The Hardest Problem AI Ever Solved, with Google DeepMind CEO

0:00 Something's obviously not quite right about the definition of intelligence. >> If we play this out, what's the limit here? >> The best use case of AI was to improve human health. It was the moment I've been waiting for that could achieve something no other system could. I want to use AI as a tool to help us understand the nature of reality around it. >> Governments are going to use AI. What would you hope that they use it for? There's two things to worry about.

0:23 One is That's Demis Hassabis, the CEO of Google DeepMind. >> Nobel Prize winner. He is one of the most important people alive on what is quickly becoming the biggest technological leap in our lifetime. Because the biggest way that AI is going to impact our lives isn't something that we can see. It's not a chatbot, it's not an image generator. It's tools that are invisible to us in drug design and natural disaster detection and nuclear fusion and quantum computing.

0:54 Tools that he and his team are building. Here he is winning the Nobel Prize for just one of those tools. So, who he is and what he chooses to build matters a lot for you and me. And he's fascinating. He's a childhood chess prodigy who at 17 turned down a reportedly million-dollar job offer from a gaming company to go to college instead and then got a PhD in cognitive neuroscience. He founded his company DeepMind with a mission to solve intelligence starting with beating video games. He then sold that company to Google specifically because they promised to let DeepMind focus on scientific research. But, as this has turned into the most intense technological battle in recent history, Demis is now in charge of much, much more. He's now behind basically everything Google does in AI. He's making decisions that affect your life and millions of other lives every single day. So, what is he planning to do with all of that power? My goal is to show you the future that Demis Hassabis wants to build so that you can decide for yourself what you think of it.

2:07 Welcome to Huge Conversations. >> >> Thanks so much for doing this. It's great. >> appreciate it. You already know that Huge Conversations is a different kind of interview. I'm not going to ask you about financials, I'm not going to ask you about your management style. I All well covered elsewhere. What I'm hoping to do in this conversation is think about it more like an explainer that we're making live together. And I have some props. This was not actually meant to be a Jenga game.

2:39 >> >> each block represents a project or a model and I want to talk about them and how they fit together. And so they were meant to be visual aids, but as we were setting up, we started playing Jenga with them and it turned out to be way more fun than anything I had planned. Also, I know that you like games. >> Yes, I love games. So, this is great. First in an interview anyway. So, yeah.

2:58 Well, so my hope in this conversation is to make this explainer together and to help people see what's happening right now in AI really and what is the future that you see coming? What are you hoping to do in this conversation? Yeah, a lot of the reasons that I got into AI 30-plus years ago now is to advance science and medicine and I've always thought of AI as potentially the ultimate tool to do that. So, I'm hoping we're going to talk about that today and really that's been my passion for what to apply AI to. But of course it can be applied to many things. Oh, this is going to be a lot of fun. Yeah.

3:33 So, in this Jenga game that we have, a lot of these are blocks that people will have heard of, right? These are, you know, this one is Gemini. >> Right. But I would argue that the ways in which AI is meaningfully shaping people's lives most are the things that are invisible to them most of the time. >> >> So, I want to start by talking about the project that you won the Nobel Prize for. Mhm. AlphaFold. Yeah.

3:59 Good Jenga play. I want to tell the story of AlphaFold with all of its drama because some people might not have heard it. But then I want to get really quickly to the cutting edge of this sort of category of science. Why did you decide to tackle this problem out of all of the many? Mhm. Well, I came across it actually as an undergrad in Cambridge. So, I had a lot of biologist friends and one of them specifically was obsessed with what's called the protein folding problem. So, proteins are what everything in your body relies on. They make biology possible and living possible. And what's important about them is their 3D structure. So, in the body they fold up into kind of 3D structures and those structures determine what function they have or partially determine what function they have. And so, the protein folding problem is really about can you predict this 3D structure just from the one-dimensional amino acid sequence? So, that's the kind of 50-year grand challenge of protein folding. So, I love challenges, I love puzzles. So, I couldn't resist it from a scientific point of view as this probably, you know, is described to me as the equivalent of Fermat's Last Theorem but for biology. So, who couldn't be interested in that? But also, when I first heard it, I I thought the kind of problem it was would be suitable for AI one day. Even though we of course this is in the late '90s, we didn't have any kind of AI that would be possible to work on this. But I thought one day that would be possible. And then the final thing was just the impact it would make if you cracked it because it would open up all these downstream possibilities for research and especially in things like drug discovery and understanding disease. So, which I think is, you know, the most important thing to apply AI to is improving human health. And the reason that this would be huge for human health is that up until now, in order to develop new medicines, we'd have to spend hundreds of thousands of dollars and years of human effort to find out the structure of a single protein by shooting X-rays at it. So, we had figured out some protein structures, >> >> but it was slow and expensive.

5:56 So, I'm skipping over an enormous amount of hard work here by you and your team. But I think by the way that I'm asking the questions, >> >> it is very obvious to people that you solved it. Yeah. So, there's this moment where you realize that it is genuinely useful and you have solved what had been called one of the most important unsolved problems in modern medicine. >> Mhm. And it's 2021. >> Mhm. You're in a meeting. Mhm.

6:19 I am so glad that there was a camera in this meeting because it is one of the most incredible moments I have ever seen. Can we use AlphaFold to to solve it? I think you're talking with your team about setting up a system where scientists could send in a request for a specific protein, like a website, and then get the protein folded. Yes. And then someone else has a very different idea. Yes. Can you walk me through what happens in that meeting?

6:43 And then you your reaction is incredible and I really want to know what you were thinking. >> Yeah. Sure. Well, look, we we it's it was funny that the cameras were there happened to be in that in that particular meeting. It's crazy that was that day, you know, they very rarely followed us, but it was fit for that meeting. And normally what happens for these sorts of prediction models is you the traditional thing is you kind of set up a server and then other scientists send you their protein sequences and they say, "Oh, I'm interested in this protein. Can you send me back the the predicted structure?"

7:15 So, you know, and that's how it's been done in the whole field for the last 40-plus years. and the reason is is because most of the prediction algorithms are quite slow. So, maybe it would take a few days and then you'd get back the, you know, you'd email your email back the structure and then you'd ask for, you know, you'd ask for the next one. But once I realized sort of in that meeting actually that how quickly not only how accurately we could fold the proteins but how quickly, you know, in a matter of seconds. And then I was just sort of doing the back-of-the-envelope calculation like how many proteins are there known to science, known in nature?

