[0:00:00] You need to pick hard problems to solve and be ambitious with that. Well, actually for me, my interest in science and AI started with games. So I started to play chess from a very young age. I was actually four years old when I learned. And I took it very seriously. I was captaining all the England chess junior teams as I was growing up.
[0:00:24] And that's how I actually came across computers for the first time, because on some of the chess training camps, you would use chess computers, physical boards to try and train and get better at. So as we were playing against these chess computers,
[0:00:44] I remember being fascinated by the fact that someone had managed to program this inanimate lump of plastic to actually play chess. So although I was meant to be training my chess openings and things, I actually got more fascinated by the program and the AI underneath it. And then I ended up wanting to program my own AI systems because of that. So that's how I got into computers and then AI.
[0:01:05] And then later on, the reason I spent my whole career on AI is because I believe it could be the ultimate tool to help with science. So to use AI systems to find patterns in data, insights in data and structure, and then help us advance scientific knowledge. So my aim was as always to build AI algorithms that were general enough to eventually be able to be applied to scientific problems like protein folding.
[0:01:33] Well, I think if you look at biology, I think I've thought of biology as an information processing system and a fundamental level. And biology is a hugely complex, emergent, dynamic system. And it's quite hard. I've always believed it's quite hard to describe that system just with a few mathematical equations. So it's not like physics.
[0:01:52] So actually, I think AI systems and AI algorithms are quite a good match, good description language for biology. So I've always thought biology would be one of the places where we could apply AI most usefully.
[0:02:06] And then specifically proteins and protein folding. I knew, actually I came across protein folding problem as an undergraduate at Cambridge for the first time in the 90s. And I thought it was a fascinating problem, kind of like a puzzle really. And also a puzzle that if one could solve it, you would unlock so many new avenues of research and important research and discovery like drug discovery and disease understanding. So it was a fundamental problem and it was a really interesting puzzle. And I thought it was also...
[0:02:35] So would be very amenable to AI techniques.
[0:02:40] Well, I've used gaming. I mean, games has been a passion of mine and a core part of my entire life. I think I've used gaming at least in three different ways. So first of all, to sort of train my own mind. And when I was a kid, you know, actually, if you play chess and things like that very seriously, then it actually, it becomes very formative with the way that you think about the world and, you know, and formative of how I approach problems. So that was the first step.
[0:03:10] The second one was actually writing AI for computer games. That was my first career and designing AI for commercial computer games and exploring the world of AI through that. And then finally as a third way is with DeepMind and my AI company actually using games as a test bed, a proving ground for the AI systems to learn and see if these learning algorithms could work. So yes, games has been a critical part of my entire career.
[0:03:41] I wrote that when I was 17, Theme Park. Yeah, and you know, it was probably my first big success, I guess. But that, of course, has AI as the heart of the game. That's why everyone liked Theme Park, because anytime you play, as you'll remember, anytime you design your own theme park, all the little people would come in and play on your rides and everything. And every time, the game would be different.
[0:04:04] Because the AI would react to how you as the player design the theme park. So that was one of the things that after I wrote that and I saw how many people enjoy playing it and interacting with the AI that made me convinced me that AI was what I wanted to spend my whole career on.
[0:04:20] I do still find little bits of time to keep my hand in with things like chess. I play chess sometimes late at night online, just a little bit like going to the gym for the mind, you know, to keep the mind sharp. So I love doing that and I play some video games and things with my, and board games with my children. We used to play League of Legends as a team with my kids and my brother and a good friend of ours.
[0:04:46] Well, I think science is the most important. The scientific method is probably the greatest invention humanity's ever come up with. And I think science is what drives all progress. And modern civilization comes from scientific and technical advances. And I think it's the most important thing we could apply AI to is to advance science and things like medicine. So for me, it's the number one thing to apply a science for the benefit of humanity too is to
[0:05:16] It's things like diseases and understanding and advancing human knowledge. Yeah, it's an amazing feeling. That's what's been so gratifying, I think, about AlphaFold. And we put it out there, open sourced it, put it out there freely for the world to use. And over 2 million researchers have used AlphaFold and the structures we predicted.
