AI Hallucinations (with Stewart Lee and Sarah Wynn-Williams)

This BBC podcast is supported by ads outside the UK.

With the state of today's economy, it's more important than ever to invest in products that last for years to come.

As the seasons shift and get cooler, make sure your closet is stocked with durable layers that stand the test of time from American Giant.

American Giant's clothes work harder and are wearable season after season.

Their greatest hoodie ever made is made from the highest quality materials that are cut and sewn right here in the United States. So you're investing right back in your local community.

Choosing American Giant means taking a stand for American manufacturing and hardworking Americans, something other megacorporations don't care about.

From fleece to knit, all in a range of colors for versatile daily wear, American Giant delivers everyday pieces designed for everyday life.

Feel the difference of quality made-to-last clothes from American Giant. Get 20% off your first order with code Staple20 at AmericanDash Giant.com.

That's 20% off your first order at AmericanDash Giant.com with code STAPLE20.

At the BBC, we go further so you see clearer.

With a subscription to BBC.com, you get unlimited articles and videos, ad-free podcasts, the BBC News Channel streaming live 24-7, plus hundreds of acclaimed documentaries.

From less than a dollar a week for your first year, read, watch, and listen to trusted independent journalism and storytelling. It all starts with a subscription to bbc.com.

Find out more at bbc.com slash unlimited.

Hello, and welcome to Strong Message here from BBC Radio 4, a guide to the use and abuse of political language. I'm Amanda Yucci, and I'm joined once more this week by Stuart Lee.
Hello.

And this week, we're going to look at the word hallucinations, in particular, in reference to AI. That's not me.

I know my initials are AI, but please don't, when we discuss AI in this episode, please do not think we're talking about me, okay?

What could possibly go wrong?

Which is why we're going to be joined by this week's guest, public policy expert and author of Careless People, a Cautionary Tale of Power, Greed and Lost Idealism, a fantastic book about the inner workings of meta, Sarah Wynne Williams.

Great. Welcome.

And I ought to explain in a program where it's all about discussing language, we actually have someone whose language has been taken away from them in that Sarah is not allowed to say anything negative about Meta, the company.

And for that reason, we'll not be speaking about Meta directly in this. That's true, isn't it? That's right.
I've got that accurate. She's nodding.

You are our first gagged guest.

Yes, I'm going to be the most quiet guest we've ever had on the podcast. Very early on when the show first started, I was singing the praises of the book.

You're not allowed to sing the praises of your book, actually.

You're not. You can't mention your own book.
no no but we can wow um

you're not allowed to publicize it but i'd say it's a tremendous christmas gift idea for anyone who's into paranoia and confirmation that everything's as bad as you thought it might be right but you're not allowed to say anything about it i mean maybe i could say don't buy it okay yes

are you are you a source of frustration to your publicist

i i i think i just make everybody's job very difficult yeah or easy yeah

either way yeah because when you say don't buy it, I'm hearing buy it. No, no.
You're hearing don't buy it. Okay, right.
Are you restricted in what you can say about matter?

Right. I have a gag order, so.
You've been booked by mistake.

It's basically going to be a conversation between you two with me just

vibes.

No, no, we've booked you for the irony of the fact.

Some of it's being picked up on little cameras. Are you able to make a disapproving face? No, my face, I have a neutral poker face.

There is no face. I'm making no expression that's how you make your living now is it poker or

i'm communicating nothing okay great no we we booked you for the delicious irony that um an industry that promotes free speech is also one that do you remember the ice man

who used to melt ice do you remember him sorry the ice man who used to melt ice what it was a cabaret actor oh right yeah

in the old days on the circuit used to melt ice live i booked him for resonance radio once and we melted ice on the radio which is really really good that's a bit what this is like And how many listeners did you get?

Oh, seven or eight. Yes.
Before we go into all that, I think we have to just tie up some of the loose knots. From we talked about the budget last week,

various things about the budget. We said that nobody understands anything.
It's all based on forecasts that are, at best, educated guesses rather than clear, defined principles and calculations.