7:48 200 million. And then how many computers do we have? And how many would we need? And then, you know, put and then if we folded one every 10 seconds, and I sort of realized sort of in the middle of that meeting while I was fiddling on my phone that it would be possible in a year. And so, why go to all the effort of building the servers and the databases and and, you know, the emails client and all of that when we could just actually fold everything ourselves, everything anyone could ever request and ever want, and then put it on a database somewhere for free for all the scientists in the world to use. So, it just suddenly hit me. We should just do that.

8:27 Now you can really Why don't we just do that? Well, so that's one of the options like we Right. you know, there's this We should just run We should That's a That's a great idea. We should just run every protein in existence. And then release that. Suddenly, all these things must have been going on in the back of my mind and I suddenly realized that that would be the obvious thing to do and it would be actually probably less effort than standing up the server. So, actually it would save us time. And you in that meeting, your reaction is something like, "Why don't we just do that? That would be way better. We should clearly do that." And then you do. All of a sudden this crucial process that had been so hard is suddenly fast and easy. And it's being used by scientists all over the world. This huge unsolved problem now solved.

9:13 Is it correct to say that we have now predicted the structure of almost all proteins known to science? >> Yes. And we keep updating it. So, every time, you know, somebody scoops a pail out of the ocean somewhere and then, you know, there's loads of sort of different types of organisms in that bucket of water seawater and then they sequence them all. And so, the sequencing technology has obviously improved many orders of magnitude since the human genome was sequenced. So, now the problem was the structural biology, the finding these 3D structures was taking was far lacking far behind the genetic sequencing. So, now with these computational resources like AlphaFold 2, we can actually keep up with, "Oh, here's a new million genetic sequences from some new strange organisms we found, oh, here are the structures. And so, we have a kind of a small team at the the at the European Bioinformatics Institute that keeps updating every year, you know, all the new sequences that have been found that year. So, we're we're now, you know, always at the cutting edge. Like, we know what all of these different protein structures mostly look like. That's so awesome. It is pretty amazing. It's especially amazing, actually, for the people researchers that work on slightly more obscure, you know, organisms or animals and things like for example, wheat. I found out that a lot of plants have way more genomic data than mammals and humans, which is very strange. They seem to have multiple copies of of of of their genome and things. And it's it's it's it's it's a kind of strange and bizarre world, I think, the plant world. But my plant scientist friends of mine is like, you know, they they they don't have the resources like with, you know, the human genome, there's a lot of work's been done on that. But some of these more obscure organisms that still really important for humanity, you know, like crops and things like that, now we're able to immediately jump to the science around what they want to do with the proteins, maybe, you know, help them be more resilient to to climate change, things like that. And and and they can jump straight to the problem they're actually interested in rather than getting bogged down with trying to crystallize the proteins that they're interested in. Another boon is for researchers who work on neglected diseases that affect primarily the more developing parts of the world, things like malaria or Chagas disease or leishmaniasis. These they affect, you know, hundreds of millions of people around the world. And but there's not a lot of money in that if if big pharma tried to research that and find cures cuz they're in the more poorer parts of the world. So, they tend to be neglected, the research that goes into that. So, there's these amazing non-profit organizations that do the research on that.

11:58 but they don't have a lot of money or resources, so giving them the structures of the proteins that are involved in, say, malaria virus is a huge boon for them, too, cuz they can go straight to the drug discovery phase. That was one of the hardest things to figure out as I was doing this research because there's this moment where scientists all around the world have access to AlphaFold. You can see like the map lights up. You can see that people are using it. But I wasn't easily able to figure out what, and a great example would be to talk to you about of a scientist using AlphaFold and then that speeding up a drug process that results in a drug that I could now take. What is your favorite example of a scientist using AlphaFold for something the audience might understand or have seen. So, over 3 million scientists are now using AlphaFold. We think it's pretty much every biologist in the world at this point.

12:47 and one actually scientist at at a pharma company said to me that you know, almost every drug developed from now on will have probably used AlphaFold in its in its process, which is sort of, you know, mind-blowing, really, and amazing that that's that that that that impact it's having. So, but it still takes time with drug discovery. So, we're we're still mostly in the fundamental biology stage of understanding the disease, what is the protein we're targeting, is that the right biological mechanism. And then as I understand it, some of these drugs are now in kind of the the the the clinical trials phase. And then hopefully we'll see in a few years' time mir you know, a whole dozens of drugs that were partially helped by at least AlphaFold. And in terms of my favorite breakthrough is more on that so far has happened with the help of AlphaFold is there's this protein in called the nuclear pore complex, and it's one of the biggest proteins in the body. It's huge for a protein. And what it does, a very important job, it's basically the gateway that opens up and closes to let nutrients come in and out of the cell nucleus. So, it's it's basically the gate It's like a big donut ring that opens and closes. And but we didn't know until very recently what the structure of this was cuz it's so big and complicated. it's pretty hard to crystallize and actually see. And so, almost I think it was pretty much 6 months or year after we put AlphaFold out, some teams used it along with experimental data to finally work out what this beautiful shape was of this this gateway protein. And that was amazing to me. It's, you know, one of the biggest proteins in the body that's and and AlphaFold was was was very useful in helping determining that structure. And so, perhaps we can design drugs or treatments that better that that use that somehow that better access the Yes, potentially. I think I think that was more for fundamental biology understanding, but but obviously there is I mean, we ourselves, we've spun out a new company, Isomorphic Labs, that actually tries to build on AlphaFold and uses it to indeed in this in this block here, uses it to as as one of the pieces of the puzzle to massively speed up drug discovery. So, on average, you know, it takes like 10 years to to to kind of develop a drug. It's crazy long time, unbelievable amount of hard work, very expensive, and huge failure rates. You know, only about 10% of drugs actually get through all the clinical stages. So, it we need to vastly improve that if we want to improve human health, I think. And I think the way to do that is by using in silico methods, AlphaFold 2 being one of those components. but knowing the structure of a protein is only one small part of the drug discovery process. You need a lot of chemistry, like what compound should you design to bind to it, all of these things. So, we're trying to Isomorphic build these, you can think of them as adjacent systems that work with AlphaFold, more advanced AlphaFold, AlphaFold 3, AlphaFold 4, you could call it. And and then end-to-end basically create these drugs that have, you know, very minimal kind of side effects and and incredibly effective at addressing the type of disease we're we're we're trying to help with.

