Curing All Disease with AI with Max Jaderberg

49m
Can AI help us model biology down to the molecular level? Neil deGrasse Tyson, Chuck Nice, and Gary O’Reilly learn about Nobel-prize-winning Alphafold, the protein folding problem, and how solving it could end disease with AI researcher, Max Jaderberg.

Listen and follow along

Transcript

Tron Aries has arrived.

I would like you to meet Aries, the ultimate AI soldier.

He is biblically strong and supremely intelligent.

You think you're in control of this?

You're not.

On October 10th.

What are you?

My world is coming to destroy yours, but I can help you.

The war for our world begins in IMAX.

Tron Aries, suited PG-13.

Maybe inappropriate for children under 13.

Only in theaters, October 10th.

Get tickets now.

I'm gonna put you on, nephew.

All right, Uncle.

Fuck on the McDonald's.

Can I take your order?

Miss, I've been hitting up McDonald's for years.

Now it's back.

We need snack wraps.

What's a snack wrap?

It's the return of something great.

Snack wrap is back.

So AI was not satisfied just whooping our ass in chess and in Jeopardy and everything else where it looks like brains mattered.

It's now taken over our physiology.

Well, no, you pointed it in a good direction, aimed it at a good place, and we're getting somewhere.

To solve our diseases.

Yeah, so now it's going to cure us of all disease

before it makes us its slaves.

Because we need healthy slaves.

All that and more coming up on Star Talk.

Welcome to Star Talk,

your place in the universe where science and pop culture collide.

Star Talk begins right now.

This is Star Talk, Special edition neil degrasse tyson your personal astrophysicist special edition means we've got gary o'reilly in the house gary hi neil former soccer pro apparently yeah and soccer announcer yes yes absolutely and you still do that don't you i do chuck nice baby hey announcing that i know nothing about soccer You're in my club then.

Announcing that you are American.

American dove going as real football.

Violence.

So we're talking about AI today.

That's a favorite topic.

We revisit that often.

Only the future of the entire world.

AI as it matters in biology.

Oh, wow.

Now that's a big deal.

I know.

I know.

Because people are thinking about composing your term paper

or winning a chess.

But it's got a whole frontier ready to be explored.

Yeah.

And so tell me what you and your producers cooked up today.

Okay.

So we've been on the case to get these guys involved for some time, but they are so busy.

So here we go.

I'll say it.

I'm made of proteins.

Yes.

You're made of proteins from strings of amino acids that fold into shapes that put

all together form us.

But there's a fundamental problem in biology

that has implications for all of medicine.

How do these proteins fold up?

Oh.

For this solution, we looked at AI and a Google DeepMind tool called AlphaFold.

The second iteration of AlphaFold 2 won the Nobel Prize in Chemistry last year for answering this very question.

Who knew AI was smart, huh?

Appreciate AI win all the Nobel Prizes.

Yeah, let's give it to her now.

Just park them all up.

Now, the isomorphic labs together with Google DeepMind developed and released AlphaFold 3.

Yes, we're on the third iteration.

And that was last year, and applied these new AI models for drug discovery.

Oh, that's great.

All right, so think this through.

Could our next generations of treatments be computer generated?

Oh, yeah.

Oh, by the way, Neil, let's introduce our guest.

I will.

We've got Max Yaderberg.

Did I pronounce that correctly?

Yeah, you got it.

You got it right.

Let's hear you say it.

Let's hear what you say.

Me say it.

Yeah.

Max Yoderberg.

Oh, that's exactly what I said.

That's what I got.

Yeah.

He was practicing.

He was practicing.

So you studied AI at Oxford?

That's right.

That's right.

And now I hear that's a community college?

Oxford Community College.

That's exactly right.

In Oxford, England.

Yes.

so uh specialized in deep learning algorithms i got your little bio here for understanding images nice yeah that was a big advance when when a search can go into an image i thought you know i died and gone to computer heaven where that started yeah yeah i mean you know this this was 10 15 years ago back before ai was cool right um right where you know you talk about ai and it's something from a sci-fi book but understanding images and videos was like the big thing at that point in time.

We couldn't actually do that very well.

I searched my 9,000 images on my computer for the word telescope.

Yeah.

And it found telescope written in Chinese

on a photo taken at an angle in one of my images.

Wow.

When I was visiting China.

Yeah, this is it.

During my PhD, we took all of the BBC's back catalog.

No.

And we ran...

my algorithm across it and created a search engine so you could pull up footage from decades ago that had this text or these objects.

and so you know seriously

if you're doing interesting when you do that do you tie so when the ai is looking at an image it's not seeing the image the way we do we're not even seeing a whole image our brains we're we're really just intuiting an image when we see it as right exactly that's how we do it but the ai it's like a holistic process of holistic processing ai actually sees the image and what it's seeing is pixels.

That's right.

And really all it's doing is just, oh, this pixel, this pixel, this pickle in this arrangement, that's this image.

So do you tie that to language and that's how we search?

Or is the search just the AI knows the actual image itself?

This was like the big breakthrough back then when I was doing my PhD.

And this is what deep learning as well is all about.

You can imagine if you have

this image full of pixels, how do you actually code up how to read text from there?

How do you tie it to the language of the text?

It's unimaginably hard to code that up by hand.

So instead, what you do is you put these, what they're called, neural networks.

They look at all of the pixels of the image, and you give it lots and lots of examples of images that have somewhere it's got the text in it, and you tell the neural network what the text is.

I see.

And the neural network, through lots and lots of training, starts to work out its internal algorithm to extract the information from these pixels, piece it all together, and spit out the actual text or spit out what the objects are.