Yeah, although Rachel Reese is now in trouble, isn't she, for

trying to manipulate those educated guesses, which is really interesting.

Yes, it says something about how bad Labour's communications machine is, that Stamer and Reeves are in trouble for not telling everyone that the finances were actually better than everyone expected.

Yeah, yeah.

If a chance to say, I admit it, we're doing rather well. I have to resign.
And that they're being recommended to the ethics committee by a man who used to make hissing sounds at Jewish people. Really?

Oh, yes.

Allegedly. Yeah, oh, yeah, allegedly.
Sorry, I forgot we have to say that. Well, he would say, whether it was his thing, I can't remember.
But it wasn't with intent. No.
We all agreed.

Do you feel you got the measure of the budget in advance of it then, last week? Well, I'm even more confused now than I was then.

When this whole dilemma over what the OBR

said about changing its forecasts and when it had told the government came out, one minister was quoted saying, I don't understand what has gone on. I genuinely don't.
And that's a minister.

And I think there is something weird about the fact that Richard Reeves is being held to account by a department,

the OBR, as we've discussed, can only recommend certain guesswork rather than issue very, very direct and literal estimations.

Because part of the problem was it downgraded its expectations of growth and then upgraded it.

So she's been lambasted for not paying full attention to a department or an organization that keeps changing its mind. Yeah, it's an impossible situation.

But when you ask, how did they manage to get a good financial outlook and still end up in a comms disaster? Yeah. Because 80% of the press have it in for them anyway.
Yes.

I mean, my argument is it also reflects badly on the government's comms. operation in that this is where they found themselves.
Yeah.

How can that happen? When you're sitting on really good figures,

it still ends up that you're in trouble and it's your fault.

I don't know if ever you have been present on when we've ever added anything to the Kiostama metaphor tree. No, but I'm really excited about that.

I think, you know, with advent starting, this is quite a good time to add something to the Kiostama metaphor tree.

We're going to hang on this metaphor, which he said in a speech earlier this week, talking up the budget.

He told his audience he was confident the UK had walked through the narrowest part of the tunnel on the cost of living. And I thought, do tunnels get narrow? the most incredibly badly designed tunnel.

Because the problem with that is you'd drive in with your six-foot-wide vehicle and it would narrow and you'd be trapped in the middle. Yeah.
I mean, Starmer's gone into the wrong tunnel.

If the tunnel is narrowing, it's not a tunnel he should have entered in the first place. I think he was, was he not thinking of an esophagus?

He should have said that. He says,

that's clearer, isn't it? You know, we've swallowed the bitter medicine.

It's stuck in our gullet at the moment.

You can't walk through an esophagus. So you have to say, I have crawled through the narrowest part of the esophagus.
He should have said that. He should have said that.
That's a good statement.

And once these fingers are squeezed through and enter the stomach and are digested, in two years' time, I promise we will be excreting a happier, rosier outlook.

Actually, if he drew all of his metaphors from digestion, that would be really good. Okay, well, I'm glad we've sorttered that one.

Another thing that was being resolved this week was the name of the party, Your Party. Yeah.

I mean, I think at some point in this series, we will look at Your Party, Greens, Lib Dams, the other parties who are now almost on an equal footing in terms of the polls with the main parties.

From the point of view of language, I was just thinking, is it going to be confusing calling your party? your party. Yes, it is, because as soon as they're in parliament,

suppose they were to become the opposition,

they would have to say something like across the benches, they'd have to go, and the problem with your party is that you've done this about, and they go, what do you mean, your party?

Your party, or your party?

It's such a total non-starter. It's idiotic.
It's also, in the interests of balance, it's really great to be able to have a go at the left about something because it's normally.

It's really been accused of being a very right-wing show. I know, yeah.

In the interests of balance, it's it's you'd think people on the left would be ashamed to be squabbling about a name after it's been so brilliantly satirized by Python years ago with the People's Truth of Judy and the Junior Forms.

It's a typical thing of the left that it factionalism tears it apart.