16:08 And we're working on, you know, I think at this point at like 18, 19 different drug programs across the gamut of things, from cardiovascular heart disease to cancer to immunology. So, I think eventually these these types of technologies should be able to help across almost every therapeutic area. In prep for this, I did a background interview with your fellow Nobel Prize winner, John Jumper. He really stressed that it's one part of a larger problem of drug discovery. And so, that brings us to the cutting edge today. I've taken some of the some of the examples that I want to talk about.

16:43 what is the cutting edge now? Sure. So, we're we're we're building many different components that can kind of go together. So, AlphaFold is one of the linchpins, so that's the structure of the protein. But if you think about it, let's say you understand what the shape of the protein is. Okay? And then now you know which bit of the protein is the important part that does its function. So, now, if you think about drug discovery, so say you want to block the effect of that protein or enhance it in some way, you know you need which part of the protein surface do you have to bind to. So, now you have to discover a chemical compound that will that will attach to the right place on the protein, right? And and you want to know how strong will it attach, and will it And then on top of that, even more important is not just will it attach to the thing you're interested in, make sure it doesn't attach to other things because if it does, that would be toxic, toxicity. So, you don't want it to have these side effects, we call it side effects with drugs. So, you want to minimize those. So, but because now we have all of these amazing algorithmic tools, we can sort of do a virtual screen of like, oh, here's a compound one of our AI systems has designed. It binds This is our prediction of how strong it binds to the protein surface. And then we can check that very quickly, like in a matter of hours, that particular compound, how does it attach to any of the other 20,000 proteins in the human body? And so, we can just do it like that, you know, within a few a few minutes, and then modify keep modifying the compound so that it has less and less side effects ideally none on any of the other proteins, but increasingly strong effect on the one that you want. So, you can see I've just outlined a self-improvement process or self-modification process. And this is extremely fast and efficient if you can do it in silico. And then you know, on on on on on computers, and then at the last stage, only at the final stage do you check it in the wet lab.

18:38 So, you still have to validate it. You can't You make all these predictions, you do all your search in silicon, but then at the final stage, you check your final proposed compounds in the wet lab. and then check it really does what the predictions say. But that, you can imagine, would save you can search thousands of times more compounds, or maybe even millions at some point, more quickly and efficiently that way, and then just at the end check that they're correct. That's so much more efficient than doing the search in in the wet lab, which is what effectively is done today.

19:09 one of my favorites also is AlphaGenome. >> Yes. So, I reached out to yet another Nobel Prize winner, good grief. >> >> Dr. Jennifer Doudna, who I've had on the show. And she sent a question for you. So, I'm going to read this question from >> >> Dr. Doudna. Okay. So, she says, "CRISPR, the gene-editing technology that she pioneered, can now target nearly any DNA sequence. But for most genetic diseases, we still don't fully understand which changes in the DNA are actually driving the problem, especially in the 98% of the genome that doesn't code for proteins. With tools like AlphaGenome starting to decode that 98%, how close do you think we are to the moment where AI can reliably point to the exact genetic change causing a patient's disease so the technologies like CRISPR can fix it.

19:55 >> Yeah, what an awesome question. So, you know, I've discussed this with her actually in the past and and it is really exciting. I think with AlphaGenome which is exactly that kind of technology, it takes the big long genetic sequences and then it tries to predict, you know, if you had made a mutation to this particular single letter single position in the genetic sequence, will that be a harmful, you know, mutation that might cause disease or is it benign and it won't and AlphaGenome which we just released is the best system in the world for predicting that. So, that's exactly what you then want if it got it's still not probably good enough yet, but you can imagine a future version of AlphaGenomes that are accurate enough to sort of really know like oh, that particular mutation in combination with this other one, that's the hard part is what if they're multigenic diseases where there's cascades of mutations cause the problem. Those are even harder to detect, but actually perfect for sort of AI to try and help with. Then, you could go in with something like CRISPR maybe one day and go in and fix that mutation. and then fix the problem. So, that would be so a kind of combination of things like AlphaGenome and CRISPR could be incredibly powerful and hopefully one day will be you know, collaborating with with with the likes of Jennifer on that. Last year you said something to the Guardian that I found really interesting. You said that if I'd had my way, I would have left AI in the lab for longer. And the quote is done more things like AlphaFold, maybe cured cancer or something like that. From the outside, it looks like the story goes you found DeepMind with the mission to solve intelligence and use it to solve everything else.

21:33 And then you sell to Google specifically because they will allow the freedom to explore science in this way. And for a long time that's your exclusive focus. And then ChatGPT comes out, Google goes code red, and you become the head of all Google AI including the consumer products that you weren't spending as much time on before. And it feels to me like watching that from afar, it mirrors somewhat the larger experience of AI which is just this incredible change in the last couple of years.

22:05 What was gained and what was lost in that change? Yeah, I think that's exactly right what you described is is sort of what how it felt from the inside too. And for me as I mentioned earlier was that the AI the the best use case of AI was to improve human health and accelerate scientific discovery. In fact, for me I I I've got I got into AI in the first place because I was interested in all the big questions in the world, the nature of reality, nature of consciousness, these kinds of things and I felt we needed a tool to help us even the best scientist to help us make sense of the amount of data and information out there and find insights in that. And that's happening which is amazing and obviously AlphaFold was the our first and you know, so far best expression of that let's say.

22:55 And I always had that on my mind and many other problems like that. So, it would have been great I think to and and given how important AGI is and how transformative a technology is, maybe the most transformative one in human history, then I thought it would be best to approach these kinds of the the the sort of latter stages of building it which we're in now in we're using the scientific method very carefully, very precisely, very thoughtfully and rigorously with all the best scientists kind of in my ideal world collaborating on in kind of CERN-like way effort on making sure each step we understood each step each as we got to the final goal of of building AGI. seems to me like that would make the most sense with a technology like this.

23:43 And then and of course you don't have to wait so that might take a lot longer, maybe a decade even a two decades longer, but I think that would make sense given the enormity of of what we're dealing with. And and then my other idea was but we don't have to wait till AGI arrives to get start getting the benefits of AI. We could use more specialized systems that maybe make use of the general technologies, the general algorithms we're developing for AGI, but are not in themselves general intelligences.