Wow.

So you're currently chief AI officer at isomorphic labs.

This is a biology place.

That's right.

Do you have any biology in your background?

Formal biology?

No.

No, okay.

No.

So they want you for your AI.

That's right.

That's right.

So I was at a place called DeepMind beforehand.

Oh, Google.

Google Deep Mind.

Exactly.

I was there for a long time.

I absolutely love this core AI technology called deep learning.

That's what I've been developing.

my whole career so far.

At DeepMind, we were working on some crazy stuff, you know, learning to play chess and go and beating top professionals at games like starcraft you know back then it was about

yeah and and because the world didn't know what ai was so we were trying to prove that this was even a thing right seems crazy now but back then it was just proving that this was actually a real thing but at the core you know i love this technology i want to see it have profound impact on our world and i was thinking these things where it begins yep

it's always where the terminator starts it's always the innocent dreamer

who says this can change the world for such good and it's in my closet now.

Would you like to see it?

And then it's always like some evil businessman who's just like, with my weather machine, I will one day rule the world.

You know, so

apart from that, carry on.

Yes.

The good thing is there's some pretty strictly good applications of AI that we can drive.

And, you know, Demis Asabis started Isomorphic Labs, spinning that out of Google DeepMind to really think how can we apply AI to actually completely solve all disease.

Okay, so it has genetic links back to its origin story within Deep Mind.

I feel better about that now.

Okay.

You happy?

Yeah, I'm happier now.

Yes.

So I moved over as part of that founding team to head up AI in this space.

And it's been about three and a half years now.

It's been a crazy journey, but it's fascinating.

It's so much fun.

So you've got this AI expertise, and Alpha Folds spins off this biological application of it.

First, tell me the word isomorphic.

What does that mean in biology?

Isomorphic is this technical term which is

a one-to-one mapping of space, right?

And the reason we're called isomorphic labs is really that we believe that biology is really, really complicated.

In the world of physics, we can write down equations for physics with maths.

And maths is that perfect description language for physics, but you can't really just write down equations in maths for biology, for the cell.

It's just too complicated.

Biology is the most complex expression of chemistry.

Yes.

that we know.

There's just so many moving parts.

Did you just make that up?

Yeah.

So we're looking for a Rosetta Stone here for the language of biology.

Exactly.

And so what could be that perfect description language for biology?

We believe AI and machine learning is that.

So that there could exist an isomorphism, a mapping between the biological world and the world of AI machine learning.

Hence the name.

Hence the name.

Gotcha, gotcha, gotcha.

All right.

So tell us about...

protein folding.

Because, you know, when we learn about chemistry, we learn about chemical reactions.

And we're not really taught that the shape of the molecule should have anything to do with anything.

It's just what is the chemical symbol.

And when you write down the chemical equations,

there's no shape in there.

There's just what elements and molecules comprise it.

And those equations don't really ever represent the three-dimensional nature.

Exactly.

You don't even know which, if it's, it has handedness.

Yes.

Yes.

Right.

So take us from there.

We think about proteins.

Proteins are these fundamental building blocks of life.

They're inside of everyone.

They make up everything we have, basically.

And they're made up of what's called a sequence of amino acids.

Each amino acid is a molecule.

There's about 20 different amino acids.

And you put them together in a long sequence.

Ever or just in life?

In life, you can have non-natural amino acids as well that you can make as well.

You can make them.

You can make them and actually use those for drugs sometimes.

You string these amino acids together, and that becomes a protein, but they don't exist as these strings.

They fold up spontaneously in the cell to create these 3D shapes.

And why that's important is that these proteins, they're basically molecular machines.

They don't just exist by themselves.

They actually create these little pieces of machinery.

They interact with other proteins.

They interact with other biomolecules like DNA and RNA.

And that interaction is a shape-fitting.

Exactly.

Exactly.

So these proteins are...

It's a puzzle.

It's a 3D puzzle.

It's a 3D puzzle.

Exactly.

A 3D jigsaw puzzle.

And it's not static.

Which is way harder than a 2D jigsaw puzzle.

And these are not static things.

It's not just static puzzle pieces coming together.

They change shapes.

So something comes in contact and that opens up something else on the other side of the protein, which changes the machine and on and on it goes.

And that's what I was going to ask you.

What speeds is this folding taking place?

Is it continuous?

Once it folds, that's it.

But you've just told me, no, it just keeps moving through the whole thing.

Yeah,

these are really, really complex dynamical systems.

Yeah.

You know, composed of thousands, millions, trillions of atoms within our cells, unfolding over the course of microseconds and beyond.

And this dynasism that you're talking about.

What word is that?

Dynozism?

That's a word?

Isn't that a word?

Dynoism?

I think I just made it up.

Mism.

Oh, dynamism.

No, dynamism.

Dynamism.

Thank you.

Thank you.

I'm correcting Grammar.

This is a first.

This is.

Dynamism.

Dynamism.

So you're kind of dinosaur.

Dynamism.

Not dynamos.

Dinosaur.

Dynasism.

Dinosaur.

Yeah, dinosaur.

But the dynamism that you're talking about within the cell.

when you look at each one of us, since each one of us is so different, even though there's a general

execution and blueprint, we all come out so different.

Is that part of the process that you are looking at and mapping?

I would say to a jellyfish, we all look identical.

True.

Okay.

They're not saying, oh, is your skin color slightly different?

Are you slightly taller?

Yes.

You're describing functions at a cellular level.

Is your job to understand that, or is your job to figure out extra ways to fold proteins that maybe biology has yet to even figure out?

