Demis Hassabis Three Quarters Of The Way To Agi

Dennis, thank you so much.

Exciting to be here. Thanks everyone for coming. It’s great to be here. We’re so honored to have you at our chocolate factory. Yes, I just heard about that. Yeah, looking forward to the chocolate afterwards. Excellent. Well, Dennis, we’re going to jump right in. We have one of the OGs in every way. Uh original thinkers, founders, visionaries in all things AI. Uh true believer, true scientist in in Dennis. uh we’re going to spend the beginning of the conversation about the early days, then early days of Deep Mind, and then we’ll get into the science and open up the room for some questions. So, let’s jump right in. So, Demis, you were a chess prodigy. Uh you also were a founder of a gaming company, you’re a neuroscientist, and you’re the founder of Deep Mind and now leader of a really big consequential company. Those seem like pretty unrelated pieces, but you’ve said that there’s a common thread. Can you walk us through that? There is a common thread and um uh maybe I made it into a common thread so it could be a post hawk you know sort of um shaping but uh I wanted to do AI for a long time. So I kind of decided it was the most important thing I could poss and most interesting thing I could do in my teenage years. Uh and then I picked things to study or do that I felt eventually would help me um build a company like DeepMind. So I I had that uh uh as a plan from about 15 16 years old. I had a detour into games because actually that was in the 90s. This is um that’s where all the most cutting edge technology was being done. Uh obviously not just in AI but in graphics especially including hardware of course the GPUs we all use today. They were designed for graphics engines and we used the I mean I was using the very first GPUs back in the day in the late 90s. Um, so there was a lot of really cutting edge technology and then all the games I made, uh, including all the games I did for Bullfrog, but also my own company, Elixir Studios, uh, all involved AI as the main gameplay component. So probably the most, you know, well-known game I made was theme park when I was about 17 and that was a simulation of an amusement park and thousands of little people came into your theme park and played on your rides and decided what to buy from the shops. So there was a whole kind of economics AI model underneath and it was one of the first games of its type along with Sim City. And when I saw it sold, you know, 10 million plus copies and when I saw how delighted people were to interact with the AI, you know, that was one of the things that made me think about spending my whole career on it. Um and then of course neuroscience um is to get inspiration from how the brain works and um and and different algorithmic ideas from that and then just bringing all those different things together for the start of Deep Mind when the timing we felt was right. And then of course we use games as the early proving ground for our AI ideas. So we’ve got a room full of founders here and you can relate because you’re a founder not just once but twice. Take us back to the first time Elixir Studios. What was that like? It’s not the startup that you were most known for, but it was one that you had incredible success with. How did you lead that? And what did it teach you about building? Well, look, we we we start I started Elixir Studio straight out out of college. Um, and uh I was lucky enough to work at Bullfrog Productions, which those of you know games, it was a kind of legendary game studio uh in the early days of the game industry, probably the best one in the UK, in in Europe. And uh I wanted to do something that combined uh pushed AI. So effectively I was funding AI back in those days through the backdoor through games development. Um and then push the the the forefront of that and combine it with cutting edge creativity. And I think that’s still relevant today with the way we do our blue sky research. But maybe the biggest lesson I learned was you want to be um 5 years ahead of your time, not 50 years ahead. So we tried to do a game called Republic at Elixir Studios which simulated a whole country and then the idea of the game is you could sort of um overthrow I think there was a dictator in charge of the country in any number of different ways and we basically simulated living breathing cities and this is bearing in mind this is like the late 90s on a Pentium. So we had to get all the graphics and all the AI for a million people um working on uh you know a home PC at the time. So, it was a little bit ambitious and um uh and maybe it was too ambitious and it caused it caused some some issues. And I took that lesson with me of like you want to be ahead of your time. You don’t want to be obviously when it’s obvious to everyone it’s too late. But um if you’re 50 years ahead then there’s probably no way you can get it to be successful. All right. So, speaking of not being too far ahead of your time, it was 2009 and you decided there would be AGI. Mhm. Yes. Maybe it was only, you know, 10 years ahead of of our time in that time. Better than 50 years. So, so tell us about again room full of founders here. Tell us about 09. How did you convince the first few brilliant talents because you pulled in really high caliber employees, early team members? How do you convince them to believe in what seemed