Demis Hassabis Thinks We’re in the ‘Foothills of the Singularity’

Full Transcript

Note: This transcript was generated by an automated system and has been lightly edited for clarity. It may contain errors or omissions.

Jennifer Aaker: Hi everyone. I’m Jennifer Aaker, the General Atlantic Professor at Stanford Graduate School of Business.

Michael McDowell: And I’m Michael McDowell, a producer here at the GSB. And today we’re live from the GSB.

Jennifer Aaker: That’s right. You’re about to hear a conversation between my friend and colleague Jon Levin, President of Stanford University and Demis Hassabis, CEO and Co-founder of Google DeepMind.

Michael McDowell: Jennifer, I think it’s fair to say that you made this conversation happen. So tell me why you invited Demis to the GSB?

Jennifer Aaker: So many reasons. But honestly, the thing that drew me the most isn’t AI itself, it’s Demis. Before DeepMind, before AlphaGo, his dissertation explored the neurological relationship between memory and imagination. He showed that patients with hippocampal damage lost not only the ability to recall the past, but to construct any rich mental scene moving forward. To imagine you have to have lived.

Michael McDowell: This was a big breakthrough back in 2007 and it raises a pretty profound question given what Demis went on to build.

Jennifer Aaker: It really does. If AI increasingly handles planning, remembering, scenario building, what happens to human imagination? What capacities do we most want to preserve, deepen and expand?

Michael McDowell: Yeah, it’s not just what we might lose, but what becomes possible.

Jennifer Aaker: Absolutely. I’m thinking about scientific advancement, economic empowerment, creative potential, expanded agency, and ultimately fuller human lives. At the micro level, I think of this as flourishing, which means feeling fully alive, creative, connected, and purposeful. And if we’re able to accomplish that, really exciting things start to happen at the macro level across organizations, communities and society.

Michael McDowell: I keep thinking about something Demis said, which you’re all about to hear, which is that the future is still unwritten.

Jennifer Aaker: None of this is inevitable. I think Demis is asking many of the same questions I am, and understands maybe better than anyone how the choices we make now may alter not only our future, but the nature of human existence. And that’s why I was so excited to bring him here to Stanford, the place where big ideas meet real application.

Michael McDowell: On that note, thank you so much, Jennifer.

Jennifer Aaker: Thank you, Michael.

Michael McDowell: Here’s Demis Hassabis in conversation with Jonathan Levin.

Jonathan Levin: Demis, it’s great to have you here at Stanford. Thank you for —

Demis Hassabis: It’s fantastic to be here, thanks everyone for coming.

Jonathan Levin: Really appreciate you doing this. So I’m going to ask you some questions. We’ll have some questions from students and looking forward to hearing your thoughts. You’ve been chronicled a lot recently, a movie, a book, so many people have heard about your trajectory. It is quite remarkable. Chess prodigy, video game developer, scientist, tech entrepreneur and leader, Nobel Laureate. That’s just the first half of your career. So if you were going to try to draw a through line through all those different things that you’ve done, what would it be?

Demis Hassabis: Well, I think there’s several through lines actually with what seems maybe somewhat unconnected subjects. First of all, I’ve always really enjoyed working at the intersection of creativity and technology and very broadly construed. So actually the games industry, the video games industry, which is my first very early part of my career in the ’90s, was one of the most creative spaces in any industry that was using cutting edge technology with art and design to create an entirely new entertainment medium. So that was really an amazing time. In fact, some of the most fun times I’ve had in my career was early in the ’90s. The chess and the neuroscience I did, all of those things I’ve tried to… I had from this idea of working on AI and AGI being the most important thing and most interesting one thing one could spend your career working on also from a very early age.

So, as a teenager, probably I read too much science fiction, reading things like Gödel, Escher, Bach, these types of books and biographies of some of my scientific heroes, Turing and Feynman and so on. So all of these were serving to inspire me to try and understand the world around us in a really deep way. And then building AI was my expression of that mission to try and build the ultimate tool for science. And I’ve tried to, because life’s short, I’ve tried to reuse and repurpose every experience I’ve had in service of that bigger North Star mission that I’ve had for more than 30 years. So my chess training is the way that I think about business and organizing things and planning and how I think I’ve been able to break down very ambitious plans into small and more manageable steps. That all comes from chess thinking, I would say.

And then using games, first of all, building games, learning about engineering projects at scale, running companies, startups, and then fusing this creativity with engineering. Actually, it’s what we do today with AI. It’s an engineering science. So you’re fusing creative work, scientific work with very hardcore cutting edge engineering. So that all served together. And then finally on games, as everyone knows, we used games in the early days of DeepMind as the perfect proving ground for testing out algorithmic ideas, probably most famously with AlphaGo, which I think we’ve just had the 10-year anniversary of and was really looking back now, maybe the start of the modern AI era.