That can then solve problems that we encounter that the natural universe has not.

So, oh, that was a good question.

You're happy with yourself.

Happy with yourself today.

Happy with yourself.

Happy with myself today.

Yeah.

This is really interesting.

We have these little molecular machines, these proteins, and we care about that 3D structure and how they work for two reasons.

One, we want to understand how our cells work, because if something goes wrong with that, which is the case for disease, then we want to understand, okay,

where do we actually need to go in and start fixing that?

Or how we can stop it from actually going wrong in the first place.

Exactly, exactly.

So that's one thing.

And then when we think about, okay, how can we go and fix that?

What we're actually saying when we're doing drug design, we're saying, can we create another molecule that will come into the cell and actually start modulating these molecular machines?

This drug molecule is going to actually attach to this protein over here and that's going to cause this protein to change shape, for example, and so it won't operate how it normally does.

Or and so we stop that protein working or we make it work better.

These are the sort of things we do in drug design.

It reminds me of messenger RNA vaccines that we developed for COVID.

Yeah, you know, there's so many different types of molecular mechanisms that we take advantage for for drug design.

A massage chair might seem a bit extravagant, especially these days.

Eight different settings, adjustable intensity, plus it's heated, and it just feels so good.

Yes, a massage chair might seem a bit extravagant, but when it can come with a car,

suddenly it seems quite practical.

The all-new 2025 Volkswagen Tiguan, packed with premium features like available massaging front seats, that only feels extravagant.

The best business-to-business marketing gets wasted on the wrong people.

Think of the guy on the third floor of a 10-story apartment block who's getting bombarded with ads for solar panels.

What a waste.

So when you want to reach the right professionals, use LinkedIn ads.

LinkedIn has grown to a network of over 1 billion professionals and 130 million decision makers and that's where it stands apart from other ad buyers.

You can target your buyers by job title, industry, company, role, seniority, skills, company revenue.

So you can stop wasting budget on the wrong audience.

It's why LinkedIn Ads generates the highest business-to-business revenue on ad spend of all online ad networks.

Seriously.

all of them.

Spend $250 on your first campaign on LinkedIn Ads and get a free $250 credit for the next one.

No strings attached.

Just go to linkedin.com/slash star talk.

That's linkedin.com/slash star talk.

Terms and conditions apply.

I'm gonna put you on, nephew.

All right, um, welcome to McDonald's.

Can I take your order?

Miss, I've been hitting up McDonald's for years.

Now it's back.

We need snack wraps.

What's a snack rap?

It's the return of something great.

Snack wrap is back.

I'm Olicon Hemraj, and I support Star Talk on Patreon.

This is Star Talk with Neil deGrasse Tyson.

Are the folding proteins generally following a set pattern?

in the way that they do fold and you're able to map them and when they misfold

that's when you're able to flag that up or have i just reinvented something or talked rubbish.

There we go.

No, no,

you're onto something.

So

great.

So, you know,

like the, I mean, the amazing thing is that we can actually, turns out, predict how these proteins fold.

So they are.

So modeling that.

Yeah, we're modeling that with deep learning, with neural networks.

That's what AlphaFold, you know, and all its generations are all about.

And that means that we can actually just take in a sequence of amino acids knowing nothing about this protein before, and then get out the 3D structure.

And normally this would take people months, if not years, to work out this 3D structure.

So how is it that alpha fold knows how

a large molecule wants to fold?

Again, it's got to know that in some way.

It's learnt this from a few hundred thousand examples.

So chemists, biochemists over the last 50 years, they've been working out these protein structures by hand.

They've been literally synthesizing protein, crystallizing it, then shooting x-rays at this to like look at the electron scattering, and from that you can resolve the protein structure.

It's a pretty hard process, but people have been doing that.

That's your way to photograph what the shape of the molecule is.

That's your way to photograph in reality.

With it, with that kind of it's basically an electron microscope at that level.

Yeah, similar.

It's like electron scattering, yeah, yeah, exactly.

And so people have been doing that for the last 50 years and depositing these structures.

And now we've taken all of that data and trained a neural network to go just from the input of what is this molecule description to try and predict all of that data.

Right.

And the amazing thing is, and this is really remarkable, is that you can then train this on the last 50 years of data.

That's a couple of hundred thousand protein and biomolecular systems.

But you can apply it seemingly to everything we know about in the protein universe, in the proteome.

Well, it has

the proteome.

Oh, we like that.

Oh, proteome.

Proteome.

Proteome.

Proteome.

So how accurate is AlphaFold and AlphaFold?

We're on the third iteration with its predictions because AI has been around a little while, as you've already said, and you're not the only AI tool that's out there, but how accurate is this particular tool?

Yes.

So AlphaFold 2

was that big jump.

where we started to get experimental level accuracy for just proteins.

And that's what won the Nobel Prize.

So you balanced it off against empirical experimental level.

Yeah, the benchmark is doing the real lab work itself.

So AlphaFold II reached that level.

Now AlphaFold 3 expands from just proteins to incorporate other biomolecular types.

So proteins with other proteins, proteins with DNA, with RNA, with what's called small molecules, which are the same.

They go to the neighborhood.

They start mixing all that up.

Or maybe not the neighborhood.

Maybe the neighborhood gets a little bit of an upgrade.

No, that's when you make the superhuman.

So

I'm trying to be the terminator.

It's not going to be the terminator.

It's going to be the superhuman.

And then they're going to be like us, and they're going to look down on us and go, you know, why do we need you guys?

And that's it.

So, anyway, you're able to predict these.