like total sci-fi at the time? Well, there were some interesting threads that I think we picked up on. I I I think we thought we were five years ahead, but maybe we were more like 10. But it was, you know, deep learning had just been invented by Jeff Hinton and colleagues sort of uh in academia, but almost no one had really realized it was a big deal. Uh we we knew a lot about reinforcement learning and we felt there was huge um progress to be made by combining those two techniques which almost had never been mixed together really certainly not any kind of re anything other than toy problems uh in the academic subjects. There were quite two quite siloed parts of AI. Um then we could see the compute uh the GPUs at the time were going to be really useful. Of course we use TPUs now. Um but the accelerated computing industry was was was going to be very helpful. Uh and then we also felt at the end of my PhD and postto and some of the other people I got together were computational neuroscientists that we had enough ideas and principles um from the brain that could be useful including the idea that reinforcement learning uh could eventually scale uh to AGI. So um so we felt we had these ingredients and we almost felt like we were keepers of a secret because no one uh either in academia or industry really believed that any big progress was possible. In fact a lot of the people in academia uh used to roll out you know literally roll their eyes up at us when we were sort of suggested we would work on AGI or strong AI it was sometimes called at the time um because it was like well we know this doesn’t work. So, you know, everyone tried it in the ‘9s. You know, I did my posttock at MIT, which was the which was the sort of center uh uh uh point for expert systems uh and first all the logic language systems. I mean, it seems amazing to think that now, but I was already feeling like that was araic then, but they they you know, that’s still how it was done both in Cambridge in the UK and also in MIT, these big centers of traditional AI. Um, and it felt like, but actually that convinced me even more that we were on to something because at least if we were going to fail, we would fail in a different way than people had failed, you know, to get to AGI in the ’90s. So, that felt like it was worth doing uh no matter what, even if obviously it was researched, we didn’t know for sure it would be successful, but at least we would we would fail in an original way if it didn’t work. Was there any common sticking point in that early belief? Was there something that you had to prove either to yourself or to your early followers to get them on board? Well, it was I mean we had put it this way. I would have been spent my life on AI um no matter what had happened. So, as it’s turned out, it’s gone, you know, it’s sort of gone on the absolutely amazing side of the optimistic side of what we thought. Still actually within what we were predicting in 2010. We thought it would be a 20 year mission. Uh and I think we’re basically exactly on track as a field for that. And obviously we played our part in that. But um even if it hadn’t transpired that way and it was still now niche subject that’s all that’s what I would still be doing because I felt it was the most uh important technology ever if it was obvious to me. You know our original state uh mission statement at deep mind was step one solve solve intelligence i.e. build AGI. Step two use it to solve everything else. So it was always I always thought it was um the most important technology that could ever be invented but also the most interesting one. So as a tool for science, as an interesting artifact in itself and actually as one of the best ways to understand our own minds, you know, like the nature of consciousness, dreaming, creativity, all of these questions I had as a neuroscientist. I felt one of the things that was missing was an analysis tool like AI, but also a comparison for that you could do sort of a controlled experiment study and compare uh uh uh two different systems against each other. Let’s talk about AI for science. You’ve been early to that. you’ve been a believer and you’ve been really a purist about this where this is the driving mission. What about the way you set up Deep Mind and set the culture has positioned it to be on the constant forefront of AI for science? Well, that was the uh the ultimate goal at least for me my personal passion there’s one there’s my own drive to build AI which was to advance science and medicine and and our understanding of the world. It’s my expression of that mission was to sort of do it in a meta way, right? build the build the the ultimate tool and then come back when that was ready and use it to make breakthroughs in science things like Alpha Fold that we’ve done and I think many more things so we’ve always had that at the heart of what we’ve been trying to do at deep mind so actually we’ve had an AI for science group division uh uh led by push me kohi that um has existed for nearly a decade now actually pretty much the day after we got back from soul and the alpha go match which is sort of 10 years to the month now uh is when we started formally started the AI for science efforts because I was waiting for the algorithms to be powerful enough and the ideas to be general enough um and for me you know cracking go was that point that time uh that we