 2 I think there's a time, you know, there's a time when I thought about this stuff a lot, but it was kind of like econ models and like, you know, kind of like these sort of theoretical abstractions.

Speaker 2 And, you know, you talk about human brain size or whatever.

Speaker 2 And I think, you know, since

Speaker 2 I think since at least last year, you know, I feel like, you know, I feel like I can see it. Yeah.
You know, and I, I just, I feel it.

Speaker 2 And I think I can like, you know, I can sort of see the cluster that AGI is going to be training out.

Speaker 2 And I can see the kind of rough combination of algorithms and the people that be involved and how this is going to play out.

Speaker 2 And,

Speaker 2 you know, I think, look, we'll see how it plays out. There's many ways this could be wrong.
There's many ways it could go. But I think this could get very real.

Speaker 1 Yeah.

Speaker 1 Should we talk about what you're up to next?

Speaker 2 Sure. yeah.
Okay.

Speaker 1 So you're starting an investment firm. Yep.
Anchor investments from Nat Friedman, Daniel Gross, Patrick Collison, John Collison.

Speaker 1 First of all, why is this the thing to do? You believe the AGI is coming in a few years.

Speaker 1 Why the investment firm?

Speaker 2 Good question. Fair question.
Okay, so I mean, a couple of things.

Speaker 2 One is just, you know, I think we talked about this earlier, but it's like the screen doesn't go blank, you know, when sort of AGI or superintelligence happens.

Speaker 2 I think people really underrate the sort of basically the sort of decade after. You have the intelligence explosion.
That's maybe the most sort of wild period.

Speaker 2 But I think the decade after is also going to be wild. And, you know, this combination of human institutions, but super intelligence, you have crazy kind of geopolitical things going on.

Speaker 2 You have the sort of broadening of this explosive growth. And

Speaker 2 basically, yeah, I think it's going to be a really important period. I think capital will really matter.

Speaker 2 You know, eventually, you know, like, you know, going to go to the stars, you know, going to go to the galaxies.

Speaker 2 So anyway, so part of the answer is just like, look, I think done right, there's a lot of money to be made. You know, I think if AGI were priced in tomorrow, you could maybe make 100x.

Speaker 2 Probably you can make even way more than that because of the sequencing.

Speaker 2 And

Speaker 2 capital matters.

Speaker 2 I think the other reason is just

Speaker 2 some amount of freedom and independence. And I think

Speaker 2 there's some people who are very smart about this AI stuff and who are kind of like see it coming. But I think almost all of them are kind of constrained in various ways.

Speaker 2 They're in the labs, they're in some other position where they can't really talk about this stuff. And

Speaker 2 in some sense, I've really admired sort of the thing you've done, which is I think it's really important that there's sort of voices of reason on this stuff publicly or people who are in positions to kind of advise important actors and so on.

Speaker 2 And so I think there's a, you know, basically the thing this investment firm will be will be kind of like, you know, bring trust on AI. It's going to be all about situational awareness.

Speaker 2 We're going to have the best situational awareness in the business. You know, we're going to have way more situational business than any of the people who manage money in the New York.

Speaker 2 We're definitely going to, you know, we're going to do great on investing. But it's the same sort of situational awareness that I think is going to be important for understanding what's happening,

Speaker 2 being a voice of reason publicly, and sort of being able to be in a position to advise.

Speaker 1 The book about Peter Thiel,

Speaker 1 they had an interesting quote about this hedge fund. I think it got terrible returns.
So this isn't the example.

Speaker 2 That's too bad, right? It's like too theoretical.

Speaker 1 Sure, yeah. But they had an interesting quote that

Speaker 1 it's like basically a think tank inside of a hedge fund. Yeah.
That's what I'm trying to build. Right.
Yeah.

Speaker 1 So

Speaker 1 presumably you've thought about the ways in which these kinds of things can blow.

Speaker 1 There's a very, there's a lot of interesting business history books about people who got the thesis right, but timed it wrong. Yeah.
Where they buy that internet's going to be a big deal. Yeah.

Speaker 1 They sell at the wrong time and buy at the wrong time during the dot-com boom. Yep.
And so they miss out on the gains, even though they're right about the.

Speaker 1 Anyways, yeah.

Speaker 1 What is that trick to preventing that kind of thing?

Speaker 2 Yeah. I mean, look, obviously you can't, you know, not blowing up is sort of like, you know, task number one and two or whatever.