Meanwhile, when the Tory party starts to go that way, a quiet man from the 1922 committee emerges, slaps a few people around, and they go quiet again and present a united front.

And it's just to be arguing about whether it's called your party, our party, pop reliance for the many is such a bad start. I know.

And your party is the worst possible name they could have chosen for the exact reason that you've just pointed out.

It will make any dialogue about the other party involved will collapse immediately under its own semantic weight. And also, you know, for example, the Labour Party is shortened to Labour.
Yeah.

Conservative Party is shortened to Conservative. So is your party going to be shortened to your?

Yeah.

It's a shambles.

I can see questions on television like, what's the your

party stance on unfair dismissal legislation? Or

if someone's in trouble, is it really your

party anymore? Or if they're asking another party, they'll go, what is your position on this? And they'll go, do you mean my position? Or the position of your.

Yes.

And it would be hell. I mean, I tell you what, would be the perfect storm if it was led by mideur.

Right, that's absolutely. That's too late.

Right. These are the light amuse bushes before we get into the heady mix of

AI. Sarah Wynne-Williams.
Hello. We've established there are certain things you can and cannot say.
Correct. You are now working on AI.

You worked for a long time at Meta, as we now call it in international policy you left you're saying nothing i'll carry on i'll keep practicing yeah

i know you're working primarily in ai what's your so

one of the things that i've been doing since leaving is um working on negotiations between the us and china on integration of ai into weapons and whether there should be rules about that right so just something relaxing yes um you know i mean when you said whether there should be rules on that that's that's the thing I pricked my ears up at because is it not assumed that there should be rules on that?

I mean, that is a subject for negotiation and has been a long negotiation because

for, you know, both countries could see the advantage in not having any rules. Right.
But there's rules about chemical weapons and attacks on civilians and

what counts as legitimate targets. So

why have they been so slow on thinking about applying that same logic to the use of AI in weaponry? I mean, I think the challenge is actually that the development of AI has been so fast. Okay, right.

And so that's just outpaced. And at the same time, you know, there have been a few issues between the US and China.
So

getting those rules agreed has become increasingly difficult. I hadn't given that any thought until you mentioned on the way in that you were dealing with it.
And I feel so depressed about that.

Well, yeah, no, no.

Let's not bring up nuclear weapons. Okay, okay.

Well, part of the reason I wanted to discuss this was this very week there was an interview with Jared Kaplan, who's the chief scientist and co-owner of Anthropic, which was one of the major AI companies, who was talking about, you know, what will happen when AI gets to the state where it can improve itself.

And he says, this is in some ways the ultimate risk because it's kind of like letting AI kind of go.

And we're going to have to make a decision about what we do. Now, first of all, I mean,

the kind of the informality of kind of, as in kind of a big decision. But also, why is that a decision? Why is it seen as a decision that has to be made by us whether to allow AI to improve itself?

Should it just not happen? I mean, it's like so much of the stuff we could just decide not to.

And yet there seems to be a real reluctance to even have that conversation

or to discuss any kind of boundary or limit. But what's the headline on that article?

Oh, well, the headline, which would be his headline, be the Guardian's headline, is AI pioneer warns of dangers of humans losing control. And what page is that on? 19, Stuart.

You think that would be nearer the front, wouldn't you?

No, I think it's probably something about

puffins or something. I don't know, making a comeback.

Is there no legislation happening, Sarah, because there's too much money involved? Here, for example, the Labour government seemed to think that making Britain into an enormous battery for

all the machines that need to run AI is the answer to our economic problems. And And there was a Labour Minister for Business on talking about AI growth zones and he was very excited about that.

There were going to be four or five AI growth zones in the UK. And he came on and said how he loved chat BG chat chat.
I always get it wrong.

The chat what's it app and that he'd used it that morning to write the speech that he'd had to give about the AI growth zones, which I thought was awful.

So that's that they kind of think it's the answer. I mean they kind of daren't legislate.
I think that is the thing about AI is projected as the solution to everything. Yeah.

And in fact, even those at the top of the companies don't really know whether that's going to be the case, or am I being overparanoid here?