thought okay now we’re ready to really apply these ideas to important real world problems starting with these big scientific challenges so we’ve always had that in mind as the the most beneficial use of of AI like what could be better than using it to you know cure diseases and uh give us healthier lifespans and and to help with medicine followed obviously by other really important areas like material science and the environment and energy uh and these kinds of topics which I think AI is also going to be have a huge part to play in the next few years and how does AI break through in biology you’re deeply involved with isomorphic this is an area of deep passion you have been a purist on the potential of AI to cure diseases from the very beginning when do we have the type of moment that we’ve had in language and coding, but in biology. Yeah. Well, I I mean, I’d argue we’ve already had one of those moments with AlphaFold. So, um you know, it’s a 50-year grand challenge. Protein folding and the 3D structure of proteins is incredibly important thing to know about if you want to design medicines or if you want to understand biology. Of course, it’s only one part of the drug discovery process. It’s an important part, but it’s only one part. So Isomorphic Labs which is our um you know latest spin out uh having a lot of fun running that as well is is to build adjacent technologies in more biochemistry and chemistry space that can actually design the compounds uh automatically to kind of fit and bind to the right part of the protein. So we now know the protein the shape of the protein. We know that what’s on the on the surface of the protein and what we have to target. Um um but now we got to build the right compound that of course binds strongly to where you want it to bind on the target of interest but doesn’t bind to anything else ideally because that would be a toxic side effect. So um the dream is to do almost all the exploration uh which is 99% of the of the work and the time in silicone and then save the the the the wet lab step just for the validation step. Right? So that would be um you know I think if we can do that and and I think we can get there in the next few years um I think we could reduce drug discovery times instead of down for taking like you know an average of 10 years down to months maybe even weeks uh and perhaps even days one day and um and then I think then all disease could be in reach and I think things like personalized medicine will become possible uh you know like personalized variations off of base medicines. So uh I think the whole of uh the whole medical area drug discovery are is going to be revolutionized in the next in next few years. Brilliant. You’ve talked a lot about AI for science. Do you think that at some point AI will create new sciences? Allah industrial revolution and thermodynamics. Will there be something new taught fundamentally in our education system? And if so, what would it be like? Well, I think there’s several things uh that along those lines that I think we is going to happen. So, first of all, um the understanding and the uh analysis of AI systems themselves I think is going to become a whole science, a kind of engineering science. These are incredible incredibly interesting artifacts that we are building and um they’re incredibly complex as well as complex eventually they’ll be as complex as the human mind and the brain and so they’ll need to be studied so we can understand fully way beyond where we we are today how these systems work. So I think there’s a whole kind of field mech is part of that but there’s a lot more I think that we can do to analyze uh these systems. So that will be a science, but I think also um AI itself will maybe unlock new sciences, which is maybe what you’re getting at. Um the one I’m particularly excited about is AI for simulations. So I I love simulations. All my all the games I wrote not only had AI, but they were simulations. And I think simulations is the way we can address some of the um what we maybe think of social sciences uh like economics um and and other more humanistic subjects because um it’s very difficult to do control studies in that you know why aren’t they just sciences like physics today because the problem is they’re emergent systems um just like biology actually and it’s very hard to do repeated control experiments. You know if you raise interest rates by half a percent you have to do it in the real world and then see what happens. You can have theories but you can’t run it thousands of times. But if you could simulate things uh really accurately then maybe there’s sort of new sciences to be done where you can sort of uh rigorously sample uh from a very accurate simulator. Um and then I think that will turn that allow us to make much better decisions in these today what are very uncertain uh domains. What will it take to get to those extremely accurate simulations? world models, what kind of science is necessary and engineering together? Yeah. Well, look, I I mean I’m thinking a lot about that in we do a ton of that work a like learning simulators basically would it be so you know these are in domains where you can’t we don’t know the mathematics of it well enough or it’s perhaps too complex. We can’t just write we can’t just write a a directly down a special case simulator. Um it’s just not accurate enough doesn’t capture all the