24:10 They're they're narrow AIs if you if you want to call them like AlphaFold which does a specific purpose and and only that purpose. And we could we could have we could and we still are I'm still doing this create you know, many types of AlphaFolds and isomorphics while we're building AGI in this careful scientific way and then benefit the humanity could benefit from the from the proceeds of that like cures for cancer or maybe new energy sources or new materials. And so, I felt that that would be you know, maybe looking at this from 20-30 years ago when I started out on all of this, that would have been the ideal way for it to play out in my opinion. Now, it didn't happen like that because technology is unpredictable and in fact, it turns out that things like language were a lot easier than we were all expecting even those of us who were obviously optimists about the whole technology and eventually we'll crack language, but it seems funny to think of it now, but language and concepts and abstractions, things that the current models foundation models like Gemini do incredibly well, we thought that maybe there would be one or two or three more breakthroughs needed before we could get there. But it turned out transformers which my Google colleagues invented and some reinforcement learning as well on top was enough to crack things like language. And and we were sort of playing around with that so with the other leading labs, but of course with ChatGPT and fair play to OpenAI, they scaled it and then they put it out there and I think even they say it was a sort of science it was kind of a research experiment. They didn't realize it would go so viral and I think none of us did and we had sort of fairly equivalent systems at the time because I think when you're building that technology, you are so close to it, you you're very aware of the things it can't do, the flaws it has and you don't realize that actually people out there would find use even though it was hallucinating and doing other things that we're obviously all still trying to improve on now, still not completely fixed, but there's still interesting use cases like summarizing things or you know, brainstorming things like that that people use, you know, everyone uses chatbots for today. Now, the downside of it is is that we're in this sort of ferocious commercial pressure race that that that everyone's sort of locked into currently. And then on top of that there's geopolitical issues like the US-China race and so on. So, there's sort of multiple levels of of of of pressure to sort of move fast. So, the benefit of that is of course you get faster progress obviously. So, you know, the progress is just like at lightning speed at these days.

26:47 so that's good for all the good use cases. The second benefit is that everybody all of the viewers out there everyone, you're all getting to use the most cutting-edge AI technology perhaps only three to six months behind what is actually in the labs. So, that's kind of mind-blowing. It's also great because I think it gives everyone a feeling for it's democratizing AI. It's giving everyone a feeling for what it's like to interact with cutting-edge AI and what it can do and what it can't do. And I think that's good for society to start getting normalizing itself to what is going to be an enormous change with this technology coming. So, it's probably better that we get to sample that in incremental steps rather than it's just a shock to the system. Here's a you know, there's no AGI and then here's AGI one day. Probably that that that's not good although I think there could have been many ways it could have rolled out.

27:40 And then the final thing that's actually on the benefit side is that you you can't really fully understand your systems until they're stress tested by millions of people. So, it doesn't matter how good your testing is and you know, your in-house testing obviously millions of smart people trying out things and then you seeing what bubbles to the top or the feedback you get is really important for building more robust systems and better systems.

28:16 So, I think there's positives about and negatives about how the way it's gone. It's not the way I dreamed about years ago where we would be sort of contemplating this philosophically and and and sort of so, we we you know, we have to deal with the world as we find it and make the best of that. And we try to do that by advancing the frontier, but also trying to be as responsible as we can with doing that as we deploy these you know, very powerful technologies like Gemini and AlphaFold.

28:46 There's another story happening at the same time as this and I want to get back to your concerns and how you weight those concerns and the cost. in order to understand that, I think we need to tell a story about AI being very creative, unexpectedly creative. And that story begins let me find my Jenga block. That story begins here. So, let's go back to March 10, 2016.

29:16 There's a very famous Go player that sits down to play against a system that you designed. And at this point computers have beat humans at all kinds of games, but Go is really interesting because there are more potential moves in Go than atoms in the universe. They're they go back and forth, they're playing. And then your system makes a move that is so surprising because it is incredibly unlikely that a human would figure out a move like that, move 37.

29:44 >> Yes. And you see Lisa Doll sitting there, He's just got this shock on his face. He's got his head in his hands like this. And it really was this moment where I think people like yourself saw ahead to the creativity that we would find in AI systems that are very different than the systems that we've talked about so far. So, there's a category where you're giving a huge amount of data and you're asking to make new predictions. And I understand this is much more complicated than this oversimplification. But then there's a category where you're not giving data, you're giving rules.

30:14 >> Mhm. Like with math or physics or games like Go. And it has this incredible opportunity for creativity. >> Yeah. Where were you when that moment happened? >> >> And what future did you see ahead? Yeah, it was an incredible moment that you're describing and it's actually almost exactly 10 years ago now, which is feels like a century ago actually, but I think in many ways it was the dawn of the modern AI era because until that point there were many AI programs that could be world champions at games, things like chess, but they were done with what's called expert systems. So, they were systems where the a team of smart programmers with a team of smart, in that case chess grandmasters, came together, tried to distill the knowledge the chess grandmasters have into a set of rules and system, kind of a brute force system, that would use a lot of compute like on a supercomputer like IBM did with Deep Blue to beat Garry Kasparov, and they would in sort of encapsulate the rules they were given by the chess experts, and then the the system would sort of dumbly execute those those rules and heuristics and do a millions and millions of searching of of moves, and then try and work out against those heuristics which is the best one to do. Now, the thing with that is, for me that was not satisfactory when I saw that in the '90s.

31:33 I was doing my undergrad at the time. I didn't feel like that was proper AI because that system, let's take Deep Blue, okay, it's it's it's world champion level at at chess, but it can't do anything else. Not only can't it do, you know, language and robotics or any of those kind of things, it it can't even play strictly simpler game like tic-tac-toe, right? So, something's obviously not quite right about the definition of intelligence, right? In the sense of like no human, you could imagine a human grandmaster not being able to learn how to play tic-tac-toe. It would make no sense cuz it's strictly simpler. So, so there's something sort of wrong about its generalization capability and the fact that it didn't learn. It was just given the answer, right? So, where if you could ask for something like Deep Blue, where did the intelligence reside of the system? Well, it wasn't in the system, it was in the minds of the chess grandmasters and the and the programmers. They solved the problem of chess and then implement and then implemented the solution. The the program just dumbly executed the solution. Now, Go, as you mentioned, is the sort of final frontier for games.