And

have you actually taken any of the modeled predictions and made the proteins?

Yeah.

Oh, yeah.

Or tell us where you expect these to lead to new and innovative drugs.

Because otherwise, it's just

a puzzle exercise.

Yeah.

It's a great Lego set.

Lego set.

We want the guests to enjoy this.

Yeah.

It's like, oh my God, how much that Lego set costs?

Only $10 billion.

Sorry, go ahead.

Yeah, yeah.

So if you take a particular disease and

we think that we can actually, you know, solve this disease by modulating a particular protein.

The question is how we do that.

So we design a drug molecule and we want it to fit to this protein in a certain way.

And so this is where traditionally you would have to actually either just guess or go into

and

crystallize each one of those combinations and then photograph them and see if it worked.

But now you can model it and the AI can do a thousand of those in like a minute.

Isn't this what you call the

target proteins?

Yes.

Wow.

So if you know you've got a certain target protein, do you not then run that against a list of drugs and think, yeah, this one, drug A works better with this, or maybe it's drug D or whichever letter of the alphabet you're on.

And now we become the sort of detective.

And has this AlphaFold 3 produced how many clues and how many answers?

Or are we still grappling with that?

That's interesting.

Yeah.

Instead of trying to figure out the drug, the AI actually figures out the drug for you.

Discovery.

Well, exactly.

If you let the guy speak, we might be able to,

the two of you.

Okay.

We're figuring out this whole industry ourselves, Gary.

Yes.

Yeah,

this is exactly where it's going.

So

we can start actually rationally designing these drugs.

Traditionally, you would take

a million random molecules and you would just throw them at these proteins and see what sticks.

And that's how so many drugs have been created historically.

You go back further and you're sifting through mud to find these sort of molecules.

This is why there's been such a low percentage of success rates with the sort of drugs for whatever the problem is.

That's part of it because

we don't necessarily understand how these molecules are working.

But with something like AlphaFold 3, you can put the molecule, put the target protein into the system, into the neural network, and you get out the 3D structure.

And as a chemist, you can start to understand, okay, how is this small molecule drug modulating this protein?

Now, still the problem is, well, how do you find that small molecule in the first place that's going to be good for this protein?

You know, it's estimated there's like 10 to the power of 60 possible drug-like molecules out there.

That's 10 with 60 zeros.

How many people know what 10 to the 60s are?

Yeah, yeah, yeah.

Don't be rude.

Even if you had the perfect alpha fold, you'd have to run that across 10 to the power of 60 molecules, which is just computationally impossible.

It's unfeasible.

Until quantum computing.

And so then what we need is something that we call a generative model or an agent which is able to actually search through that space, understand that entire molecular space and come up with molecule designs for you.

Oh, because the 10 of the 60 is if you just did it randomly.

Exactly.

Right, but you're throwing it out, anyway.

If you don't do it randomly, then you can.

Yeah, but randomly is the state-of-the-art method.

That's how people do it.

It's how people currently do it.

It's how people currently do it.

Old school.

Well, he calls it state of the art.

You're the state of the art.

Thank you.

Let's use the word properly here.

So what if the protein turns left

when you've mapped it to turn right?

Is that when we have issues that even AlphaFold 3 has a problem with?

Exactly.

These are not perfect models at the end of the day.

They're very, very accurate, but they will make some mistakes.

So you still do...

currently need to go into the lab occasionally but the amount of lab work you have to do is so much less right than

And often you can find the area of molecular space where these models work really, really well.

And we then go out to the lab later down the line, we crystallize these things and we see, yeah, like this is a perfect mapping of what the model predicted.

So back to an earlier point, in the old days, like last month,

the pharmaceutical companies, big pharma, would

spend millions, maybe not quite a billion, hundreds of millions of dollars developing a drug.

We think that holding aside what might be abuses of pricing, the fact that

there's some truth to

this first pill

cost $50 million.

The second pill cost 10 cents because they had to research to get the formula for that first pill.

If you have narrowed the search space,

then the cost of developing that first pill can be manifold smaller.

It costs on average $3 billion to create a new drug.

Wow.

That's on average.

Yeah.

And so yeah, there's a reason.

I was low when I said $100 billion.

You were low-balling.

I was low-balling it.

So this is a massive opportunity to completely change just like the cost, the speeds.

The business model.

And the business model as we do that.

Is it proprietary?

So here's my real, because here's where you would revolutionize.

So if I come up with it and I'm company A, right?

It's mine and I get to determine everything.

If you're an AI company company and you're just doing this, okay, so that you can sell it, then it's yours.

Which one will be, will make prices lower for the consumer?

Our goal is to is to like really redefine this way you do drug design.

So it becomes so much cheaper.

We have so much more abundance of potential drugs and chemical matter that it really does change.

the business model and it changes the economics of the space.

So you can actually revolutionize the cost of making drugs.

Yeah,

that's where we're going.

That's where we're going.

All right.

Is one of the next steps with AlphaFold, whichever it's three or maybe the next iteration or so, going to investigate why and what drives the misfolding of a protein so as you can kind of get ahead of even the story of that happening.

Wow.

So actually the misfolding of a protein is another thing.

That's what causes some types of disease, where you'll have a genetic mutation, a mutation in your DNA, which will change a particular amino acid in that protein.

And so it doesn't fold the normal way it should fold.

And so it doesn't function as it normally should as a molecular machine.

And so things like AlphaFold can help us understand what are those mutations that cause misfolding.

They're called missense mutations.

And then, you know, these could be potential drug targets.

So we could think about molecules that could mitigate against that.