I mean, I think they're explicitly putting that in the nation's newspapers.

It's just a whole series of top AI executives saying we've got to be careful with what we're doing while expanding their company and asking for more investment in the company.

But I think to some extent it serves their purpose, right? Like, we've got to be careful. This is so powerful.
This is so important. The more that you hype it up, the more that people feel.
So

they are, it's dressed up as a warning, but it's also talking up their own book. Right.
And is that because they just keep, they need to keep growing?

I mean, it's, you know, growth underpins all of these technology companies. That is the beating heart of technology is more.

So what you're saying is talking about the danger of it is a form of marketing as well, because it makes it look like an exciting and powerful thing. Look,

if something feels dangerous and like it shouldn't be happening,

people want to know more.

I say that as someone someone who's not allowed to say more, but you know, that does feel like the surest way to make someone want to know and be part of something is to say, oh, it's a bit dangerous.

You've talked about

the kind of the Minister for Business and the Minister for Technology and so on, and how they're, we discussed this last week, how politicians are bamboozled by technology and by business, and are far happy to just allow companies that deal with very complex systems to hand over a contract to them with freedom to do what they want.

It's insane.

And Peter Thiel, who purportedly thinks that Greta Thunberg is actually the Antichrist, and who, when pressed, doesn't sound entirely convinced that he thinks the human race should endure.

His company, Palantir, which he has chosen to name after an evil orb owned by bad wizards in the Lord of the Rings, that company has been given all our NHS data to manage. So

all our NHS data has been handed over to a man who, when questioned about whether he feels humanity should continue, is not quick to answer in the affirmative. That's right.

Very long pause.

Yes,

I think, for the sake of transparency, I ought to read out what it is.

You don't want to misrepresent Peter Thill's position on the necessity of the continuing existence of humanity.

So it was in an interview, New York Times, the reporter says, do you think people are raising money by pretending that we're going to build a machine god? Is it hype? Is it delusion?

Is it something you worry about? Thiel, um, yeah. Reporter, I think you would prefer the human race to endure, right? Teal, uh.

Reporter, you're hesitating. Theal, well, I don't know.
I would, I would, reporter, this is a long hesitation. Teal, there's so many questions implicit in this.

Reporter, should the human race survive? Teal, yes.

It's such a low ball of a question, right? Do you think humanity should endure? Like, I was like, I try to think of a softer question, and the answer is,

no, no, like

if you wanted to have a soft interview, I don't know how you could be softer. Yes, I mean, yes, reporters are often accused of being unfairly asking the interviewees

leading questions or oversimplistic yes or no questions. But if ever that was a yes or no question, do you want humanity to endure, you know, there should only be one answer to that.

Well, or, well, you know, what's in it for me? Or what are the hours or whatever? Yeah. Yeah.
Yeah. Yeah.

At the BBC, we go further so you see clearer. Through frontline reporting, global stories, and local insights, we bring you closer to the world's news as it happens.

And it starts with a subscription to BBC.com, giving you unlimited articles and videos, ad-free podcasts, the BBC News Channel streaming live 24-7, plus hundreds of acclaimed documentaries.

Subscribe to trusted independent journalism and storytelling from the BBC. Find out more at bbc.com/slash join.

This very week, there was reports that

a company called Fullutoy, who make this bear called Kuma, that was linked up to OpenAI,

that when prompted, Kumar would respond to questions about kink, suggesting bondage and role plays as ways to enhance a relationship, and also the easiest place to find knives.

So,

this was a consumer organization testing what Grimmer would say in front of children if certain questions were put to it.

So, the company, of course, withdrew it, did an audit, and then put out a statement saying, Full toy remains committed to building safe, age-appropriate AI companions for children.

I'm quite, you know, I'm quite a good person to talk to about this because I'm only just starting to get to grips with this. I didn't really, I don't use Chat GBT or any of those kind of things.

And I suppose

all it's doing is it's just taking in the information, isn't it, and regurgitating it?