Speaker 2 I mean, you know, I think this investment firm is going to just be betting on AGI, you know, betting on AGI and super intelligence before the decade is out, taking that seriously, making the bets you would make if you took that seriously.

Speaker 2 So, you know, I think if that's wrong, you know.

Speaker 2 firm is not going to do that well. The thing you have to be resistant to is like you have to be able to resist and get one or a couple or a few kind of individual calls, right?

Speaker 2 You know, it's like AI stagnates for a year because of the data wall or like, you know, you got, you got the call wrong on like when revenue would go up.

Speaker 2 And so anyway that's pretty critical you have to get timing right um i do think in general that the sort of sequence of bets on the way to agi is actually pretty critical and i think a thing people underrate so

Speaker 2 all right i mean yeah so like where does the story start right so like obviously the sort of only bet over the last year was in video

Speaker 2 and um you know it's obvious now

Speaker 2 Very few people did it. This is sort of also, you know, a classic debate I and a friend had with another colleague of ours where this colleague was really into TSM, you know, TSMC.

Speaker 2 And he was just kind of like, well, you know, like these fabs are going to be so valuable. And also, like, NVIDIA, there's just a lot of idiosyncratic risk, right?

Speaker 2 It's like maybe somebody else makes better GPUs. And that was basically right, but sort of only NVIDIA had the AI beta, right? Because only NVIDIA was kind of like a large fraction AI.

Speaker 2 The next few doublings would just like meaningfully explode their revenue. Whereas TSMC was, you know, a couple percent AI.

Speaker 2 So you know, even though there's going to be a few doublings of AI, not going to make that big of an impact. All right.

Speaker 2 So it's sort of like the only place to find the AI beta basically was NVIDIA for a while.

Speaker 2 You know, now it's broadening, right? So now TSM is like, you know, 20% AI by like 27 or something, is what they're saying.

Speaker 2 One more doubling, it'll be kind of like a large fraction of what they're doing. And, you know, there's a whole stack.

Speaker 2 You know, there's like, you know, there's people making memory and co-as and power. You know, utility companies are starting to get excited about AI.

Speaker 2 And they're like, oh, it'll, you know, power production in the United States will grow not 2.5%, but 5% over the next five years. And I'm like, no, it'll grow more.

Speaker 2 At some point,

Speaker 2 like like a Google or something becomes interesting. And, you know, people are excited about them with AI because it's like, oh, you know, AI revenue will be, you know, 10 billion or tens of billions.

Speaker 2 I'm kind of like, ah, I don't really care about them before then. I care about it, you know, once it, you know, once you get the AI beta, right?

Speaker 2 And so at some point, you know, Google will get, you know, $100 billion of revenue from AI. Probably their stock will explode.
You know, they're going to become a $5 trillion, $10 trillion company.

Speaker 2 Anyway, so the timing there is very important. You have to get the timing right.
You have to get the sequence right. At some point, actually, I think

Speaker 2 there's going to be a real tailwind to equities from real interest rates.

Speaker 2 right so basically in these sort of explosive growth worlds you would expect real interest rates to go up a lot um both on the sort of like uh you know

Speaker 2 basically both sides of the equation right on the supply side or on the sort of demand for money side um because you know people are going to be making these crazy investments, you know, initially in clusters and then in the robo factories or whatever, right?

Speaker 2 And so they're going to be borrowing like crazy.

Speaker 2 They want all this capital, higher OI.

Speaker 2 And then on the sort of like consumer savings side, right, to like, you know, to give up all this capital, you know, know, the sort of like Euler equation, standard sort of intra-temporal transfer, you know,

Speaker 2 trade-off of consumption.

Speaker 2 It's a standard, right?

Speaker 1 It's a standard.

Speaker 2 Some of our friends have a paper on this.

Speaker 2 You know, if basically, if you expect, you know, if consumers expect real growth rates to be higher, you know, interest rates are going to be higher because they're less willing to give up consumption.

Speaker 2 They're less willing to give up consumption today for consumption in the future. Anyway, so at some point, real interest rates will go up.

Speaker 2 If sort of ADA is greater than one, that actually means equities, you know, higher growth rate expectations mean equities go down because the sort of interest rate effect outweighs the growth rate effect.

Speaker 2 And so, you know, at some point, there's like big, the big bond short. You got to get that right.

Speaker 2 You got to get it right that, you know, nationalization, you know, like, yeah, you got, yeah, anyway. So there's this whole sequence of things.
You got to get that right.