If I understand correctly, if you look at the PDR, it is this thick, and

the physician's desk reference.

Thank you.

It's this thick.

So it's the size of an old-style Manhattan phone book.

Okay, it's very thick, multiple inches across.

And it's chock full of existing medicines available to the doctor to prescribe.

Is it true that 100% of those medicines are interacting with the patient chemically

rather than through protein folding?

So that if that's the case, does that mean that where proteins misfold, we can't combat it with any kind of folding algorithm.

We just prescribe chemistry for your body to handle the impact of that.

Is that, did I say any of that right?

I think that that is the majority of drugs.

They are chemicals that we take.

The chemical is not going to fix the fold.

It's going to treat the symptoms of the misfolding that happens.

We're not changing the mutations mutations of the proteins.

That could be something like gene therapy.

But these are chemicals that come in and will attach themselves to these proteins and somehow mitigate

something like a misfold or it'll change an interface, change how these molecular machines work.

That's weird.

Is there a particular disease isomorphic labs are focusing upon right now or is this a more of a broad spectrum?

Let's go for proteins and cherry-pick out certain things or are we really looking at one particular the technology we're creating it's really really general we want to be able to apply this drug design engine on any protein any target any disease area that comes our way now saying that you know practically as a company and and that you want to focus on a particular area we're focusing at the moment on you know a lot on cancer and a lot on immunology of course

two biggies

and the two that probably lend themselves best to what you're trying to accomplish actually okay the question everyone's gonna want me to ask right now that's watching this and listening to this is, how are you getting on?

I mean, it's going really well, to be honest.

We're seeing these algorithms actually

change the way that we're able to do drug design.

We're able to discover completely novel chemical matter against some of these targets that people have been working on for even over a decade.

So really, really hard stuff, making amazing progress.

Still really early in the company, but it's super exciting.

And have you sent anything to be photographed yet?

We send some things for molecular photographs, yeah.

Okay.

Yeah, yeah.

I know you're not allowed to talk about it.

I know.

I get it.

But we're all very happy.

Okay, there you go.

Listen, I'm with you.

I'm picking up what you're putting down.

That's cool.

Yeah, but so if you if you I see this work as fundamental research so that you publish a result you publish the image as they published the

image of the DNA molecule to know that it was a double helix.

Exactly.

That becomes public knowledge at that point.

So someone with tools access to AlphaFold 3, would any company have access to this once you have published the blueprint for it?

In drug design, often these blueprints come out in the patents.

So, when you're going to go into clinical trial, you need to patent these molecules.

And in those patents, you'll have a lot of data around the molecules, the formula.

I was talking about earlier.

Yeah, yeah, yeah.

Okay.

Yeah.

All right.

So,

the

immune system, the cancer, these are leading leading causes of maladies in this world.

What of the

genetic disorders that affect one in 100,000 people?

Wow.

You bring them together, there's enough of them, you know, they'll fill a stadium in the world, but that's so uncommon as to not really trigger anybody's interest.

Yeah, it's also not profitable.

Well, yeah, because you don't have enough of a market there to sell the drug.

Right.

I mean, you know, it's exactly that point, Chuck.

Traditionally, it might not be that attractive commercially to go after

very small patient populations.

But in a world where it's so much cheaper, so much easier to get to these drug molecules, then that opens up all of this space.

The cheaper it is,

the easier you can justify going down that risk list.

And this is a big guiding star for us.

This is why we're doing this.

That's beautiful.

I see what you did there.

Guiding star.

You know, great though.

He's picked up in the environment.

He's not sitting.

Why is chief AI officer?

I'm interested how active the company is in shaping policy around what you're doing because there's going to be a great deal of legislative policy that is going to be tied to what you're doing.

All of the patent implications, there's going to be

research implications.

There's going to be a lot of things tied to this.

Yeah, yeah.

I mean, we've been talking in this conversation about drug design, but then once you've designed the drug, you've got to go into patients in clinical trials.

And that's a really long process.

Yeah.

And the mice.

But even these mice models, they're not actually very predictive.

Like you do all these studies in mice, and then they don't, you know, it doesn't translate into success in people.

Right.

You've got to go up the evolutionary scale and then get to the human bit.

Yeah, exactly.

And so there's, you know, you can imagine a world where we can design loads of new drugs.

We've got to be changing the way that we're doing clinical trials, you know, how we can actually get these drugs to patients who really, really need them in a timely manner.

So I think there's a lot to be done and like rethought there.

Is the ultimate goal for AlphaFold and I think medical science as a whole to be able to bespoke medication for you as the individual rather than the broader spectrum medication that we find ourselves with all the side effects?

So are you able to then design a drug?

or a medication that has zero side effects and works exactly for me?

This is the goal, right?

This is what we're shooting for.

You know, imagine a world where we can sequence your particular cancer mutations.

Right.

And then based on those, your individual mutations be generating specific drugs for you.

Right.

That even these are like, you know, 3D printed or something around the corner.

Okay.

Yeah, this is.

We're not there yet, right?

We're in the very nascent stages of that right now with immunotherapy for cancer treatments.

But how many of these yet-to-be-cured diseases lend themselves to solutions that involve protein folding?

And how many are just plain old, old-fashioned chemistry?

Proteins make up like pretty much all of our molecular machinery.

So

there's a class of disease which is due to misfolding, but then there's many, many other diseases which are due to, for example, a protein not being expressed properly.

Right.

Or, you know, a cell going wrong in a certain tissue.

If I have a bacterial infection, I give myself antibacterial chemicals and then I'm done.

Do I need you for that?