And it's not, I mean, there was a band I liked in the 80s called The Wild Poppies, and they had an album out in the 80s that was one disc. And I knew that in 2014, a two-disc version came out.

I saw that it was on Amazon, but it wasn't clear which one it was. And you can't ask anyone anything on Amazon because it's all gone to AI.
So I said to the Amazon AI, is this the one-disc version?

And it said, yes. I thought, I'll just check that.
So I said, is it the two-disc version? And it said, yes. Yes.
Because it's found evidence of a two-disc version.

So in this most simple binary form, what AI does, isn't it, is runs with the leading leading question, which is why Musk has had fun with people asking it, is Elon Musk better than Christ or whatever?

And it will go

yes. But if you ask him, is he worse? He might put up a bit of a fight before agreeing.
But it's going to go with the leading question. Yes, that's, I mean, you talked about weapons.

The other thing about AI is, as we've discovered, it's not infallible. It's not foolproof.

Is that acceptance of its fallibility ever factored into how these companies are expanding or how they're selling contracts to governments?

I mean, I don't think anyone's solved the hallucination issue with AI.

And to me, that seems like a very big issue that is yet to be solved. And, you know, there are tremendous possibilities from the technology.

It's not all like, how do we, you know, get teddy bears to recommend knives to children? Like, there are a lot of very,

but to really fulfill them, you have to address the hallucination issue.

You have to be able to have information that is reliable, which there isn't a good answer as to how you're going to address hallucination. And the other thing that hallucination, again,

that is a word that you in your industry use to describe the fact that if there's enough wrong information on the internet, then when the AI transposes it all together, it's just going to come up with wrong information itself.

Right, and it's a really good point because

what you're implicitly saying there is like it's far nicer to say hallucination, which sounds like kind of a pleasant, almost pleasant thing to do, than to say wrong information

or lies.

It's tripping rather than it's incorrect. Right.
And again, that might be the West Coast positive spin. Yeah, that's a great, that's a great thing.
Yeah, that's what it is, isn't it?

Oh, it's hallucinating. Oh, that's nice.
Yeah, right.

It's lying. It's lying about that.
And that's a very different thing. And my concern is like a lot of these companies, for example, Musk with his rockets, when they blow up, he says, yes, no.

it's all about trying out you know uh expecting to fail because we learn from a failure whenever it blows up we get so much data and information on why it blew up so that we can improve it.

So it's all about test it to destruction and from that, the information gleaned will then keep improving our model, our service, or whatever. That's fine.

But you don't want something that could go wrong theoretically applied to weapon systems. Presumably, you want something that's been absolutely guaranteed 100%

accurate, open to inspection. So isn't that the worrying thing that?

That's part of what it is. Have an in-built errancy.
Well,

that's part of what it attracted is exactly the wrong word, but part of the reason that I was drawn to that work is like it's one thing, you know, technology is incredibly buggy, right?

And the solutions are often just as brutal as like turn it off and turn it on again, even when you're working with the most sophisticated engineers who've come out of the best universities around the world.

And a lot of times, they don't even understand why the technology is doing things it shouldn't be doing. And that's one thing if it's, you know, playing, you know, games or some, you know, very,

that's a very, very different thing if it's second strike capability in the South China Sea or, you know, a drone load

that can autonomously kill humans, right? Like, I do think that there are distinctions.

And part of the reason that I wanted to work in this area is it seemed very important to me that there were some discussion for a start, ideally some policies, even better, some rules.

Can I come back a bit on something? Because I'm worried that you talking to us is like a god talking to ants that have just realised something.

But, you know, when you talk about hallucinations, the idea that AI is hallucinating, but with Musk's Grok, it's not just that it's hallucinating, it's that algorithms are skewing it towards certain.

I mean, if you ask Grok about me, it says Stuart Lee is a British comedian renowned for his surreal verbose stand-up. Humour thrives on a shared worldview and his doesn't align with Trump's America.