Jonathan Levin: When you went into AI professionally in 2010 or so, you started DeepMind, you had this very ambitious vision. You were going to solve intelligence and then solve everything else. How’s it going? Let me expand a little bit. What has gone according to plan and what has been off plan?

Demis Hassabis: Well, the broad arcs of it have gone, I mean, unbelievably well, perhaps. When we started DeepMind in 2010, can you imagine we used to try to go to VCs in the UK, which there weren’t very many. And with that as the business plan, it was literally step one, solve intelligence, step two, use it to solve everything else. And people were quite confused. But we really meant it. And actually we may could go back to exactly using that mission statement because by solve intelligence, we meant build AGI. Ideally also, understand the nature of intelligence on the way to building AGI and perhaps using AGI to help us understand our own brains and minds better, things like the nature of consciousness, what is creativity, dreaming, all of these deep mysteries of the mind.

And one of the reasons I studied neuroscience was try to learn from what we understood about the brain as inspiration for algorithmic ideas. So step one was to try and build AGI and then we always had in mind what sort of happened, which is that of course it’s a general purpose technology, maybe the general purpose technology. And if it was built in the right way, so it was a learning system that was very general, what would be the limit of what it could be applied to? It could be applied to almost anything, was the dream, and I think that’s what’s borne out. I had specifically in mind for that step two, advancing science and medicine. So that’s what I meant by using it to solve everything else. I meant the big questions in science, all of them.

I was fascinated by all of them, the nature of time, the nature of reality, maybe that’s the most fundamental one. And I loved physics when I was at school. That was my favorite subject. And I think when you’re interested in the big questions, you end up doing physics probably. But the reason I decided that there were too many interesting big questions. So how was one to try and tackle all of that in a lifetime? And that meant, in my view, building new tools and to aid us are the best scientists, the best experts to make much faster progress in the fields that they were tackling and the big questions and important questions that they were tackling. And then of course, AI in itself is also a fascinating artifact in itself, scientific object one could call, worthy of study itself. It’s almost a new field.

To me, it would felt like the most fascinating and most important thing to spend one’s life on and I would have been doing it even if it hadn’t worked out, I would have found some way to be doing this in academia or wherever. This is what I always planned to spend my life working on. And all those things I did earlier were different expressions, gathering the experience and I suppose the knowledge to be able to attempt something like DeepMind in 2010 when we felt that we were ready to make fast progress. And of course the second part of that, use it to solve everything else is now much broader than just science and medicine, although that’s where I’ve tried to personally do my work in as well as running the overall organization. But obviously it’s going to be amazing for productivity and many other things in the world outside of just science and medicine.

Jonathan Levin: As you’ve been building these different models at DeepMind, you started with games and then you went into science. Were there particular moments where that… I mean, you started with a lot of conviction, but I’m curious if there were particular moments where you saw this is actually going to work with the AlphaGo and playing more with the — [inaudible].

Demis Hassabis: Yeah. There were many moments where I thought it wasn’t going to work, put it that way. Some of the ones I remember really well are, we started with games because they’re self-contained. They obviously were designed by other humans to be challenging or fun for other humans to play. I love games. They’re often microcosms of a lot of real world scenarios. If you think of Go or Poker or chess, I often thought as one of the think courses I’d have in an MBA or business school course would be a games module to study those types of games, diplomacy. They all have really interesting aspects of the best games of real life. And you can obviously practice many times in a safe scenario. That’s what I think games are really useful for. And that applies to AI systems that are learning too. They’re neat environments, they’re challenging and they have clear objective functions, which also was very important for our early days of reinforcement learning.

Almost no one had used reinforcement learning for any kind of scaled up problem. It was obviously an academic discipline, but it was used mostly for toy problems like little grid worlds. It wasn’t clear it could scale up to anything major. And so we started with, I would say, the most famous set of but most basic games that had become world popular, which were Atari games from the 1970s. And we started with the simplest game of all, which was Pong, just the bat and the ball, just two bats and a ball. And there’s an inbuilt AI system. It’s not really an AI system, an inbuilt system that controls your opponent and uses all the information that the game has about where the ball is and so on to move the bat around. And what we wanted to do was could you play Pong just from the pixels on the screen. So the raw data, the raw visual input and no other information, no privileged information about, no access to the insides of the program about where the ball is or the speed of it and so on, which obviously the program knows.

But we didn’t give the DQN system as it was called, our Atari system, any of that information. It just got the 20,000 pixels on the screen. And 20,000 pixels, I mean, it seems trivial now, but back in 2010, that was an enormous amount of input data. No one had ever dealt with something that complex and then multiplied by all the frames that you were doing. And for about, it felt like six months, maybe it was only two months, but we couldn’t win a single point at pong. So it was jerking the bat around. It was like, “Oh, is it ever going to even be able to control the bat?” And of course we had no notions of any of these things and it was just losing 21-nil to the inbuilt AI. And I did think, and we had a couple of different ways of trying to attack this. And we had almost no money, the runway, the couple of million dollars of funding that we had, which wouldn’t even cover an intern these days, which is good for all of you, students, was our entire funding.