But those chemicals are interacting with the proteins in the bacteria.

Okay.

So proteins are the fundamental machinery and the chemicals which are drugs are modulating those proteins, whether it's like in our cells, in bacteria.

All right.

So basically everything you're talking about is all happening on the cellular level.

If what you're describing is happening inside of cells, proteins doing their thing, their 3D jigsaw puzzle, and you have a solution for that, a remedy, you have to get your remedy inside the cell to interact with that folding.

The delivery soldier.

And how do you do that other than through like a Trojan horse virus or something?

Because viruses get in there pretty

on command.

Yeah, well, if you think about the drugs that you take as pills,

drug design is really hard because it's not just about targeting these proteins.

We've got to get them to the right place.

We might want a pill that you take.

So you take this pill.

It's got to be absorbed by the body.

So it's got to be soluble.

It's got to go through the gut wall.

It's got to go through the bloodstream to the particular tissue type, the cell type you care about.

Then it's got to go through the cell membrane to be able to actually target maybe a target which is within the cell.

So you need all of these properties in a single molecule.

So we're actually designing these molecules not just to hit the protein, but also to be soluble, to be cell permeable.

There's so many different factors.

And then you don't want this molecule to be toxic.

So you want it to hit the target of interest,

but not...

hit anything else.

Right.

So I can see how a molecule gets through the cell wall.

Yeah.

A simple molecule, but a full-up protein, red-blooded protein, how's that getting through the cell wall?

Exactly.

There's different types of drugs.

Some are what we call small molecules, things you could take as pills.

Others are made from proteins.

There, they're often things that you would inject directly inside.

And some of those might be cell permeable.

There would be things like peptides.

But often these protein-based drugs, things like antibodies, they're injected, but they don't go in the cell.

They're just interacting with proteins on the the surface of cells.

So you don't need that permeability.

So it really depends on what your target is and how do you want people to be able to take that drug.

Sometimes a pill is the best thing, but actually sometimes injecting is the best thing.

Yes.

And remind me what a peptide is.

A peptide is a really small protein.

So you've got full-blown proteins, which are these big molecular machines.

And then you've got small proteins made up of

five to fifty amino acids.

Those are peptides.

Those are peptides.

They're smaller, so sometimes in some configurations they can get through the cell wall.

That's the sound of the fully electric Audi Q6 e-tron and the quiet confidence of ultra-smooth handling.

The elevated interior reminds you this is more than an EV.

This is electric performance redefined.

Honey, do not make plans Saturday, September 13th, okay?

Why, what's happening?

The Walmart Wellness Event.

Flu shots, health screenings, free samples from those brands you like.

All that at Walmart.

We can just walk right in.

No appointment needed.

Who knew we could cover our health and wellness needs at Walmart?

Check the calendar Saturday, September 13th.

Walmart Wellness Event.

You knew.

I knew.

Check in on your health at the same place you already shopped.

Visit Walmart Saturday, September 13th for our semi-annual wellness event.

Flu shots subject to availability and applicable state law.

Age restrictions apply.

Free samples while supplies last.

At Capella University, learning online doesn't mean learning alone.

You'll get support from people who care about your success, like your enrollment specialist who gets to know you and the goals you'd like to achieve.

You'll also get a designated academic coach who's with you throughout your entire program.

Plus, career coaches are available to help you navigate your professional goals.

A different future is closer than you think with Capella University.

Learn more at capella.edu.

So with this computer science, are we upending chemistry as we've known it?

And are we going to find it kind of moving off?

Even with AI?

Yeah, we go from there.

With the AI, is it then sort of moving into the other sciences?

Are we going to just see it stick in one particular area?

Chemistry is always going to exist.

It's like, to me, it's like any field of science at the moment.

Doing science like chemistry without maths, you wouldn't think about that now.

And it's going to be the same thing with AI.

It already is in my mind.

You just wouldn't do chemistry without AI.

You wouldn't do biology without AI.

It's just that fundamental tool that allows us to understand the world better.

So chemists will not one day be like coal miners.

Just like I remember, my grandpa used to go into the lab.

So going back to an issue challenge.

He comes home smelling like chemicals.

Who's actually going to be able to access Alpha Fold 3 or AIs of this iteration?

Is it exclusively isomorphic labs or this comes out in license?

I want a home kit.

That's it, exactly.

Right, right, right.

My DNA goes in.

There's an alpha fold going on, out comes a pill, and I take it, and I don't even need you at that point.

That would be so cool.

You just finger prick or at least put your finger on a sensor or something, and it figures it out.

You make your own pill at home.

Yeah, yeah.

Look up the Theranos story for that one.

Wow, that's the woman that went to prison.

But no, you can access AlphaFold.

So if you search for AlphaFold server,

there's a whole web-based system where you can fold proteins there for academic use.

It's really cool.

So you can just put your system in, get the 3D fold out from AlphaFold 3, download it.

Yeah.

Oh, cool, man.

I mean, how far away are we from modeling an entire human being, which I suppose touches onto your fears?

Modeling or creating?

Either or.

Once you model,

the next step is creating.

Yeah, that's true.

That's all there is to it.

That's true.

That's the dream.

We kind of need to work up the scales here.

So

we can model how two atoms interact.

We can write down those equations.

We can simulate small atomic systems.

With things like AlphaFold, we get into bigger atomic systems, things on the scale of like multiple proteins.

Now we've got very...

very accurate AlphaFold 3.

Maybe we can actually bootstrap off that to get to more sort of even bigger systems,

what we call pathways, how all of these things interact.