32:40 It's it's the most complex game humans have ever invented. It's also the oldest game, so it's just amazing in many ways. And it's also very beautiful. So, in Asia where they play in China and Japan, Korea, instead of, you know, it takes all of it they play instead of chess basically, occupies that intellectual echelon. But it's a much more intuitive game, sort of artistic game almost. So, you you play patterns that look beautiful and they turn out to be, you know, really strong, which is why the game has a little bit of a mystical element to it. Like almost encapsu- the top Go players would say to you encapsulates the mysteries of the universe in the game. I think that's how the ancient Chinese thought about it. And so, and and also just its raw complexity, as as you mentioned, has more possible board positions, 10 to the power 170, than there are atoms in the universe.

33:26 So, what that means is there's no way you can brute force it in the way that we did with chess. And furthermore, because the game's so intuitive and so esoteric, there aren't really these rules that you can encapsulate easily for a machine to follow. So, when you talk to a Go master, unlike a chess master, they'll tell you things like, "Why did you play there?" They'll say, "It felt right." Okay? That But as a chess player will never say that. They would say like, "I did it cuz I'm calculating this, this."

33:54 And then they'll tell you the calculation. So, that intuitive intuitive feeling is obviously very hard to encapsulate in a system. You can't really program that directly. So, it's the perfect proving ground, I would say, for these new techniques that we were pioneering in the early days of DeepMind of deep reinforcement learning. Can you use build systems that learn from themselves directly from experience? So, in the in the in the case of AlphaGo, AlphaGo started by looking at all the games on the internet that humans have played and learning the types of moves humans would do, but then we overlaid it with a Monte Carlo tree search that allowed it to sort of discover new branches of the tree of knowledge, if you like, in Go. Starting with what humans knew and then going beyond that.

34:40 And that's what we hoped was going to happen. So, so the amazing thing about that match, which was ended up being watched by 200 million people around the world, was that not only did we win the match 4-1, that was the main objective, but in game two, specifically, it played this famous move 37 that you talk about, this creative move that was it was on the fifth line of the board and it early in the game and it's it's sort of a big no-no to do that in Go, right? Like Go if you were being taught by a Go master, they would slap your wrist playing on that because it's just regarded as a bad move. And and and but not only was that great move, it ended up winning the game for AlphaGo. Like 100 moves, 200 moves later, it was in the right place. As if it sort of presciently put the stone there. So, it was a critical, not only was it a surprising move, it was the critical move for later for it to be exactly in the right place to decide the game. So, obviously it's changed the way all Go players play Go, but for me it was the moment I'd been waiting for in terms of building a system that we'd already spent six years by then building these types of learning systems that could achieve something no other system could, you know, this sort of Mount Everest of games AI, you know, the final frontier, if you like, of can you beat the Go world champion.

35:54 But also, not only did it win the match, but it was how it won and with these creative new ideas like move 37. And that for me was the signal that we were ready to turn it to scientific problems like AlphaFold. To say this back to you, the reason why it's important that this audience that wants to understand the future understand what happened with move 37 and Go >> Yeah. is because the implication is if DeepMind can build a system that can do that, it can also perhaps build a system that can play any game.

36:25 >> Yes. It can also perhaps build systems that can figure out in real world problems what is the best solution in quantum computing or in nuclear fusion or in matrix multiplication or what else do I have? Chip design. >> Or >> >> etc., etc. Could you tell me about the cutting edge here? Yes. Pick one of these systems. What is the move 37 of Yeah. the surprising creative element going on?

36:57 >> Yeah. I think AlphaZero is very interesting to talk about, which was the evolution of AlphaGo. So, after we we we we won, you know, got to the pinnacle of Go and showed that it could come up with new ideas, at least in Go, move 37 and actually many other ideas that it came up with, which has revolutionized how people, professionals, play Go now. We then generalized it further to a system called AlphaZero, which I think is going to turn out to be a very important system for today as well, where with AlphaGo, we started with all the human games that that we could find on the internet. And also there were a few other couple of things that were specific about Go that we were built into the AlphaGo system like the symmetry of the board and things like that. So, we wanted to get rid of all of those assumptions completely and actually start from scratch as if the the the the program and the algorithm didn't know anything about what it was trying to do to start off with. And that's why this is what the zero refers to in AlphaZero is sort of like AlphaGo, but now removing any knowledge, human-crafted knowledge, both in the data and in the any of the kind of heuristics that we'd given the system. So, AlphaZero starts like Tabula Rasa almost. Obviously, it's a has a learning system. It's a it's got a neural network. We we set up the parameters, but we didn't give it any domain-specific knowledge about Go or any other game. And then what we tested AlphaZero on was first of all, could it learn Go from scratch and then beat AlphaGo, right? So, and we managed to do that. So, it takes 17 evolutions of the program. So, you can imagine what happens is AlphaZero starts off random to begin with. It just it only has the rules of the game, plays randomly. Obviously, it's terrible at playing.

38:42 It creates its own data set by playing 100,000 games against itself, right? And then it can see what which moves won or lost. And even though it's playing more or less randomly to begin with, there'll be some moves that are slightly better than other moves, okay? So, now it takes the 100,000 it it we train a new version of itself on version two now of AlphaZero with that new data. That version two is slightly better than version one. So, now it's not random anymore, but it's not great, it's not good, but it's playing like okay moves.

39:10 And then those okay moves end up being better. And so then a version two gets trained a version three, a version four. And so, each time that new system gets played against the old system and sees is it significantly better or not. And it turns out that at least in Go and chess and things like that, around 16, 17 generations of that is enough to go from random to better than world champion. And at least in the case of chess, which I actually once watched live happen cuz I was fascinated by I was even playing chess myself, is, you know, it starts in the morning random, then, you know, by lunchtime I could still just about compete with it myself, and then by tea time it's better than all grandmasters, and then by dinner time it's better than the world champion. And you've just seen the entire evolution of that from scratch.

39:56 And also it's playing interesting new chess that that even chess computers like Stockfish, you know, with the more the kind of expert system brute force ones, haven't discovered those types of new types of moves. So, AlphaZero was the full generalization of the AlphaGo ideas. And interestingly, I think we need these types of ideas back here now with, our foundation models, the new, you know, Gemini and these kinds of things, which you can think of a generalized models of everything, language, the world around us, not just a game, obviously like Go, but we still need the this ability to search and think and reason on top of those models. And, sometimes we call those world models.