And we were taking no salary and it was running out, the money, and I was like, “Oh, well maybe it turns out, maybe we are still 10 years too early. Maybe we’re 20 years too early.” And then magically it got a point and it was like, “Oh, maybe it was just luck.” And then it started winning a lot of points and then it started winning the games and then it was like, “Okay, we have liftoff now.” So now, and those of you working in machine learning will know this, if you get a foothold, you can usually heel-climb your way out of that. That’s been the history of AI, I would say, right? Once you have something working, there’s usually a way of optimizing it more. And that’s what turned out with Atari, so that was our first big result. And our first nature paper was really the first deep reinforcement learning model, certainly at scale.

Combining deep learning to learn the domain and deal with the perceptual inputs and the complexity of the input, find the patterns in it and then reinforcement learning built on top of that to kind of make the decisions and do the planning. And then of course that culminated in AlphaGo, which was always our aim. Dave Silver and I who was head of that project, we were undergrad friends as undergrads at Cambridge and we were just discussing it since our undergrad, about we were there in the mid ’90s, the Deep Blue-Kasparov match happened while we were at college. Of course, I was fascinated by both from the chess and the AI point of view, but I was more impressed with Kasparov’s brain than I was with Deep Blue because Kasparov with his incredible mind, still one of the biggest chess geniuses that there’s been of all time, he was able to basically compete on an equal footing with this supercomputer brute force machine next to him.

But of course he could do all the other things with his mind, speak five languages, do his politics, drive cars, all the rest of the things humans can do. And to me, that was incredible. That’s much more impressive. So there was something missing from the Deep Blue system and obviously those techniques, those expert system techniques where you hand curate the heuristics and then you use brute force search on top, which is still how a lot of traditional chess programs work today, that works for chess, but it’s never worked for Go because Go’s too esoteric a game, it hasn’t got material, every piece is worth the same. It’s all about patterns and intuition. Even the top Go players, that’s how they play it. So we thought, “Okay, if someone could actually get to world champion level at Go, it’s not just about…” Really that was an aside getting to that level.

It was more about the approach we would have taken would probably be a really interesting algorithmic approach and maybe and hopefully would generalize to other domains. And that’s what turned out with AlphaGo and then it went beyond our wildest dreams really because not only did it win the match against Lee Sedol in 2016, it also created famously new strategies that had never been seen before, even though we’ve played Go. Go is the oldest game humanity has invented, it’s 2000 plus years old, and been played professionally for hundreds of years, and we hadn’t discovered those strategies. So that was double whammy for me. I was waiting for that moment that AI was able to come up with something novel and there’s more levels of creativity than that, but beyond that, but at least it was a novel idea. And then that was for me was what I was waiting for to then start using AI for science. So the moment we got back from SOL, we started the AlphaFold project.

Jonathan Levin: So let’s talk about the science a bit, because then you went into the protein folding problem. And again, you picked a problem where there was data and where there was a clear objective function in terms of thinking about protein folding and it worked. I mean, you actually managed to solve this longstanding problem of predicting protein structure. You did something very interesting when you came up with AlphaFold, which was, it was obviously a huge science, Nobel worthy scientific breakthrough. Probably also of commercial value and you just gave it away for free. I’m curious, how did you come to that decision? Was it something, did you think about other ways of going about it? Why give it away?

Demis Hassabis: Yeah. So we picked the protein folding problem. I had my eye on that since also my undergrad days at Cambridge, just when I first came across it. I had a few biologist friends who were obsessed with the protein folding problem and actually they ended up becoming structural biologists, of course, in their career. And one specifically, I remember every time we were in the pub playing table football or something, he would be talking obsessively about how this was the most important problem in biology. And more importantly, I think of it as a root note problem, like if you could unlock that and find the structures of proteins, that would unlock whole new avenues of research, things like drug discovery, obviously we’re trying to push that, but also fundamental biology and disease understanding. So this was a problem worth really spending a lot of tension and time on because of the downstream effects that it would have.

It was a fascinating problem. It felt to me like the ultimate puzzle, 3D puzzle of how does this amino acid sequence, think of its genetic sequence, fold up into this 3D structure. It’s amazingly interesting, intricate thing. And the more I looked into proteins, the more incredible my respect and wonder is for biology. These unbelievable little bio nano machines, everything on life obviously depends on proteins and as you start looking at their structure, you start understanding their function. So this was fascinating to me as a science question. And then yes, there was the clear objective, like minimizing the free energy in the system. Presumably this is how physics is doing it. It’s why the body, these proteins fold in milliseconds in your body, billions of times a second. So somehow physics has solved this. There must be some topology, let’s say, that you could learn maybe with a deep learning system that would guide the search.