He has limited mainstream appeal with past tours here hardly packing arenas. I've never been to America.
Few outside niches. Well then you're not packing there every day.
Yeah yeah

that's true. That's true.
That's not packing American arenas. His pledges amplify his ego.
He joins a long line of entertainers who signal virtue while ignoring broader audiences.

I mean, it's really glossed. I mean all those things are true.

But it's very politically glossed, isn't it? Yes. Whereas if you ask ChatBGBT to do a parody of me, it's not it's not doesn't skew right, but it does conclude with the sentence, blah, blah, blah.

I mean, what is a comedian if not an algorithm trained on resentment and Radio 4? It's pretty funny given

I'm on Radio 4 now resenting things.

But, you know, the the as soon as you put in something you know about

Grok, it's terrifying. I had to run in with Grok

because it said it somebody asked it about a a still from a film and it said it was from my film Dave Stalin. And I said, No, it's not.

And it started engaging with me, saying, So, which scene in your film is it in? And I said, No, it's not in my film. Okay, so it must, what is it in was a deleted scene then? And I was saying, No.

And then I lied to it and said, It was a deleted scene, it was from a dream sequence. It was a musical dream sequence.
It didn't fit the tone of the rest of the movie.

And they went, All right, right, great, great to hear it from me. I said, No, I'm lying to you.
It's not in my film. And I went, So, where in your film is it then?

It just couldn't accept that I was right and it was wrong. Right, right.
I I don't know what that says other than...

Well, what it says is that you know the actual fact and you can't defeat, you can't correct it,

the machine.

But is Grok an outlier for those reasons that Stuart has said? I mean, I think, no, I mean, sycophancy has been an issue across a bunch of different AI models. And,

you know, OpenAI came out and said that they were toning down the sycophancy.

And there was a lot of concern around, like, actually, is that going to impact growth and usage?

Because are people who are using AI for things like therapy, are they going to use it less if they're not hearing such nice things and that they don't have this compliant sycophantic?

And I think some, I mean, I think some of the stuff with Grok, like the sycophantcy around Musk got very extreme, sort of saying he was fitter than LeBron James. And, you know,

so

it's actually said this thing this week about who is funnier, Elon Musk or Jerry Seinfeld?

And Grok says, Elon Musk takes the crown for funnier because his tweets blend sharp absurdity with world-altering stakes. Seinfeld masters precise observational comedy about life's quirks.

Yet Elon's chaotic, meme-driven style lands broader, unexpected punches. Humour evolves, but Musk's wit disrupts norms in ways Seinfeld stand-up entertains without appending them.

It's interesting that he's done that thing that all the tech people talk about, disrupting norms. Yes.

And he thinks the purpose of comedy is to disrupt norms in the way that he would talk about doing it to business. It's really interesting.

What's that phrase that we had at the start about world changing or something? World altering stakes. World altering stakes.

It's kind of defining, it's assuming that comedy is, as well as making people laugh,

is about world-altering stakes. When Tommy Cooper comes on, and does he's dead now, obviously.
He died on stage. He died on stage.

But he didn't sort of go you know his kind of aha is it there

you see that uh it wasn't meant to confront our assumptions about society in no way no no no no it wasn't meme driven certainly no certainly not yeah

it's it took jesus three days to rise from the dead would elon musk have figured out a way to do it faster what do you reckon and grok says elon optimizes timelines relentlessly

jesus didn't do that did he he didn't optimize timelines relentlessly

so elon would like to engineer a neural backup and rapid revival pod to cut it to hours. But resurrection is a divine hack beyond physics.
Jesus nailed the miracle without venture capital.

Three days sets the bar high. Faster might have sparked theological debates on overlooking it and overclocking eternity.
I mean, it's quite funny by its technical language. Yeah.

Can we just go back to the kind of

stuff that? Yeah. Well, I mean, we quoted Peter Till saying you're having to think about whether humanity was worth saving.
Do they understand how human beings work?

Are you not allowed to work?

they understand how humans work in as much as they know how to keep us online yes by by drip feeding him exciting little bits of information and then so they understand our addictive side or dopamine uh yeah hit

hit driven side but that's not the all of our nature is it it's it's a it's a specific aspect of our nature yeah i think if you basically know that humanity's biggest value is to be farmed for data, then maybe you have to stop thinking about people as people and just think about them as farmable machines.