40:37 And, I think that's still hasn't fully been cracked yet how to do that. Bring back Bringing back some of these AlphaGo ideas, but now instead of just a narrow game applying it to that, but to the whole world, and maybe, interestingly, parts of science, too, like material design, and things like chip design, and, quantum computers, all of these cool projects that, you know, there's so many I just want to see all these bricks. I can't believe we're actually working on all these things, but it's true. Is And this is sort of the dream is like I get to I love all so much of so I mean I love every branch of science, and I get to, indulge myself in all these different areas of science because AI is such a general tool, it can really, make a huge difference to all these areas. So, maybe at one example I give is just designing new materials.

41:24 You know, if you want a material with a special type of property, can we go beyond, what is currently known, in material science? And I think AlphaGo-like processors could be very useful there. And the equivalent of a move 37 would be like AlphaTensor finding a new algorithm that makes, you know, matrix multiplication better. >> faster. Exactly. Exactly. So, you can apply it in algorithmic space, which is quite exciting cuz then the algorithm itself gets faster, so there's some circular circular sort of improvement there. And, yes, AlphaTensor, just making the matrix multiplication, which is the the basis of all neural networks.

42:00 It turns out everything's matrix multiplication. you know, if you just make that 5% faster, that's a huge, you know, the tens of billions being spent on training, that's a huge cost saving. And, and so these are good examples of of of, ideas in And I think we're still early, you know, also like things like the design of chips on a on a on a on a on a die, you know, making it as efficient as possible the routing. You know, it's a kind of, NP-hard problem, you know, in terms of like the the traveling salesman, like what's the shortest distance you can wire up all of these things. And AlphaChepan programs like that are really good, better in some cases than human chip designers at dealing with that. So, I think we're just scratching the surface, I would say, of what's going to be possible in the next few years with, today's kind of more general systems, combined with these types of ideas from from AlphaGo I and and AlphaZero I think are going to come back.

42:54 These two categories, the story that starts with AlphaFold, the story that starts with AlphaGo, these are the kinds of AI that make me feel really optimistic. >> Mhm. I also think that being really optimistic, and you do this a lot in public, which I appreciate, is fully thinking through the ways in which something can go wrong and what we can do to prevent that. >> Yeah. So, I want to insert one other in here. Sure. What are we doing?

43:20 >> >> This one. Yes. And the reason why I bring up this game is this is a real-time war game. Yeah. And in the videos where this system is absolutely crushing humans, >> Mhm. you can see the engineers cheering for the victory of their system. But of course, as someone who didn't build the system, I'm thinking to myself, what if that's real? >> Mhm. And we're speaking right now during a time when the debate about militaries and governments using AI is a huge topic of conversation.

43:59 >> I want this conversation to last for 10 years. I want it to be useful for that long. So, I don't want to talk about specific companies, specific terms of service. >> I also think people are in some way missing the forest for the trees here. Because bigger picture, governments are going to use AI. >> And so, what I want to know from you as someone building these systems is if you could wave your magic wand, >> Mhm.

44:23 what would you hope that they use it for? Well, look, I think governments, and governments should be using, AI, and, you know, we want to support all sort of democratically elected governments. And I think, the things I would love to see them use it for and what we're trying to build our systems to be good for is, things like improving public health, education. I mean, all of these things need to be rethought. The efficiency gains and the amount of good we can, do with it, governments could do with it for their citizens could be incredible. And I think some countries are doing it like Singapore and UAE, I think are leaning into, these types of use cases. I would love to see it being used for, things like energy, like optimizing energy grids.

45:07 we did that with our data centers and save 30% of the, you know, energy used for the cooling systems. I think there's enormous societal gain from applying AI at scale to these types of areas. So, that's what we, you know, I've always thought about and and and and, hope that, governments will pick up and use for and we we, you know, want to support all of that. of course, you know, the geopolitics of the of of the world is very complicated right now, and these are dual-purpose technologies.

45:37 And, you know, I worry about a couple of use case things that can go wrong with AI that, you know, in the bigger picture, as you say, I think sometimes the the is the the the kind of details are get people get bogged down in the details, but actually there's big the big picture, there's two things to worry about. One is bad actors, whether that's individuals or all the way up to nation states, using, repurposing these technologies that we're trying to build for good, like curing diseases and advancing material science and energy and so on, for harmful ends, right?

46:09 And whether that's, inadvertently or intentionally. and then the second branch of things I worry about, is the AI itself, going rogue, or going off the rails it it it if as they get more powerful. That's not today's systems, but maybe in the next two, three, four years, especially as we go towards more of the agentic era, which we're entering now. And by agents I mean systems that are capable of completing entire tasks on their own. So, you can Of course, we want those cuz they'll be very useful, like as an assistant or something like that. But also that means they'll be increasingly capable and autonomous. And so, how do we make sure as as one of the frontier labs, and the frontier labs all have to think about this, is the guardrails are put in place that they end that we can ensure that they do exactly what they've been told to do or the goals they've been given, and they've been specified clearly enough, and there's no way of them circumventing that or accidentally, breaching those guardrails. And that's going to get That's an incredibly hard technical challenge if you think about how powerful and how smart and capable these systems eventually going to get. So, I tend to worry about those You could call them medium term now, even though three, four years is not really medium term, but those are the things I think people are perhaps not paying enough attention to at the moment. And I think we'll, be the biggest, issues that we're going to have to contend with if we're going to get through the the AGI moment in in a in a in a way that's beneficial for for humanity. Yeah, one of the biggest questions I came in for you with, you know, if I get an hour with you in my life, was next time I read a headline, how do I weight the concerns that we're all going to have over the next 30 years? You know, like what are the things that people are worrying too much about, and what are the things that they are not worrying en- enough about?

47:58 >> Yeah. So, I think the two things I just mentioned are the things that maybe the average person is not worrying enough about, but even I think some of the experts and the scientists in the field. I feel like those are the key things that are more societal affecting, that if we if we There are other things that we need to worry about, too, like deepfakes and and we and we try to help Those are immediate term worries, right?