Just like we’d done with AlphaGo to find a great move in Go, a great strategy out of the more possibilities than there are atoms in the universe in Go. And protein folds are even larger search space than that, but there’s some way to narrow that down in a sensible way. You learn a kind of heuristic using the deep learning models to then guide your search for that to become tractable. And it felt like a really analogous problem in science to what we’d solved in Go, applying some of those same approaches, those same theories to this domain. And then the other thing was there was obviously 50 years worth of painstaking crystallography, structural biology work by many great labs and people. And after all of that effort, there was about 150,000 structures in the PDB, the main database, which isn’t a lot actually. Obviously it’s a huge amount of effort that’s gone into that, but there’s 200 million proteins, and 150,000 also for machine learning systems is a very small amount of data.

So most people thought it was at least 10, 20 years away before we would have enough data and the right types of algorithms to tackle that. But we felt that using every technique we knew in the end that we could make progress with that and it turned out to be the case. And then when we decided to, well, how would we make the maximum impact with this? It was obvious to me that we should fold all the proteins because not only was AlphaFold accurate, it was extremely fast. It could fold a protein in a matter of seconds. And then collaborate with, in the end European Bioinformatics Institute in Cambridge which hosts many of the biggest biology databases scientists use, and just host the entire 200 million protein structures on their database and just allow it to be as simple as a Google search to just find your protein structure.

Along with the confidence intervals the machine learning system had about which parts of the protein structure it was confident on, which is very important for biologists to know. So we put that all together and it was, of course, could have been very valuable. I don’t know how many billions of dollars or whatever. I mean, it depends on how you calculate it, to do that experimentally would be incalculable cost. But it would have been hugely valuable to keep proprietary, but for us it felt like we would only be able to scratch the surface of the downstream impact that putting all those structures out in the world could have on our own. Because there’s three million researchers around the world that use AlphaFold pretty much every day, almost every biologist, medical researcher in the world. There’s no way one organization could have done that. So it was obviously the right thing to do.

We also had depended on public data to train the first version of AlphaFold, so it only felt right to give back to that community, the structural biology community, this amazing resource that was amplifying the resource they had painstakingly built. And so it wasn’t even a question for me and it was great that also the executives at Google also loved science and totally got that. I don’t think all companies would have made that decision, so I give them a lot of kudos on that too. That was an easy discussion. And then we’ve tried to ourselves push that downstream with Isomorphic Labs, an Alphabet spinout that is building, you can think several more type of AlphaFold level breakthroughs, putting them together into a way that will accelerate hopefully drug discovery, take it down from years to months, maybe even one day, weeks, just like we did with protein structures, which used to take years for a single one and then we could do it in seconds.

Jennifer Aaker: We’ll be back with more of Demis and Jon after this.

Jonathan Levin: I want to turn for a minute to something you said earlier this week. You were in the news this week because at a big Google event you said that “we’re in the foothills of the singularity.”

Demis Hassabis: Yes, it got quite a lot of pickup, that line.

Jonathan Levin: It got a lot of pickup and I understand that maybe the Google press team might not have been so thrilled about it, but since you’re out there saying that, what did you mean by that?

Demis Hassabis: Yeah. So the full thing I said that closed the conference with was, when we look back at this time, I think that maybe I’m thinking 10 years from now, I think we’ll realize that we were standing in the foothills of the singularity now. What I mean by that and the reason I chose that word is that, so there’s the technology which is AGI. We’ve been calling AGI this next version of really general artificial intelligence. I believe that we’re only a few years away from that, maybe like 20, 30 plus or minus a year, which is astounding to think really. And then the era, I think that will be such an enormous transformative technology, it’s going to effectively be a new human era.

And that’s what I’m meaning by the singularity is that, and what many science fiction writers have written about that, is that it’s describing the era that we will be in and around when the advent of AGI happens. And I think we can feel this year, I would say, even though I’ve been working towards this for 30 years, I think this year with the way the agents are working and tool use, it started to become really useful for, you know, still early days of it, but genuinely useful in people’s workflows. And we can see what extra things are needed to be done and all of us, the leading labs are working on that. I think this is the beginnings of that, but the foothills, I still think there’s a lot more work and it’s just the beginnings.

And it’s not any one thing, it’s several different technologies, several use cases that I see, several things that I thought were maybe a bit further out, turned out to be now that are coming together that make me feel that in aggregate and to the extent that I wanted to say that. Because I think society needs to hear that because we don’t have long to prepare for what that means and it’s going to be enormously profound, I think. And the future in my view is still to be written, but these next few years are going to be very critical as to which way that will go and how we collectively want that to look like.

Jonathan Levin: If you look at surveys of how people perceive AI in this country in particular, it’s very negative right now, and maybe more negative here than even in other countries. And there’s probably lots of things driving that, concerns about privacy or state control or the size of the tech companies or jobs. I mean, you’re running one of the leading labs and how do you think about that public concern about the technology?