I quoted before, but it's worth quoting again Alex Karp's statement to his shareholders several years ago, who's one of the bosses at Palantir,

to his shareholders saying, Palantir is here to disrupt and make our institutions we partner with the very best in the world, and when it's necessary, to scare enemies and on occasion kill them.

We hope you are in favor of that. We hope you're enjoying being a partner.
That was to your share.

It's that kind of, there could be a pop, you know, there is every opportunity there to say, you know, whether or not you agree with Palantir and what it does, there is every opportunity to say, in conflict situations, it's messy, people get hurt, but we're out to protect our national interests or something.

You know, you can, you can see the language there,

whether you agree with that or not, but the kind of the coldness in which killing becomes

part of your metric.

Do you think that there's a kind of bloke, I mean, that was the nerd at school that becomes powerful in the tech industry that enjoys naming his company after an evil orb from Lord of the Rings and enjoys saying they have to be killed for a joke?

Is it kind of ironic almost that they do they like taking on these

are you can you is that not something no okay you can't talk about that is it is it is it I think what is uh of concern to people is that a lot of these um companies when they expand exponentially become so major they're more powerful or more valuable than major economies.

And therefore, a lot of power and influence is invested in one or two individuals. So

it's therefore natural that we try to work out what these individuals are like.

Is the question of legislation,

regulation, morality and ethics, are these annoyances to them or are they something that they're prepared to work with?

That is the response of citizens is policy.

And the problem is it's not sexy or funny,

but

it does matter.

And I think part of what we do when we think of technology companies as like the Death Star or Death Vader is that we make it seem impossible. We make it seem like it can't be done and that nothing,

it's all too big. It's all too far away.

And it's not. And how about legislators? Do you find them,

do they fully understand what it is they're grappling with? I mean, a lot of the time, no, but that like that's perfectly understandable, right?

Like, I didn't understand lots until I came in here today.

Now I'm

really frightened.

I'm not here to frighten you. No, this is normally really funny, but it's like a horrible

chill wind.

I'm sorry, I'm sorry. No, don't worry.
I'll go back to being quiet. I'll go back to being quiet.
I'm really grateful.

But, you know,

it's meant to be the other way, which is like,

the stuff is literally in your hands. You literally hold the technology and you can decide how you use it.
You know, I think we've sort of got into this perspective of it feeling too overpowered.

And it's not. Okay, but how do we, I mean, you said a lot of politicians don't understand.
How do we get to that point where they do and therefore know what they need to do to make it work?

Yeah, so I think that is, you know, you talked before about, you know, data centers and politicians being like, okay, we'll just turn, you know, this country into a giant data center.

And I have to believe that they've never actually been in a data center because there is no way if you have been in a data center that you think it's going to be a huge generator of jobs um it you know it's it's a silent warehouse right um so

again it's it's it's gonna i i hate the answer but it comes back to us right asking the questions making sure like it's all about having that sort of accountability and responsibility pushed up

well okay on a practical level all right i I tell my children not to use a chat GBT or AI for anything, right? And I, and they're doing pretty well.

And I, because I've, and I can show them that it told my friend, a film director, Bernard Mann, was coming over here, and he thought, I wonder what Stu's up to.

And he looked me up on chat GBT and it said, I was dead, right? So

I can show them that. And so I can sort of don't use it for your essays.
It says, I'm dead, right? But I don't have any control over my nieces and nephews or over my sisters. You're a hallucination.

I am. Yeah, yeah, I am a hallucination.

I'm trying to say to people, you've got to fact check. Don't you like, you know, but they just think you're like some stupid old fart now.
It's gone. It's done.
It's done.

And when the politician, when the labor minister came on, the business guy, and said that he'd written some of the questions, he'd written his speech using ChatGP, this is what I thought.

What do we pay you for?