48:19 Misinformation, deepfakes, these kinds of things. And, you know, we work on this system called SynthID, which is, you know, a watermarking system, actually an AI watermark, probably somewhere one of these bricks. Yeah, one of them. And and and, we need AI to sort through them. And and and it's it's it's uses AI to actually digitally watermark any generated image. So, all the things all the Google technologies, VO and everything else and nanabanana, they all they all they all have this, watermarking technology. So, we can detect and flag to the user or or government or whoever that these are fake. and I think actually I would advocate all, companies working on generative, AI, should build in something some kind of technology like that. So, at least, it can be detected or they can detect which things have been built with their, technologies. And I think that's going to be increasingly important. But I think that still pales in into into, you know, as as a small issue compared to some of these bigger issues around, AGI itself, becoming very capable and how do we put make sure that, you know, guardrails are put in place that we understand, what that those types of systems are are capable of as we get towards AGI. And, you know, I think a lot more research, a lot more effort needs to go into that from from everyone. And actually, I would love to see international cooperation and cooperation around amongst the you know, leading labs around the safety issues. And and including you know, with with with places like the AI safety institutes and and also academia to help kind of work out how we navigate that next step cuz it's unprecedented to create technology like that. If we play this out, what's the limit here?

50:06 What are the things that you think AI cannot do that humans can do? You've called this the central question of your life. >> it is and it's very related to you know, some some scientific thinking of some of my all-time heroes like Alan Turing. You know, he described Turing machines which were these theoretical constructs that actually all modern computers are basically Turing machines that are able to compute anything that's computable. So, anything that can be described as an algorithm this this type of machine can compute.

50:40 And I think that the systems we're building are approximate Turing machines and potentially a lot of neuroscientists including me think that maybe the brain a good model for the brain is an approximate Turing machine. So, the question is and but there are others like friends of mine like Roger Penrose and who you know, believes there might be some quantum effect in the brain, right? And I'm sure you probably done videos about it that that and he you know, we've had some very good-natured debates about this. But so far neuroscience hasn't found any quantum effects in the brain. Doesn't mean they won't be found, but so far people have looked quite carefully and they haven't we haven't found any. So, it looks like most of what's going on in the brain is kind of classical computation. And so therefore, it's not clear what the limit would be in terms eventually what an AI system could do and could mimic. But you know, I think that's an empirical question. I think that's one of the you know, the questions around consciousness. I mean, I don't think it's very well defined what it is, but we all intuit what it is. And I think this journey we're on of building an intelligent artifact, I think we'll have almost like a controlled study comparison to the human mind and then I think we'll see in this journey like what are the differences and what's unique about the mind. And I'm very open-minded about that. I think there could be unique things and certainly unique connections between humans that will never be replicated by you know, these AI systems. But I think a lot of things that we currently are not in reach like long-term planning and reasoning and maybe some forms of creativity, I think eventually AI systems will be able to do.

52:22 I want to be honest about what's happening in my mind right now and it is that I am doing exactly the thing that humans have done throughout history. I am trying to find the reason why we are special. It is that we have to be at the center of the universe. Oh wait, we're not. We have to be the ones that are emotionally attuned. Oh wait, elephants have funerals. Oh we must be the ones that can be creative and create art. Oh wait, Gemini can do that or like what Oh, we must be special. Do you find yourself doing that as well? That's my reaction as you're describing the future of AI.

52:53 Yeah, no. I think I think we are special and I think there is some there's a lot of deep mysteries about how the universe works and including a lot of things that that are in our minds, but also are things out there in physics. You know, I I I I think that's why I got I think I decided from a very young age to do AI is because I was obsessed when I was a kid at school with with with the big questions. And normally when you you know, physics was my favorite subject at school because that is the subject you're supposed to study when you're interested in all the big questions. And but the thing was I just realized I guess as a young teenager reading all these science books and biographies on the best scientists. Richard Feynman is one of you know, my all-time heroes at that they actually although they we discovered a lot and we know a lot about the world, there's so much we don't know. Like there's just incredible like we don't know what time is. I mean, this is this is insane to me. Like we you know, we don't we can't even describe something as that. It's just we're swimming in it, but what is it? We you know, of course it's you know, entropy and things like that. But it's nothing it's nothing satisfactory about what it really is. And you know, we don't understand a lot of quantum effects and gravity properly and and consciousness all actually most of the things we we care about. And and but we just sort of I feel like most people we just distract ourselves all day with you know, TV shows and games and things and don't worry too much about it. But I'm I've never been like that. I've it's just it it sort of it's it's it's these deep mysteries kind of play on my mind all the time. And I think I'm quite open-minded about what the answers might be eventually about what's going on here, the nature of reality. I think that's ultimately what I'm after and I want to use AI as a tool to help us understand the nature of reality around it. And I'm quite sanguine about whatever the answer might be. I'm not you know, I'm I guess I'm a true scientist in that sense of like I don't actually I don't really have any prescribed notion of what the answer should be. I just want to know the answer. Me, too.

54:49 >> >> One way to describe what you're trying to do is effectively this. Which is to say to create a system that wouldn't be especially good at one thing or another thing, but rather to create as you've been saying AGI, artificial general intelligence that would be good at it all. Yes. I know you're a fan of sci-fi. I am, too. Could you play out for me the plot of the sci-fi movie in your head that is the future where you actually do this? Yeah. I can. I I think I love sci-fi, too and I probably I read too much of it when I was when I was a kid might explain a few things, but one of my favorite series was the Culture series by Iain Banks.

55:34 I think it just paints a really interesting actually post-AGI world. He didn't call it AGI, but that's what it we've describing like a thousand years in the future. But I think even 50 years some of this could happen where we've we've got through the AGI moment safely. It's built. It's it's it's helpful for society and it's it's and and you know, it's it's here and maybe we will have it in our pockets even. And we've used it to crack some of these what I call root node problems in science. AlphaFold was one of those, right? So, these are problems if you think of the tree of all knowledge, these are kind of root node problems which if you cracked it, it would unlock a whole branch of new research or new applications. And I think there are other things like fusion we briefly mentioned or better maybe room temperature superconductors at atmospheric pressure that you could then combine with optimal batteries and things like that. I think though there will be a solution to the energy problem. So, free pretty much free renewable clean energy one way or another fusion or you know, better solar. And then that will unlock us to really travel the stars because the main cost of you know, Elon does amazing work with SpaceX and those things, but the main cost is still the rocket fuel, right? The energy cost. So, if that's sort of zero somehow because we can just make infinite rocket fuel out of seawater because we have we cracked fusion. So, we can have you know, catalyst plants and desalination everywhere, then you know, that unlocks the really unlock space. And then we'll be able to get a lot more resources because we can mine asteroids. All of these things that the purview of science fiction become I think very plausible in the next 50 years. Dyson spheres around the sun.