It's only just now clicking within me because you can look up in a book the tables of

action potentials for the interactions of atoms and molecules.

And so you'd have a very good sense of which molecules will combine.

Is it exothermic?

Is it endothermic?

But these are atoms and molecules.

And as powerful and as convenient as that is, that's just the first rung in in this ascending ladder of complexity that you are gaining control over.

Yeah, and there's trillions of atoms within a single cell, let alone the whole human body.

It's just unfeasible to simulate the whole thing.

But what you can do is you can, you know, we do have good measurement techniques.

at different levels of scale.

So we can measure things like protein folding.

We can measure the amount of protein within a cell.

We can measure the number of cells of a certain type within a tissue.

And so we have these.

The bigger it is, the easier it would be to measure.

So we have these little windows into this sort of microscopic world.

And then we can use AI to sort of fill in the gaps and bootstrap off the stuff we can do well, the atomic level, and start building up that scale of modeling, if that makes sense.

We can rebuild it.

I read this article like...

God knows how long ago.

God, how long did Jack read the article?

Because you know, God.

Anyway, it was talking about when a fertilized cell starts to proliferate and become a person.

And basically what it determined, what the scientists determined at that point, and this is many years ago, is that the only way they could describe it is there's a bunch of noise.

Like there's just a bunch of noise.

We can't really see anything.

We can't make sense of any of it because it's just basically, if we were to look at it as data, it would just be noise.

Are you able to pierce that veil and see into that?

I mean, we haven't been looking at that specific thing.

Okay.

But this is where you start to understand more about, you know, a really granular scale.

And then you can integrate that and create these sort of, I don't know, coarser measurements and coarser predictions.

This is what we do in lots of areas of science, right?

We don't simulate the whole universe at the atomic scale, but we find these rules of thumb or ways to describe sort of broader collections of molecules.

And that's what we can start to build up and actually learn with these neural networks.

Cool.

So question on behalf of Chuck.

Could AlphaFole discover a hallucinogenic that could make him see God

or any other deity or being?

Thank you for asking.

Welcome.

I would like a very real answer, please.

You can go on the AlphaFole server and try that out.

Okay.

All right.

Invitation.

Hey, listen, I'm all about it.

So we've looked

in previous shows talking to biomedical engineering.

And if we are to travel off-world and deep space we are probably going to need different upgrades for us to be able to do that

are we going to be able to with alpha fold or AI like this be able to upgrade ourselves to make this sort of deep space travel

or or upgrade ourselves for anything anybody

fire you know the yeah now now we're sort of beyond you know solving disease into like actually can we enhance ourselves right yeah I don't know I think like there's there's probably potential right to to think about creating chemical matter that we can take or ingest that that i mean aging alone is would be aging huge

application for this i mean there's crazy research on aging aging is basically cellular degeneration yeah and if you're able to on a molecular level kind of restart that process or jumpstart it or you know boost it aging is an interesting one you know this is a really nascent area of research where people are just starting to work out what are some of the um factors that that reverse the age of cells.

There are these things called Yamanaka factors.

Yes.

And there's even potential that people are finding of creating molecules that stabilize particular proteins.

Yamanaka factors are proteins.

They're transcription factors that read DNA.

Can stabilize these things.

Maybe that is what reverses some of the age of cells.

This is super nascent.

So what is the connection between wanting to modify a genome and your ability to fold proteins to interact with our physiology?

I ask that because I'm reminded there was a scene in the film Gattaca

where they didn't manipulate your genome, but they selected your pre-existing genome for certain properties.

And there's a person giving a piano recital, and it was a very rich sound.

I mean, it was beautiful.

And then the camera came around to the front, and the person had 12 fingers.

That's right.

And

bread for that.

Yes.

Right?

You get two extra notes for every

everything going on.

She could only play the stuff.

Nobody could play with you.

Nobody could play with you.

Nobody could play with you.

So this would be modifying not to go into space, necessarily, but just to sort of enrich the diversity of the human species.

We're not doing genetic modification.

So he says.

He's English.

You must trust him.

To be honest.

Thank you.

Thank you.

Well done.

Well played.

Can you?

Is it the same thing?

Some particular types of drugs are

things that would manipulate your genome.

That's how people

start to target some diseases.

This is not the class of drugs we're working on.

When we think about the big ambition of solving your disease, maybe this is something that we need to be doing over time as we want to really crack the whole spectrum of disease.

Is it even possible to consider that without considering the whole area itself as it all bleeds in together at some point?

Yeah, I mean, we need to understand the genome and all the effect,

how changing a particular particular base pair on your DNA is going to change

what proteins are expressed or in what abundance and how that, all the knock-on effect on the pathways.

You really want this like basically this virtual cell to be able to manipulate this cell on a computer to do experiments there.

Here's what I want you to do.

Talking about prioritizing?

And I'm not asking much because it already happens in the animal kingdom.

You know, for so long, decades, even centuries, we imagined ourselves at the top of some evolutionary

triangle structure with the gay pets.

Without any arrogance whatsoever.

And, all right.

Yet, a newt can regenerate a limb, and we can't.

Yeah.

And so it seems to, and they're vertebrates.

So it seems to me there ought to be some

way

to extract from animals that do things that we could benefit from and then make that a priority.

So people, but especially veterans who've lost limbs

in conflict.

Or even geckos with their sticky hands, like maybe I could be Spider-Man one day.

So organize.

Let's prioritize that, Chuck.

We'll become superheroes.

So the regeneration of limbs,

that's got to be a protein thing going on in there, isn't it?

Yeah, I mean, you know, all of our mechanisms are proteins.