[0:05:37] for their important work. And it's been used for so many things, so many great research that we couldn't have imagined. And that's been great to see. And I think we're probably only at the beginning of the impact that AlphaFold will have on things like drug design and understanding biological processes. Yeah, of course, like any career, a successful career, there's obviously been ups and downs and mine's no different. I've had a lot of successes, but
[0:06:08] I've also had some setbacks. My original first games company, we had some difficult times, in part because I think some of the ambitions that we had for the game design and the technologies were a little bit...
[0:06:24] So I sometimes I've described it as being, you know, 20 years ahead of your time. And actually, what I realized was that you need to pick timing is very important. You need to pick hard problems to solve and be ambitious with that. But you've also got to pick the right time when the sort of the world and the context that you're in is the right kind of environment for those ideas to flourish. And that's what I took into my later career to that learning.
[0:06:53] I was better able to calibrate that later on in my career, I think, in terms of picking really hard problems, but at the right time for where we could make serious progress.
[0:07:04] I think absolutely. I actually don't really regard it as failure because it's so fundamental to advancing anything in science. If you're working at the frontier of knowledge where no one's been before, which is exhilarating, then of course it means sometimes you'll make hypotheses that turn out to be not right. But that's just part of the past of exploring the frontier. And actually, if you design experiments well, then either answer whether it works or it doesn't work.
[0:07:34] So I actually, you know, I don't even think of it as failure. I just think of it as a necessary part of the scientific process.
[0:08:00] Yeah, something I've always had as growing up as, I guess, at the time, a sort of child chess prodigy and then programming. I did that early. And so I've always been in a bit of a hurry. I think I've always had this notion of that life's pretty short if you want to explore and experience all the amazing things in the world and maybe add a little bit of impact and value to that. And I've always had an unbelievable drive since I was...
[0:08:28] I can remember. So maybe that's just part of my genetic makeup and character. And then the way I approach successes is each time I've had success in my career, it's just maybe more motivated.
[0:08:43] For the next thing to think even bigger. And so it means I'm not that good at celebrating successes. Perhaps I need to get better at that because I'm always looking straight away as soon as I do something onto the next thing. But it served me quite well in terms of my career and driving me forward to do more and more ambitious things.
[0:09:02] Well, I still have a lot of ambitions to carry out. So actually, my original ambition in AI is to build what we sometimes call artificial general intelligence, which is a system that can exhibit all the cognitive capabilities that humans can. And then we know we have a truly general system at that point, a kind of Turing machine.
[0:09:23] There's still that to build. And I think that would be one of the most consequential inventions humanity will ever invent. So I guess that was my original goal when I got into AI 30 plus years ago. And that's still my goal now. And we still have a lot of work to do to get there.
[0:09:41] Well, I've had many amazing people influence me at different times, but a lot of it actually comes from reading the greats and trying to understand and take inspiration from them. So actually, when I was a kid again, growing up, and the kind of formative books and people I read about were things like Richard Feynman. Actually, not so much his physics books, but his more layperson books. Surely you're joking, Mr. Feynman. You know, the joy of finding out things.
[0:10:11] So that actually I really recommend to any students who are watching this to read those books because I think he really explains and gets across how enjoyable, how thrilling it is to be exploring at the frontiers of science.
[0:10:26] And then my other kind of all time heroes are people like Alan Turing, Claude Shannon, Von Neumann, the people who sort of founded the computer era and the information theory era and also sort of started off artificial intelligence, I suppose.
[0:10:42] I think when I was probably the most important teacher for me was my high school maths teacher for my A levels. Mr. Lovegrove was his name. And he not only taught us maths really well, but actually he was, I realized looking back on it, he was teaching us how to think.
[0:11:01] And how to break down problems in a really disciplined, efficient way. And that combined with my chess upbringing, I think was really great informative for me in terms of the way that I try and approach problem solving now. So that was great. And then probably in terms of like my biology knowledge and being inspired to apply AI to biology, it'd be people like Paul Nurse, who I've known for over 25 years now. And I've discussed with him many times about...
[0:11:30] You know, biology as an information processing system or an information system. Well, what I would say is actually I'm a really big believer in interdisciplinary research and interdisciplinary science.
[0:11:44] And I think that is what a lot of the next 10 years or so, a lot of the big advances will come in the combination of two or more subject areas and sort of the in-between points where they connect. And so for me, that was AI and neuroscience originally, and then now, of course, AI and biology. And I think there's so much potential now of combining a couple of subjects together and then find something new. So what that means is...