Demis Hassabis: I think the public’s right to be concerned. I think that there are things that, and I’m concerned about several aspects of what the technology is as a dual purpose technology. It’s something this profound. I sometimes describe it, quantify it as 10 times the impact the Industrial Revolution was, 10 times faster. And so taking into place over a decade instead of a century. So that’s like a hundred X of the Industrial Revolution and it’s probably an underestimate to be honest, but that’s probably enough for us to try and comprehend and deal with. And so of course I think there are super exciting. There’s going to be amazing things that are going to happen. We’re trying to do that with solving all disease. I think a lot of the other challenges facing society today from climate to energy to disease will be helped by AI, I’m sure of it.

And in fact, I’d be much more worried about those challenges if I didn’t think something like AI was coming down the line, but it’s going to cause a lot of change and disruptions and actually both on the technical side, economic and philosophical. And I think we’ve got to think through very thoughtfully and bring together all parts of society to discuss this, not just the technologists. The technology and the safety of the technology is just one piece of this. It needs economists, social scientists, human and humanity experts to kind of chart out what is going to happen next. And I think one of the reasons it’s negative here is that specifically, because it’s different in other countries.

For example, we came back from the summit in India, it’s hugely popular with the youth of India because they see the opportunities that it’s going to democratize for them having access to basically the same tools that you would have needed to go to Silicon Valley for. We’re in an amazing moment in that everyone can access pretty much what’s going on in the frontier labs, but only with a delay of just a few months. That’s unheard of really if you think about that. So these incredible things, but I think it’s also partly the way some of my peers are articulating that. I think they’re being very careful with their communication and understanding they’re being way too certain, I would say, with some of their pronouncements. Where I think actually there’s just huge uncertainty and that is worrying in itself, but it’s also means that nothing is decided, in my opinion.

I think it’s unknown and I think anyone who says that… I think directionally I could tell you some things, but I think a lot of it really depends on the actions taken in the next few years. And also what the youth of today, the many students in the room today, you’re going to be forming, you’re the first generation to grow up in AI native, shall we say, like I did, computer native. And just like every generation, you’ll master these technologies, become super productive with them. And I actually think over the next 10 years, at least, hard to predict beyond that, you’ll almost be super powered with those if you use them in the right way. The amount of creativity and projects you can do and the amount an individual will be able to do. But maybe that will change the nature of jobs. There’ll be more entrepreneurial, small entrepreneurial things, rather than big companies. I don’t know. It’s going to change a lot.

And I think part of that is for society to come together and really take this exponential seriously. Not just the technologists, economists and others need to take this seriously right now and start charting out what does that look like. If we’re in a post-scarcity world, for example, how does everyone benefit from that? It’s obviously not correct for just a few people or a few companies or even a few companies or even a few nations to be benefiting from this technology. It needs to be broad. It’s going to affect all of humanity. Has to broadly accrue the benefits to everyone, but how’s that going to be done? A lot of us have been talking about this for a while, but we really need answers now and concrete things and actions to be taken. And I plan to do my bit on that.

I’ve been thinking a lot about this over the years and planning and building influence on this and I will do what I can. Obviously we’re an important actor, but we’re only one actor in this space. The good news is, I know that all of the leading labs and the leaders of those labs, although they disagree on a lot of things, they do worry about these sorts of issues coming down the line. But we need more forums to allow them for us to come together to discuss these things more candidly. I think that’s what probably the public is detecting, is this slightly skewed discussions about what’s going to happen and maybe there’s some ulterior motives behind some of that messaging, raising money, other things. But I think we need to use the scientific method, be really rigorous and thoughtful about this critical moment in history.

And then maybe the final thing I would say is, I would love to see, I think it’s incumbent on the industry and the field to show more unequivocally what the benefits are. And not just talk about them, but demonstrate them. So in health, in medicine, in science, these things are all, in my view, are unequivocal goods like AlphaFold. But there aren’t enough examples. There should be 20 AlphaFolds and there should be… We got to stop talking in the hypothetical about curing cancer and actually cure cancer. And so these are the things that I think are going to be needed to demonstrate to the public why are those of us who are excited about it and many of us are in this room. Why are we excited about this? Why have we spent our whole life building towards this? And also, how are we going to concretely mitigate the risks while enabling all of the amazing things that we would like to see and I think society needs.

Jonathan Levin: I think a lot of great points there that if there were some tangible benefits that were realized because of AI breakthroughs, say to human health or drug discovery, that might change people’s perception in some ways. And I love the suggestion of trying to think farther out about a world that might look very different in terms of productivity and so forth. It’s hard actually to do that. Rarely in social science can people get out of the current frame they’re in and actually project way forward. I think of Keynes’ great article during the Depression when he looks out the economic lives of our grandchildren. It’s a rare case that you were saying we need another Keynes right now.

Demis Hassabis: Maybe it’s someone in the audience.