We pay you to, you've got that job to use your own brain and your own thoughts, not to defer it to what essentially is going to be data harvested by someone with its own business interests as well.

Yeah, I think that's it.

It's politicians who sound impressive because they say they have a keen interest in business.

But that doesn't mean to say that they actually have been in business and understand the business of AI and the business of tech.

My favorite quote about, it was Peter Kyle, who was, I think, the science and technology minister, but he's now moved on.

But my favorite quote about him was from Marina Howe to describe Peter Kyle as someone who always comes off as a man playing a businessman in a play about business,

which is that thing of there are certain politicians who like to display themselves as the people who do understand what's going on. And those are the ones that I don't know if you've met them.

Are you convinced they really do understand the ins and outs of AI and where it's going?

The optimistic take on what both of you are saying is that eventually there's going to be a premium placed on people who are not relying on this technology.

Real ale. Right.
Like

eventually the value.

You know, the value of

the human is going to exponentially increase. I mean,

that's what you have to do. Okay, I believe that is true.
Yes, I believe that's true, right? Like there is a logic to it. And it should, if anything, make us feel a lot better.

My daughter last night showed me an AI-generated American rock song about

what a great bloke Charlie Kirk was.

Because obviously, probably no musician is going to write that, but it sounded like a sort of soulful power ballad about that.

But I think you can tell, I think you can tell with music, and

I prefer to see models in films that have been handcrafted

than CGI, because I like to see the human element at work.

But I wonder whether the premium will be, will have the same premium as sort of organic food does, and whether 90% of the population will be eating the chicken nuggets of AI-generated content, whereas the elite will go and listen to someone playing an acoustic guitar beautifully.

You know, is it going to be, and it will cost 10 times more, is that what's going to happen? Will there be an economic premium placed on human-generated content?

I go really sorry to interrupt, but we're going to need to wrap it up in a moment. Oh, right, we are.

We can't wrap it. I don't know the answers.
I'm still frightened. I mean, I think we can't stop until

I think we can't leave from. We've got to sort everything.
I think we have to come up with our own ecosystem.

Can I quote?

I think we like to think there's good people involved in all this technology and and bad people, but they're all so interconnected. This AI company's not so bad.
Okay. Sam Altman, the CEO of OpenAI,

the parent company of ChatGBT, met his husband in Peter Thiel's hot tub at 3am.

Small world.

And that's all real, you see. Yeah.
That's the value of human contact.

That's where you should wrap it up, because you've managed to make it sound optimistic in some way. Okay, before we go, we're going to be be doing a listener's mailbag episode of the new year.

So, we want your most hated political phrases or any thoughts on things that we should cover or indeed have covered. Please send your emails to strongmessage here at bbc.co.uk.

Thanks for listening to Strong Message Here. I'll be back next week.
And why not go and listen to some of our episodes from our back catalogue?

We've got episodes on the climate, the tepid bath, of managed decline, and even weird Turkish barber shops. Just search for Strong Message here and subscribe on BBC Sounds.
Goodbye. Goodbye.

Artificial intelligence is going to change everything from healthcare to transport, from social media to every search you make online, and it's already all around us. It sure is.

So, if you want to get to grips with what all of this means for you, then the artificial human from BBC Radio 4 is your essential guide to understanding this latest technological advance.

I'm Alex Kratoski. And I'm Kevin Fong.
And in every episode, Alex and I are here to get answers to your questions about AI and even ask a few of our own.

Whether that's can AI make me fitter or why do I feel so sad when my favorite chatbot gets an upgrade? Kevin, do you have a favorite chatbot? Alex, as you well know, I love all my chatbots equal.

Well, that is the artificial human. Listen now on BBC Sounds.

At the BBC, we go further so you see clearer.

With a subscription to bbc.com, you get unlimited articles and videos, ad-free podcasts, the BBC News Channel streaming live 24/7, plus hundreds of acclaimed documentaries.

From less than a dollar a week for your first year, read, watch, and listen to trusted independent journalism and storytelling. It all starts with a subscription to bbc.com.

Find out more at bbc.com/slash unlimited.