57:16 Mercury's sort of conveniently in the right place actually made of the right material which is kind of amazing if you think about what's going on in the universe. And and and then that should hopefully to you know, maximum human flourishing and we help cure all these terrible diseases. So, we live much longer healthier lives and traveling to the stars bringing consciousness to the rest of the galaxy. That would be I think an amazing outcome and I think could happen within the next 50 years.

57:46 I believe you. You're saying these things and I like when you're saying that my belief you. That's the that's what I'm trying to do at least. Yeah, so this is my last question. If I were a fly on the wall at my own funeral after they said she loved her husband and her family and her friends I would hope that they would say >> >> that she spent her life trying to help people see optimistic futures so that they can be part of making them happen. That they can make them happen more quickly or better for more people or whatever it is that people decide to do with the vision that they see. And so my last question for you is what do you hope that they say about you? I would hope that they would say that you know, my life was of benefit and service to humanity.

58:31 That's I think what I'm trying to do. So, that's maybe would be the best thing. Thank you so much for your time. Thank you. Really appreciate it. >> Thanks. Awesome. If you want to play Jenga anytime we can play we have a modified version of Jenga. >> You did that very well. So, yeah, this is actually awesome. I can't believe how many projects we've got. It's really crazy when I saw the bricks. So, they all got up. Yeah, they have all got our projects on them. Did you memorize where everything was? Okay. Of course. So, the the game is you pull it out and we were playing this. It It's unfair to play with you, but it would be you have to say what that project was. You don't get the point if you get it wrong.

59:06 >> >> So, for example, it would be gnome this is material science. Yeah, it's it's a little bit unfair on you. I mean, I would hope I would win this game, >> >> but you're probably a way better at Jenga than me. Yeah, let's see let's do this one. There you go. Okay, alpha code. Yeah. That one that one's clearer, right? Code code forces. Yeah. From a sense. This is genetics, but the 2% that codes for proteins?

59:32 >> Yes. We have to do this now. I've got time. I can push back my next >> Great. Wait, wait, I have one I have one more question. You know AlphaFold? AlphaFold is coding. Yeah, we can be used for coding. >> programming? It's it's combining genetic algorithms with with Gemini. So, this is this is our this is one attempt at doing like AlphaGo stuff beyond what is known. So, I wouldn't get a point for that one.

59:57 >> No, half a point, half a point. Okay, one more question for you then. While I have I'm just going to keep going while I have you cuz why not? >> we're still rolling. obviously. >> >> Okay, what did I not ask you that you think is important for people to know? what did I not ask me? I think we covered a lot actually. GenCast, this is weather prediction. >> Yes. Oh, yeah, we didn't cover that.

60:18 Navier-Stokes, I completely forgot about solving solving that whole branch of things. >> forgot about that thing you did solving So, that was one interesting thing is simulations. We didn't talk much about that or Genie, which is the the role of simulations to DQN, of course, started it all off, the Atari stuff. Simulations to help you understand some area of science that or even social science like economics that you can't a very hard to run either expensive to run experiments or you can't run controlled experiments in. So, I've always loved simulation. Oh, yeah, I said there we go.

60:57 We're both very competitive, I think. So, this is going to this is be quite serious. Actually, are you in Jenga, are you is the rules that if you touch it, you have to move it, do you or not? We are playing a a loose a loose version. >> The easier version. >> Also, because we we were doing a creative thing where you're allowed to like push them together, you can use two hands also. >> Oh, okay, you're not allowed to normally do that, right?

61:18 I'm just going to take this one. I'm going to cheat with AlphaCode again. One of the questions I think people will have for you is if they're watching this and they you know, are are very optimistic. Gemini, everybody is. They're very optimistic about the futures that you've described. They're they have all of your same concerns. They generally have gotten to the end of this conversation and they're thinking I believe in this future and I want to be part of it.

61:47 How would you Code Mentor? I think that finds bugs in code. >> Yes. Very good. How would you advise them to participate if this is all about helping people participate in the future? >> I would when I do sort of talks at at universities and schools, I would say they you've got to just go with the flow of the direction. I would immerse myself in every tool available and just become almost like super powered, fantastic super powered with those tools and those those capabilities cuz I think my impression is even at the frontier labs, we are the so much work has to go into just making the next versions of these frontier models and then all the adjacent models. So, for us like VO and Nano Banana and Gemini that we even we can only explore a fraction of what the the applied things you could do with it, the applications you could you could make with it. So, and I think that gap's getting bigger and bigger in terms of like the overhang of the capabilities, all the cool stuff on the latest models and that that time the the release schedules are getting faster and faster on that. So, I think the opportunity space is getting huge if for people who are really expert and at using those tools and then apply it to some new domain. So, I think a a kid these days could probably start a multi-billion dollar business in some ways using these tools in some new way that no one had thought about. And I think things like Open Claw is a good example of that. Yeah.

63:24 Maybe we should call it a draw cuz I think I think I don't think I don't think either of us could bear to lose that, right? So, It's your move. It's your move. Yeah, it's your move. We can I will try my my inner my move. Go on then. >> >> You're going to make me make a fool of myself. In 2016, you had a sticky note on your board that said solve protein folding smiley face.

63:46 >> Yeah. What is >> >> Did I Yes. Okay. What is on the board now in your proverbial sticky notes? >> To answer it, I've got a pile of about a hundred sticky notes on my desk, so What's what's on it? Alpha This is What's on it? What's in it? I can't actually remember. It will be a list of about 30 things that need to be done by like this evening. So, >> >> I better probably get to them.

64:12 But look, great. Should we Do you want to you want to actually I'm going to keep going until you stop. So, you can stop whenever >> Yeah, let us about to What time is it? Okay, I'll I'll do one more move. But now we're now we're kind of cheating. We're we're using our the pieces that already I'm going to I'm going to try and be I'm going to go ambitious in the last Oh, come on. Come on. Come on. If I get this one, I get another question. Yeah, okay.

64:38 That seems fair. Oh, god. How is that going to balance? Surely not. No. Yes. >> >> That was awesome. Thanks. That was a great great idea to have that.