Jonathan Levin: Maybe there’s someone in the audience who will do that. Let me ask, one of the things you’ve talked about for a lot of years is the need for the frontier labs to, in a sense, regulate themselves. That is to sometimes not release certain kinds of technologies that might be threats to safety and so forth. Right now it’s pretty clear that the labs are just at a breakneck competition. They’re investing everything. They’re going all out. Do you still feel the labs ought to be self-regulating? Do you think the government ought to step in and regulate AI in some way? How do you see the current dynamic relative to the way you talked about it in the past?

Demis Hassabis: Yeah. Well, look, first of all, just to give some historical context to this, this is not what… In terms of we talked earlier about how the technology’s gone, I think the technology has gone amazingly and maybe even on the better side of what I imagined 20 years ago. But the environment it’s been birthing in is not the ideal. Far from it. I was very worried about 15 years ago, 10 years ago about this race dynamic happening as more and more people, more and more companies, more and more ambitious tech leaders realized what I had known for 20 plus years of how important this technology was going to be. And we talked about some of this in the room about the dangers of this kind of race dynamic. And unfortunately we’ve ended up, because of the way the technology’s gone. So if I could have waved a magic wand, what I would have done was build AGI, the general technology more in a research facility, perhaps like a CERN.

Maybe all the best minds helping critique each other’s ideas and making sure we were rigorous with the scientific method and the testing of it and understanding each step that we took. But then we wouldn’t have to wait for that. Of course, that means AGI would arrive later, maybe 10 years later, but we wouldn’t need to wait for that to get the societal benefits of it because at the same time we would break off bits of that and use it for specialized systems like more AlphaFolds, curing diseases. That can be done because those are, AlphaFold is a specialized hybrid system, uses a lot of the ideas that general purpose systems use, but it’s specialized to protein folding. And you can see that was my vision for it because that’s what we were doing. But then chatbots changed that because effectively, and that was probably the only surprise to me over the last 15 years on the science side is how effective transformers ended up being for language.

And the fact that you could separate language and learn it just from the internet without having to act in the world, either robotics or simulations. It’s kind of very interesting and that would be a whole nother topic why that was. And I have some theories on that. Language is more grounded than linguists probably thought. There’s some grounding coming from the reinforcement learning feedback that the human testers are doing because obviously we’re grounded in the real world. So when we say yes, no, to certain things, our grounding is then ending up in a very low bandwidth way, but still ending up modifying what the foundation model understands. So there was these unexpected things, I would say, that happened. And then that made it a very important commercial technology that could be scaled with engineering and money and so on, which is what you see today. And that changed the dynamic.

And then has created what we see today, which is probably the most ferocious competitive environment, I would say there’s ever been. I mean, certainly in the tech industry, tech era, maybe ever. Maybe other historians here from the business school will tell me otherwise, but it feels unbelievably intense being in the middle of it and it feels like that for all of the participants. And then on top of that, you layer the geopolitical complexities. So there’s a double race going on. There’s the race between the companies and it’s pretty life or death for them. And then there’s the U.S.-China dynamic and others, the geopolitical dynamic, and there’s a race there. So it’s a double layered one, very tricky. Now I still have hope that there can be some cooperation and coordination between… We certainly discuss this as lab leaders on the safety elements and the security elements. Everybody wants that.

Nobody wants something catastrophic to go wrong. The problem is, we’re in a prisoners’ dilemma where anyone who… By definition, if you take more time to release something or make something safer, that’s harder than just putting it out there and letting it see what happens. So a defector has some advantage and this is the classic problem with the race to the bottom dynamic. And we’ve got to change that somehow and I think urgently. And I think part of that is some form of government involvement. The hard part there, of course, is that anything to do with regulation, it’s too slow. Every week there’s something new. If we were to regulate something two years ago, it’s like ancient history now. So it would almost be almost certainly the wrong thing. So whatever is designed, and I have some ideas in this and I’ll probably be talking about this later this year is, it needs to be dynamic, which usually that word doesn’t go with regulation.

So it’s got to be light, fleet-footed and able to be informed by the latest developments so that it can adapt to where the actual risk is, rather than some kind of perceived risk that turns out not to be the case or not the critical thing many years before. It’s just not going to work for AI. And even today, the leading scientists wouldn’t necessarily agree on a short list. In fact, I know they definitely wouldn’t agree on a short list of what checks and balances are needed. And that’s because the science isn’t settled and that’s partly the speed, but also the pace of the progress is running ahead of the understanding of it. That’s just how it is as part of the race dynamic. But we need to somehow rebalance that. And I think some form of really almost a smart regulation is required that is dynamic and can adapt with the times very quickly, and probably informed by the leading labs because they’re seeing what’s actually at the coalface.

Jonathan Levin: I think there’s so much more to discuss there in terms of the prospects of how you set up a regulatory system for AI and do it in a way that didn’t prevent some of the breakthroughs, positive breaks you’re talking about to the geopolitics.