Speaker 1 And the unknown unknowns. Unknown unknowns.

Speaker 2 Yeah. And so you've looked, you've got to be really, really careful about your like overall like risk positioning, right?

Speaker 2 And because, you know, you know, if you expect these kind of crazy events to play out, there's going to be crazy things you didn't foresee.

Speaker 2 You know, you do also want to make the sort of kind of bets that are tailored to your scenarios in the sense of like, you know, you want to find bets that are bets on the tails, right?

Speaker 2 You know, I don't think anyone is expecting interest rates to go above 10%, like real interest rates.

Speaker 2 But, you know, I think there's at least a serious chance of that, you know, before the decade is out.

Speaker 2 And so, you know, maybe there's some like cheap insurance you can buy on that. You know,

Speaker 1 very silly question. Yeah.
In these worlds,

Speaker 1 are financial markets where you make these kinds of bets going to be respected? And

Speaker 1 like, you know, like, is my fidelity account going to mean anything when we have 50% economic growth? Like, who's who's like, we got to respect his property rights?

Speaker 2 The bond short, the sort of 50% economic growth. That's pretty deep into it.
I mean, again, there's this whole sequence of things. But yeah, no, I think property rights will be restructured.

Speaker 2 Again, in the sort of modal world, the project, yeah.

Speaker 2 At some point, there's going to be figuring out the property rights for the galaxies, you know, and that'll be interesting.

Speaker 1 That will be interesting.

Speaker 1 So there's an interesting question about

Speaker 1 going back to your strategy about, well, the 30s will really matter a lot about how the rest of the future goes. Yeah, and you want to be in a position of influence by that point because of capital.

Speaker 1 It's worth considering, as far as I know, but there's probably a whole bunch of literature on this.

Speaker 1 I'm just riffing, but the landed gentry during the before the beginning of the Industrial Revolution, I'm not sure if they were able to leverage their position in a sort of Georgist or pickety-type sense in order to accrue the returns that were

Speaker 1 realized through the Industrial Revolution.

Speaker 1 And I don't know what happened.

Speaker 1 At some point, they just weren't the landed gentry.

Speaker 1 But

Speaker 1 I'd be concerned that even if you make great investment calls, you'll be like the guy who owned a lot of land, farmland before the Industrial Revolution.

Speaker 1 I'm like, the guy who's actually going to make a bunch of money is

Speaker 1 the one with the steam engine. Even he doesn't make that much money.
Most of the benefits are sort of widely diffused and so forth.

Speaker 2 I mean, I think the analogue is like you sell your land you put it all and sort of the you know the the people who are building the new industry um um i think the i mean i think the sort of like real depreciating asset you know for me is human capital right you know yeah no look i'm serious right it's like you know there's something about like you know i don't know it was like valedictorian of columbia you know the thing that made you special is you're smart right but actually like you know that might not matter in like four years you know because it's actually automatable right um and so anyway a friend joked that the sort of investment firm is perfectly edged for me and it's like you know either like AGI this decade, and yeah, your human capital is depreciated, but you've turned that into financial capital, or, you know, like no AGI this decade, in which case maybe the firm doesn't do that well.

Speaker 2 But, you know, you're still in your 20s and you're still

Speaker 1 excellent.

Speaker 1 And what's your story for why AGI hasn't been priced in? The story.

Speaker 1 Financial markets are supposed to be very efficient. It's very, very hard to get an edge.

Speaker 1 Here,

Speaker 1 naively, you just say, well, I've looked at these scaling curves and they imply that we're going to be buying much more compute and energy than the analysts realize yeah shouldn't those analysts be broke by now what's going on

Speaker 2 yeah i mean i used to be a true emh guy yeah i was an economist you know yeah i am

Speaker 2 you know i think the thing i you know changed my mind on is that i think there can be kind of groups of people smart people you know who are you know say they're in san francisco who do just have alpha over the rest of society and kind of seeing the future

Speaker 2 And so, like COVID, right? Like, I think there's just honestly kind of a similar group of people who just saw that and called it completely correctly.

Speaker 2 And, you know, they showed it to the market. They did really well.

Speaker 2 You know, a bunch of other sort of things like that.

Speaker 2 So,

Speaker 2 you know,

Speaker 2 why is AGI not priced in? You know, it's sort of, you know, why hasn't the government nationalized the labs yet, right?