[0:12:13] Being quite a generalist and a sort of polymath in a way and getting quite deep understanding and knowledge of at least two areas, which is quite hard to do, but I think it's extremely worthwhile. And I think increasingly we'll see connections across subject boundaries. For me, maybe it would be to believe in your ideas.
[0:12:39] And even though they seem quite outlandish, of course, I did believe in them. That's why I pursued them. But perhaps I had less faith in that than I should have had, that those ideas would bear fruit and with enough passion and work and effort. So, you know, I think having confidence in your ideas from an early stage is very, very important. Well, I think this is where I think intrinsic
[0:13:07] Passion for what you're doing becomes important. So this happened with me when starting on AI. First of all, 30 years ago, almost no one was working on AI. And then even when we started DeepMind back in 2010, nobody was really working on AI, certainly not in industry. And then you fast forward to today, and everybody seems to be involved in AI or interested in it.
[0:13:29] And we sort of foresaw that. But at the time, everyone thought we were pretty crazy. Everyone was like, well, we know AI doesn't work. But I was going to do it no matter what, because I think it's the most fascinating technology to try and develop.
[0:13:46] It also has so many aspects to it, including trying to understand the workings of our own minds. So for me, there's nothing more fascinating than a phenomenon of intelligence and trying to understand what it is. And I've tried to do that from many angles, both from neuroscience and computer science. And so for me, I would have spent my whole life doing that no matter what. So I think that gets you through the hard times because actually it doesn't really matter what other people are saying. You're doing it through your own passion and interest.
[0:14:17] So we're huge believers in open science and open access, and we've done that with all of our scientific work. And it's because, I mean, that's the way that science progresses quickly, I think, is the sharing of ideas, the critiquing of each other's ideas, and people building on top of other people's work. And I think if you look at something like AlphaFold, that's why we put it out there for everyone to use and all the predictions we made of the structures, because we knew...
[0:14:42] That we could only do a tiny, tiny fraction of the downstream work that AlphaFold could enable. So we are trying to do some of that ourselves with drug discovery and so on. But it's been amazing to see what hundreds of thousands of researchers around the world have been doing with AlphaFold that we could never have imagined or never even heard of. That has been spectacular to see. So I think that's a good example of how science can be disseminated quickly.
[0:15:11] And then everyone can use it. And then the whole of humanity advances that way. Yeah, it's really important. I mean, first of all, it's well understood if you have a diverse sort of research group, you get better, more useful ideas.
[0:15:27] I'm really keen on sort of diversity of expertise and background. So in DeepMind, in our research groups, you know, we've always had at the beginning multidisciplinary groups. So at the start, engineering with machine learning, with neuroscience, mathematics. And then since then, we've added philosophers, ethicists.
[0:15:51] Chemists, physicists, and so on. So actually, it's a real melting pot of expertise and ideas. And I think that's a lot of the strength of the work we do comes from that, including things like AlphaFold, which is a big collaboration between people with very different world-class skills in their own areas, but come together into a multidisciplinary group. So I think that's hugely important for science.
[0:16:12] I'm also really gratified to see AlphaFold used all around the world. So it's 2 million researchers from around the world, but from 190 countries. We ourselves have done collaborations with places like the DNDI, Drugs for Neglected Disease Institute, part of the WHO, that concentrates on helping with diseases that affect the poorer parts of the world. Sometimes in the global south, things like leishmaniasis or dengue fever or these.
[0:16:41] These kinds of diseases that are sometimes under-researched by big pharma and it's been really pleasing for us to just be able to give those non-profit institutes the protein structures so they can get on straight away with designing drugs and trying to find cures or mitigations for those diseases. So I think that's been one of the most satisfying collaborations we've had.
[0:17:06] Well, I think the possibilities with AI are almost unlimited if you think about what intelligence is.
[0:17:13] And let's take human intelligence, first of all. Human intelligence always astounds me. And I don't think we think about this enough. It's created modern civilization around us. And sometimes when I'm flying over to the US for a business trip or something on a 747, I sometimes look out the window and think, how have we as humanity managed this with our sort of primate brains? It seems incredible to me. And I don't think people stop and think how magical that really is.