Demis Hassabis: Exactly. We want all that innovation, right? We want to solve the disease. So exactly, how do you enable the good use cases and mitigate the bad?

Jonathan Levin: I’m looking forward to when you bring out your plan for that this year. I think that’ll be fantastic and that’ll give us a lot to talk about here on campus and everywhere.

Demis Hassabis: Yes, next time.

Jonathan Levin: We’ve got a couple of student questions. I want to give a chance for some student questions.

Student: Hi, Demis. I’m Arinze, second year at the business school. My question is, how do you balance pushing the frontier of AI with ensuring that the health and scientific dividends is evenly distributed in places like Africa and the Global South where the need is the greatest, but the infrastructure for deployment and research is most limited.

Demis Hassabis: Yeah. We think about that a lot actually and that goes back to one of some examples of that I can give is back to the AlphaFold question where we folded all the proteins, we put that out on databases you could access from anywhere around the world. So these three million researchers come from 190 countries, just to be clear. It’s pretty much every country, every researcher. And what was great, what we did actually in the early days of seeding some collaborations, what you could do with AlphaFold, we worked with the DNDi, Drugs for Neglected Disease, part of the WHO in Switzerland, which work on diseases in the poorer places of the world that don’t have good healthcare systems. And some of those diseases are neglected, as you know, because big pharma can’t make money in those markets. So then the diseases that affect primarily those areas of the regions of the world don’t get as much research resources behind them.

So what we were able to do in collaboration with this institute and many actual universities on the ground is jump them straight to not needing to try to figure out malaria virus or the structures of Zika virus or something like that, which they would’ve had to do all the painstaking structural biology. They can just start that as a given and work straight away on the drugs. So that allows to speed up massively the whole process. They can take the structure of interest and move forwards from there. Same with crop resilience affected by climate change. We work with Jennifer Doudna’s Institute and many others on these things because lots of plant proteins, we didn’t know what the structures were of them because obviously most of the structural work has gone into human proteins. So if it’s animals or plants, there’s a lot less data out there. So it’s even more differentially impactful in those types of areas.

And then the final thing I would say is, and I think this is where the capitalist engine can actually work for good here is, if we can make the drug discovery platform that we’re working on at Isomorphic as efficient as I’m talking about, down from years to months. So instead of it costing billions of dollars, it costs tens of millions of dollars, maybe single millions of dollars. Then suddenly what I’m hoping we’ll be able to do with Isomorphic is, we cure these terrible diseases that maybe affect the richer parts of the world, that makes money and that fuels the engine, but then we can do philanthropically, the company could find cures to diseases where we don’t need to make any return because it’s fast enough and it’s cheap enough that it can just be done and in a short amount of time. So I think that’s my dream for how I can make Isomorphic help the whole world.

Student: Hi, Demis. Thank you so much for taking the time to talk to us. My name is Miki, I’m a senior in the Doerr School of Sustainability. And you’ve described extensively how AGI could be humanity’s most transformative technologies and I’m just curious the responsibility or how you think about the societal impacts alongside this intellectual pioneering and productivity that AGI presents, particularly when thinking about how is this going to redefine and reshape people, the challenges that we’re trying to solve today, but the downstream effects that that could bring forward. Thank you.

Demis Hassabis: Yeah, thanks for your question. I think about this all the time and have done from the beginning because we were planning for success. Even though it seemed very improbable back 15, 20 years ago. And I think this is why I like doing talks like this and meeting folks in these kinds of places is it is a bit of a call to arms now. It’s very urgent that we really think about the second order consequences. And I think many of you in the room and many of you in the humanity subjects, it now is your time in my opinion because, okay, we got to get the technology right. But then if we do that, then there’s the economics question. And if we get that right, there’s the philosophical questions about the human condition. And I’m very excited about, and I’m very optimistic. Obviously, I’m a cautious optimist is the way I would say it is.

I’m very optimistic that we’re going to get this right and I’m a big believer in human ingenuity, especially when the pressure’s on. I think humanity’s always figured it out when the chips are down and they are now, but we do really need to start taking that… I think the technologists are taking it seriously, but the other parts of society need to as well. Economists, I’m always a little bit astounded when I talk to economists about what’s happening and they’re pretty skeptical. Where’s it coming in the GDP? And it’s like, look, it’s 10 times the Industrial Revolution. Can we start planning for that now? And we’ll be in a world where we do need some giants of these fields, like Keyne’s was, for now. Why would that hold in a post scarcity world? We’re going to be in a world for the first time if we get the technology right, where we’re a non-zero sum world for the first time in humanity’s existence.

How can that not need a new type of economic system? It has to. And I don’t think it’s any of the ones we’ve tried because they were all done under the guise of zero-sum and a limited, a scarce world. And I’m talking about traveling to the stars and utilizing all the resources that are out in the solar system, not just the limited ones on earth. And I think that really is going to happen if we get the technology right in the next 10, 20, 30 years. And then after all of that, there’s the even harder question of how do we want to evolve our society and what is virtuous, what is meaning, what is purpose? And I think that’s going to need lots of great philosophers. So that would be my appeal to people in those fields is now, in my view, could not be more of an exciting time if you’re working on those types of projects, as long as you understand and really viscerally understand and lean into what’s actually happening here.