Speaker 2 It's like, you know, this, you know, society hasn't priced it in yet, and sort of it hasn't completely diffused. And, you know, again, it might be wrong, right? But

Speaker 2 I just think sort of,

Speaker 2 you know, not that many people take these ideas seriously yet. Yeah.

Speaker 1 Yeah.

Speaker 1 Yeah.

Speaker 1 A couple of other sort of ideas that I was playing around with with regards to we didn't get a chance to talk about, but

Speaker 1 the

Speaker 1 systems competition. Yeah.
There's a very interesting

Speaker 1 one of my favorite books about World War II is the Victor Davis Hanson uh-huh

Speaker 1 um uh

Speaker 1 summary of everything uh-huh and he explains why the allies made better decisions than the axis why did they and so obviously there were some decisions the axis made that were pretty like blitzkrieg whatever that was sort of by accident though well in what sense that they just had the infrastructure left over well no i mean the sort of i think i mean I don't, I mean, I think sort of my read of it is Blitzkrieg wasn't kind of some like ingenious strategy.

Speaker 2 It was just kind of, it's like more like their hand was forced.

Speaker 2 I mean, this is sort of the very Adam Susian story of World War II, right? But it was, you know, there's sort of this long war versus short war. I think it's actually kind of an important concept.

Speaker 2 I think sort of Germany realized that if they were in a long war, including the United States, you know, they would not be able to compete industrially.

Speaker 2 So their only path to victory was like, make it a short war, right? And that... that sort of worked much more spectacularly than they thought, right?

Speaker 2 And sort of take over France and take over much of Europe.

Speaker 2 And so then, you know, the decision to invade the Soviet Union, it was, you know, it was, it was, um, look, if it was, it was about the Western Front in some sense, because it was like, we've got to get the resources.

Speaker 2 You know, we don't, we're actually, we don't actually have a bunch of the stuff we need, like, you know, oil and so on.

Speaker 2 You know, Auschwitz was actually just this giant chemical plant to make kind of like synthetic oil and a bunch of these things. It's the largest industrial project in Nazi Germany.

Speaker 2 And so, you know, and sort of they thought, well, you know, we completely crushed them in World War I. You know, it'll be easy.
We'll invade them. We'll get the resources.

Speaker 2 And then we can fight on the Western Front.

Speaker 2 And even during the sort of whole invasion of the Soviet Union, even though kind of like a large amount of the sort of, you know, the sort of deaths happened there, you know, know, like a large fraction of German industrial production was actually, you know, like planes and naval, you know, and so on that was directed, you know, towards the Western Front and towards the, you know, the Western allies.

Speaker 1 Well, and then so the point that Hanson was making was.

Speaker 2 By the way, I think this concept of like long war and short war is kind of interesting and with respect to thinking about the China competition, which is like, you know, I worry a lot about kind of, you know, America, the decline of sort of American, like latent American industrial capacity.

Speaker 2 You know, like, I think China builds like 200 times more ships than we do right now

Speaker 2 in some crazy way.

Speaker 2 And so it's like, maybe we have a sort of superiority, say in the non-AI worlds, we have the superiority in military material to kind of like win a short war, at least, kind of defend Taiwan in some sense.

Speaker 2 But if it actually goes on, it's like maybe China is much better able to mobilize, mobilize industrial resources

Speaker 2 in a way that we just don't have the same ability anymore.

Speaker 2 I think this is also relevant to the AI thing in the sense of like, if it comes down to sort of a game about building, right, including like maybe AGI takes the the trillion dollar cluster not the hundred billion dollar cluster maybe or even maybe AGI takes the you know is on the hundred billion dollar cluster but you know it really matters if you can run you know 10x you can you can do one more order of magnitude of compute for your super intelligence or whatever um that you know maybe right now they're behind but they just have this sort of like raw latent industrial capacity to outbuild us yeah and that and that matters both in the run-up to AGI and after right where it's like you have the super intelligence on your cluster now it's time to kind of like expand the explosive growth and you know like will we let the robo factories run wild?

Speaker 2 Like, maybe not, but like, maybe China will. Or, like, you know, will we, will, yeah, will we produce the, how many, how many of the drones will we produce?

Speaker 2 Um, and I think, yeah, so there's some sort of like outbuilding in the industrial explosion that I work with.

Speaker 1 You've got to be one of the few people in the world who is both concerned about alignment, but also wants to make sure that we'll let the robo factories proceed once we get the ASI to beat out China.