[0:17:43] So really the power of intelligence, human intelligence, has created the wonders of modern civilization. So we shouldn't be surprised that intelligence is sort of like this incredibly powerful force of nature in some sense. And I think the science of AI is about trying to explore and understand what that is, what intelligence is. And the best expression of understanding something is actually trying to build it. And so that's what
[0:18:12] That's what the engineering science of AI is about. And if you build it in a general enough way, like we're trying to do with artificial general intelligence, the whole point of that is that it could be applied to anything, any kind of problem, any kind of data, much like the human mind. We seem to be infinitely adaptable in a way, and we've invented modern society around us with our brains.
[0:18:36] Well, I think it's hard and I think part of that is outside the scope of the technology. It's a geopolitical question. But I think the technology itself should bring incredible productivity gains, economic gains.
[0:18:49] And scientific gains. And I think then it's important for society to decide collectively what we should use this incredibly powerful general purpose technology for, what we shouldn't use it for, and then how the benefits of that should be spread and shared around the world.
[0:19:08] And I think it's important for technologists and scientists like us who are working on the technology to engage widely with civil society, governments, academia, industry, all to come together to start debating these questions. And at Google DeepMind, we've tried to facilitate a lot of those conversations and convene those. And there's a lot more to do on that front. I think there are two concerns on AI for me. One is
[0:19:37] These are general purpose systems and very like any
[0:19:41] Powerful general purpose technology. It can be used for good, which is of course things like medicine and climate and help us with some of humanity's greatest challenges. But it can also be repurposed by bad actors for harmful ends. The same technology. So it's really, the technology itself is kind of neutral. It depends on how we as scientists and humanity and society decide we're going to use it and deploy it. So I worry about...
[0:20:10] Well, I hope that
[0:20:39] We'll be part of, and AI as a field will be part of, helping humanity solve some of our greatest challenges. So for me, that's things like finding cures for terrible diseases, hopefully using tech tools like AlphaFold, but also helping with things like climate change, maybe through designing new materials or helping with new technologies like fusion or better batteries. And I think AI could help with all of those things.
[0:21:06] I have very little spare time. I try to make sure I get enough sleep. But yeah, gaming still is one of my main outlets as a relaxation. And then probably football as well. You know, I'm a big follower of Liverpool Football Club, another game, of course. And I used to love playing football as well myself when I was younger. So maybe those are my main outlets.
[0:21:32] I don't really think of things as work-life balance. So for me, I guess I've been very lucky in that I've designed my work to sort of be what my life passions.
[0:21:43] So whether that's gaming, whether that's neuroscience, whether that's AI, whether that's biology. And so I have a lot of interests and I get into them very deeply. And I love learning. That's my main thing. And sort of mastering new skills or new domains. That's what I love doing. So I sort of designed, I guess, my career to allow me to explore that to the maximum. So I don't really see any distinction in my life between.
[0:22:12] It's all just one big adventure. Well, when we got the call, you know, the famous call from Stockholm, to be honest with you, I was in total shock for the whole day. I think I was stunned and it just felt surreal and like a dream almost. And I think for the next two, three days, I was just, my mind was just...
[0:22:36] Scrambled. So, yeah, it was an incredible feeling, amazing feeling, and one of kind of shock and, yeah, amazement. Actually, what I did was, which was very suitable talking about our games stuff is, so I think I found out on a Wednesday, and it so happened that the next day...
[0:22:57] I already had something long arranged, which was some of the world's best chess players and poker players were in town in London for a big chess tournament, including people like Magnus Carlsen, the ex-world chess champion. And one of my old...
[0:23:11] One of my childhood chess friends was hosting a poker and chess evening at his house and all the world's top chess players, several ex-world champions of chess and poker turned up and we had a really fun games evening playing poker and chess and long into the night. So actually I couldn't, for me that was the perfect way to celebrate and it couldn't have been more apt so that was lovely happening literally the next day.
[0:23:39] Well, this is, you know, it's kind of like the realisation of a lifelong dream and it's such a great honour. It's the honour of a lifetime, of course. And it's always been my dream, really, from reading about those greats when I was a kid to maybe do some work one day that would have the impact, that would be worthy of being included in the pantheon of these great laureates. And so it's sort of...
[0:24:05] It's kind of unbelievable that that's actually happened and all of my scientific heroes from the past and sort of joining them in this exclusive group.