Same for nukes as well.

So there is some mechanism there.

I don't know what it is.

Okay.

Gotcha.

But that would be a mechanism to emulate if you could.

If you could, yeah.

And then install it into our own physiology.

It's a big if.

Yeah, that's a lot.

Yeah.

How about this?

Would you be able to look at drugs that are already here?

And there are some drugs that are just not well tolerated.

And you'd be able to reconfigure them in such a way that you get the benefit of the drug without the side effects.

Yeah, exactly.

So you often have these first generation of drugs that do something, but they have these side effects.

Then there's a big opportunity to understand better how these drugs work.

Things like App Fog 3s, things like our models that understand the toxicity of drugs, can then allow us to potentially modify these to become better drugs and have less side effects, less toxic effects.

Cool.

Wow, that's going back through the medical catalogue.

Yeah, man.

Reanalyzing, which is exactly what an AI would be perfect for.

Right, exactly.

Yeah.

Yeah, you guys are going to make a lot of money, man.

I don't know how I get

dollar signs.

I see some

benefits of this company.

I'm going to invest in a company.

I need to get a piece of this company.

You guys are going to make, I mean, I can't even imagine the amount.

The gobs and gobs of money.

More money than cells in my body.

This is amazing.

We're on the doorstep of quantum computing, and I know what impact that would have in my field.

In your field, would it make your entire life's work look like it was done on an abacus?

Yeah, I mean, this is going to change things.

I think open question how this changes machine learning.

Like, what can AI do with quantum computing?

But for chemistry, even near term, there are some real applications of quantum computers for understanding the properties of small molecule drugs.

Because actually, some of the things that people do do today with quantum computers is simulate these small chemical systems.

We actually, even in the company, we have a quantum simulation team

that are not using quantum computers, but simulating...

you know, the quantum effects of molecules.

Now, if you had a quantum computer that could work on that scale, you could use that instead.

Wow.

I think of so many needs on the frontier of chemistry in modern society.

One of them is: you know, what do we do with all the plastic that's in our environment that's still there in the ocean?

Is there some life form you can create that'll digest the plastic and turn it back into its original molecules?

Are proteins something that could be applied there?

If not, in your world, then you're describing an ability more than you're describing a specific solution to a problem.

You're empowering the chemist in ways never previously imagined.

Yes.

So you can use the capability of AlphaFold to

understand structure of proteins.

People are using this outside of drug design.

Exactly.

Exactly.

People are using this, for example, to create bacteria that have enzymes that could potentially digest plastics, like you're talking about.

You could think about this for engineering

more resilient types of crops, these sort of things.

So just like AI, this is a platform upon which you can rest the the technology of any field.

Yeah, I mean, that's the amazing thing about the protein folding problem.

Once you start to solve that, you unlock so many new things for a whole broad spectrum of science.

There's a lot of downstream benefits.

Okay, last thing.

Here's the last thing.

How do I get a piece of this coming?

How do I chuck his good dollar time?

How do I get a piece of this?

Last thing.

Last thing.

What is the worst possible outcome of your work?

Ooh, what a question.

What guardrails are necessary as we go forward?

Because any new technology with awesome power comes awesome responsibility.

Yeah, I mean, I think, you know, you have this with AI.

You have this, you know, creating new biology

or chemicals.

You know, you just need to think about how to use this responsibly, like what you're putting out into the world openly versus what you close off for, you know, many safety reasons.

So I think there's a lot of things to consider there.

Because famously in one of the Jurassic Park films, they withheld lysine,

amino acid, from one of the dinosaurs in case it escaped.

It would die because it would need the lysine for its survival.

And that was a kind of an insurance plan that put in.

But life always finds a way.

There you go.

Anyway, Maxwell Yoderberg.

Yeah.

Thank you for joining us on Star Talk.

We're going to be watching your company, and Chuck wants a piece of it.

Yes.

I don't know what that means.

But anyhow,

I'm delighted to just be able to look through your lens at the birth of an entire frontier in human physiology.

I mean, what a time this is.

No, thank you so much.

It's been super fun to talk.

Thank you.

Excellent.

Excellent.

All right.

I think we're done here.

This has been another installment of Star Talk Special Edition, talking about AI, human physiology, and the future of drugs.

Oh, yeah.

Gary, good to have you always.

Pleasure, Neil.

All right, Chuck.

Always a pleasure.

As always, this has been Star Talk, Neil DeGrasse Tyson, your personal astrophysicist.

Keep looking up.

So, um, I was just parking my car and then I saw you, the Gecko, huge fan.

I'm always honored to meet fans out in the wild.

The honor's mine.

I just love being able to file a claim in under two minutes with the Geico app.

Well, the Geico app is top-notch.

I know you get asked this all the time, but could you sign it?

Sign what?

The app?

Yeah, sure.

Oh, that means so much.

Oh, it rubbed off the screen when I touched it.

Could you sign it again?

Anything to help, I suppose.

You're the best.

Get more than just savings.

Get more with Geico.

Honey, do not make plans Saturday, September 13th, okay?

Why, what's happening?

The Walmart wellness event.

Flu shots, health screenings, free samples from those brands you like.

All that at Walmart.

We can just walk right in.

No appointment needed.

Who knew we could cover our health and wellness needs at Walmart?

Check the calendar Saturday, September 13th.

Walmart Wellness Event.

You knew.

I knew.

Check in on your health at the same place you already shop.

Visit Walmart Saturday, September 13th for our semi-annual wellness event.

Flu shots subject to availability and applicable state law.

Age restrictions apply.

Free samples while supplies last.