Speaker 1 Which is like very interesting.

Speaker 2 It's all part of the picture. Yeah, yeah, yeah.

Speaker 1 Yeah.

Speaker 1 And by the way, speaking of the asis and the robot factories one of the interesting things

Speaker 1 yeah one of the interesting things um

Speaker 1 there's this question of what you do with industrial scale intelligence and obviously it's not chatbots yeah but it's a i think it's very hard to predict

Speaker 1 yeah yeah yeah um

Speaker 1 But

Speaker 1 the history of oil is very interesting, where in the, I think it's in the 1860s that we figure out how to refine oil. So I'm a geologist.

Speaker 1 And so Ben Sandard Oil gets started there's this huge boom yeah it changes american politics entire legislators are getting bought out by oil interest and presidents are getting elected based on the divisions about oil and breaking them up and everything

Speaker 1 and all of this has happened yeah the world has revolutionized before the car has been invented

Speaker 1 and

Speaker 1 I so when the light bulb was invented, I think it was like 50 years after oil refining had been discovered.

Speaker 1 Majority of standard oil's history is before

Speaker 1 the car is invented. It was a kerosene lamp.
Exactly. So it's just used for landing.

Speaker 2 So then they thought oil would just no longer be relevant.

Speaker 1 Yeah, yeah. So there was a concern that standard oil would go bankrupt when

Speaker 1 the label was invented.

Speaker 1 But then

Speaker 1 there's sort of,

Speaker 1 you realize that there's an immense amount of compressed energy here.

Speaker 1 You're going to have billions of gallons of this stuff a year.

Speaker 1 And it's hard to sort of predict in advance what you can do with that.

Speaker 1 And then later on, it turns out, oh, transportation, cars, with

Speaker 1 that's what it could be used for. Anyways, with intelligence, maybe one answer is the intelligence explosion.
But even after that,

Speaker 1 so you have all these ASIs and you have enough compute, especially the compute they'll build

Speaker 2 to run hundreds of millions of GPUs, Wilhelm.

Speaker 1 Yeah, but what are we doing with that? And it's very hard to predict in advance, and I think it'll be very interesting to figure out what the Jupiter brains will be doing.

Speaker 1 So look, there's situational awareness

Speaker 1 of where things stand now.

Speaker 1 And

Speaker 1 we've gotten a good dose of that.

Speaker 1 Obviously, a lot of the things we're talking about now, you couldn't have prejudged many years back in the past.

Speaker 1 And part of your worldview implies that things will accelerate

Speaker 1 because of AI getting into the process.

Speaker 1 Many other things that are unpredictable fundamentally.

Speaker 1 Basically, how people will react, how the political system will react, how foreign adversaries will react.

Speaker 1 Those things will become evident over time.

Speaker 1 So

Speaker 1 the situational awareness is not just knowing where the picture stands now, but being in a position to react appropriately to new information, to change your worldview as a result, to change your recommendations as a result.

Speaker 1 What is the appropriate way to think about

Speaker 1 situational awareness as a continuous process rather than as a one-time thing you realized? Yep.

Speaker 2 No, I think this is great. Look, I I think

Speaker 2 there's a sort of mental flexibility and willing to change your mind that's really important. I actually think this is sort of like how a lot of brains have been broken in the AGI debate, right?

Speaker 2 There's sort of the doomers who actually, you know, I think we're really prescient on AGI and thinking about the stuff, you know, like a decade ago,

Speaker 2 but they haven't actually updated on the empirical realities of deep learning. They're sort of like, the proposals are really kind of naive and unworkable.
It doesn't really make sense.

Speaker 2 You know, there's people who come in with sort of a predefined ideology. There's kind of like, you know, the EX a little bit.

Speaker 2 You know, like they like to shitpost about technology, but they're not actually thinking through it.

Speaker 2 Like, you you know, I mean, either they're sort of stagnationists who think this stuff is only going to be, you know, a chatbot, and so of course it isn't risky, or they're just not thinking through the kind of like actually immense national security implications and how that's going to go.

Speaker 2 And, you know, I actually think there's kind of a risk in kind of like having written this stuff down and like put it online.

Speaker 2 And, you know, there's a, there's, I think this sometimes happens to people as a sort of calcification of the worldview because now they've publicly articulated this position.

Speaker 2 And, you know, maybe there's some evidence against it, but they're clinging to it.