Jonathan Levin: It’s a good charge to university. Okay, but one more student question.

Student: Hi, Dennis. I’m Janai. I’m a second year MBA student. My question to you is, what do you not want AI to touch in this lifetime and what do you hold sacred from your perspective? Thanks.

Demis Hassabis: Yeah, that’s a great question. Look, AI is going to be, in terms of the scientific world of things, it’s a fully general technology. You can think of it as a Turing machine. That’s the way I think about it. It was my favorite course at college and I think our minds are actually fully general. So we’re approximate Turing machine. So as Turing showed, anything that’s computable, a Turing machine can compute. And most things we know about in the universe, non-quantum things are computable. So that’s a pretty large set of general things that we can turn our minds to. Hence, we built modern civilization, which is miraculous if we stop to think about it and I don’t think we wonder enough. We don’t keep our sense of wonder for long enough about that.

But it also means these systems that we’re building, they’re also going to be Turing powerful as well. One thing I would say is that there are very big questions to come that I think it would be better if we took more time over. So one example that’s pretty topical right now is consciousness. And it’s not a very well posed problem from philosophy and neuroscience still, although I think we all have intuitions as to what are the important aspects of that. My feeling is the current systems don’t exhibit any, are not, but others disagree. What I would recommend though in terms of like what area should AI not touch is that we build our first systems as tools, intelligent tools. That’s enough of a challenge already, in my opinion, because that’s already AGI.

And then using those tools, I think we should study neuroscience and other things like that and philosophy and actually come up with a more rigorous definition of things like consciousness. I think that is possible. And then test things against that and then maybe society decide if we want to cross the second Rubicon of trying to make entities that at least seem like conscious to us. So we may not want to make that decision. I think that intelligence and consciousness are dissociable. I don’t think you have to do that to have an intelligent system. I think it’s a choice. And so you can probably feel that when you use some of the leading chatbots, there’s differences in opinion that come through. And my view is it’d be better to take that as two steps. They’re both enormous for humanity, rather than conflate the two.

Jonathan Levin: Demis, we’ve got an auditorium with many students in it. If you were back in school, what would you be studying? How would you be thinking about what… What would be your advice on how they should be thinking about to study in their careers?

Demis Hassabis: Well, look, I would be really excited if I was back at college now. My recommendation would be those of you doing science and STEM subjects and mathematics and computer science, still do those things. I think you’ll be able to take better advantage of these tools if you understand how they are put together and what they’re capable of. I think that’s going to be true for the next period, the next 10 years at least. And I would also lean in though to not wish it away. The genie’s not going back in the bottle. Lean into what these tools can do. I can tell you that the leading labs are so busy making the tools that we probably only scratch the surface of, not even probably, of what they can actually do. Even today’s tools. I have this, sometimes people call it capability overhang.

There’s so much potential these things can do if you figure out how to pair them with other things or pair it with another domain you’re an expert in, build it into your workflow in an interesting way. You have those tools. They’re the most powerful tools anyone’s got. You have them in the palm of your hand. There’s so much more you can do as an individual. I think it should unleash creativity, like those of you studying humanities or product or business, maybe you didn’t have coding skills before, but you can produce a lot of what’s in your mind now using these tools. But I think also the coders, the people who are expert at that could do a hundred X more in the terms of the size of a project you can do if you’re a expert at coding. So I think it enables both, the democratization and the people that are specialized in those areas. So I think it’s an amazing time, but I get it’s also worrying because everything’s going to change.

So that’s the only thing I can tell you for sure. Everything is going to change in the next 10 years, probably more than people assume. But that also, in any time there’s enormous change like that, there’s enormous opportunities. There has to be. And the world’s your oyster really. And I envy some of you now because you’re the first generation that will be AI native, just like my generation was computer and internet native. And it’s going to be in your hands in the end, the students in the room, how that future world gets built. And I think it’s a very exciting time if you think about it in the right way from the right angle and with a lot of imagination and creativity. But I think that’s always been true and maybe it’s more so now in periods of enormous change like this, that accentuates it.

Jonathan Levin: We were saying in a period of a lot of change where you don’t quite know what the future holds, but you have to be able to be adaptable and have broad domain of knowledge. It’s going to be a golden era for liberal education.

Demis Hassabis: I mean, the main thing is to just make sure you double down your own agency. The future’s still to be written, I would say. So don’t listen to anyone who says it’s not.

Jonathan Levin: Demis, thank you for joining us. This was amazing.

Demis Hassabis: Sure.

Jennifer Aaker: This podcast is a production of Stanford Graduate School of Business. Thanks so much for listening.