Speaker 2 And so I actually, you know, I want to give the big disclaimer disclaimer on, like, you know, I think it's really valuable to paint a sort of very concrete and visceral picture.

Speaker 2 I think this is currently my best guess on how this decade will go.

Speaker 2 I think if it goes anywhere like this, it will be wild.

Speaker 2 But, you know,

Speaker 2 given the rapid pace of progress, we're going to keep getting a lot more information. And,

Speaker 2 you know, I think it's important to sort of keep your head on straight about that. You know,

Speaker 2 I feel like the most important thing here is that, you know, and this relates to some of the stuff we've talked about, and, you know, sort of the world being surprisingly small and so on. You know,

Speaker 2 I feel like I used to have this worldview of like, look, there's important things happening in the world, but there's like people who are taking care of it, you know, and there's like the people in government, and there's, again, even like AI labs are sort of idealized, and

Speaker 2 people are on it, you know, surely they must be on it, right? And I think just some of this personal experience, even seeing how kind of COVID went, you know,

Speaker 2 people aren't necessarily, there's not some, not somebody else is just kind of on it and making sure this goes well, well, however it goes.

Speaker 2 You know,

Speaker 2 the thing that I think will really matter is that there are sort of good people who take this stuff as seriously as it deserves and who are willing to kind of take the implications seriously, who are willing to, you know, who have situational awareness, are willing to change their minds,

Speaker 2 are willing to sort of stare the picture in the face. And,

Speaker 2 you know, I'm counting on those good people.

Speaker 1 All right. That's a great place to close, Leopold.

Speaker 2 Thanks so much, Tarkash.

Speaker 1 this is excellent

Speaker 1 hey everybody i hope you all enjoyed that episode with leopold there's actually one more riff about german history that he had after a break and it was pretty interesting so i didn't want to cut it out so i've just included it after this outro you can advertise on the show now so if you're interested you can reach out at the form in the description below Other than that, the most helpful thing you can do is just share the episode if you enjoyed it.

Speaker 1 Send it to group chats, Twitter, wherever else you think people who might like this episode might congregate. And other than that, I guess here's this riff on Frederick the Great.

Speaker 1 See you on the next one.

Speaker 2 I mean, I think the actual funny thing is, you know, a lot of the sort of German history stuff we've talked about

Speaker 2 is sort of like not actually stuff I learned in Germany, it's sort of like stuff that I learned after.

Speaker 2 And there's actually, you know, a funny thing where I kind of would go back to Germany over Christmas or whatever, and then suddenly understand the street names.

Speaker 2 You know, it's like, you know, Gneis now and Scharnhorst, and there are all these like Prussian military reformers, and you like finally understood, you know, Sansa and you're like, it was for Frederick, you know, Frederick the Great is this really interesting figure.

Speaker 2 Um, where, um, so he's this sort of, in some sense, kind of like

Speaker 2 gay lover of arts, right? Where he,

Speaker 2 you know, he hates speaking German, he only wants to speak French, you know, he like plays the flute, he composes, he has all the sort of great, you know, artists of his day, you know, over at Sansouci.

Speaker 2 And he actually had this sort of like really tough upbringing where his father was this sort of like really stern sort of Prussian military man. And

Speaker 2 he had had a

Speaker 2 Frederick the Great as a child, as sort of a 17-year-old or whatever, he basically had a male lover.

Speaker 2 And what his father did was imprison his son and then, I think, hang his male lover in front of him. And again, his father was this kind of very stern Prussian guy.

Speaker 2 He was this kind of gay, you know, lover of arts. But then later on, Frederick the Great turns out to be this like, you know, one of the most kind of like, you know,

Speaker 2 successful kind of Prussian conquerors, right? Like, he gets Silesia, he wins the Seven Years' War.

Speaker 2 You know, also, you know, amazing military strategists, you know, amazing military strategy at the time consisted of like, he was able to like flank the army and that was crazy, you know, and that was brilliant.

Speaker 2 And then, and then they like almost lose the Seven Years' War. And at the very end, you know, the sort of the

Speaker 2 Russian Tsar changes and he's like, ah, I'm actually kind of a Prussia stan. You know, I think I'm like, I'm into this stuff.

Speaker 2 And then he lets, you know, lets Frederick the Great loose and kind of let's let's let's let the army be okay. And

Speaker 2 anyway, sort of like a

Speaker 2 yeah, kind of bizarre, interesting figure in German history.