variables. Um we’re doing that we’ve done it with weather. Um uh we have the most accurate kind of weather simulator in the world uh weather next and it’s far faster than what the meteorologist use the weather yet. No, we can’t we can’t no and I’m not sure that would be a good idea but the f the first step is to understand it better. Um and but then even biology you know we’re we’re working on a kind of what I call a virtual cell. So you know hugely um dynamical emergent system and I think biology is is perfect uh uh sort of machine learning is perfect description language for biology in the same way maths is for physics because I think in biology and in lots of these natural systems you have loads of weak signals weak correlations tons of data far too much that any the human mind can analyze but there are connections and correlations and interesting causalities within that mass of data. So I think it’s sort of it’s always struck me that machine learning is the perfect tool uh uh to describe those kinds of systems where until today you know mathematics hasn’t been able to do that uh either because we can’t manage it as as top mathematicians because too complex or the expressive power of maths is not enough for uh uh to to understand these sort of highly emergent dynamical systems. Is it also because of the messiness and stochastic nature? Sure. And and I mean eventually you could by the way once you learn these simulators it maybe there’s this is maybe there’s another branch of new branch of science you could maybe extract some equations from the once you have the simulator so you have this sort of implicit simulator intuitive simulator and then maybe you could extract explicit equations from that um because you partly because you could also sample it as many times as you want as fundamental as Maxwells or something maybe if if I don’t know if that exists for such emergent systems. Um, but if they do exist, I don’t see why we won’t be able to find them in this with this with this methods. That would be amazing. You’ve talked about this theory that the basic building block of of everything in the universe could be information like this is more theoretical. How do you think about it that and what does that mean for a traditional classical touring computer? Well, look, I think you can of course all the famous, you know, equals M MC² and all the stuff Einstein did and energy and matter are kind of equivalent. Um but I actually think information has a kind of equivalency in the same way. So you can think of you know the organization of matter and structure and especially things like biology that are resisting entropy as um basically information processing systems at their heart. Um so I think one can convert all of those three kind of quantities into each other. But I have this feeling information is most fundamental. So it’s a little bit the opposite way round to the classic physicist thought in the 1920s and things where you know it’s sort of energy and matter primary. Um I actually think it’s a better to a better way to understand the world the universe is to think about it as information first. And if that’s true, and I think there’s quite a lot of evidence for that, then of course AI is even more uh sort of profound in a sense totally than uh we think and it’s already pretty profound because it’s it’s also about organizing information and understanding information uh and constructing uh uh informationational objects. So um AI in my opinion is all about uh information processing. Um, so I think there’s something sort of very deeply connected with with uh these different areas if you think of it through the lens of information processing as the primary way to think about it. And do you think a classical touring machine will be able to compute everything? Well, I I sometimes think of, you know, I sometimes sort of think about what we’re doing and refer to ourselves as Cheuring’s champion because cheering machines I think Alan Cheuring is one of my all-time, you know, favorite scientific heroes. Uh I think what he did obviously laid the foundations for the comput for computer science but also AI. Um and I think it’s one of the most profound results ever is the cheuring machine result. You know everything that is computable can be computed by a relatively simple description of a machine. Um so I think our brains are likely to be approximate cheuring machines. Um, and I think it’s interesting to think about the connection between chewing machines and quantum computers and quantum systems, but I think at least what we’ve shown with things like Alph Go and especially alpha fold uh is that a a classical cheuring machine obviously in the guise of a modern neural network uh it can model what was thought to be in the case of protein folding it’s a quantum system you know in some sense it’s very you know it’s dealing with very small uh uh particles Um, and one might think you’d have to take into account all the quantum effects uh of the water bonds and all sorts of things, but it turns out you can get to approximate um optimal sort of solution on a classical system. So it may turn out that a lot of things that we think are qu that would need a quantum system to model or run uh might be modelable on a classical system if thought about in the right way. So you’ve talked about AI consistently as a tool. like like a telescope or a microscope u astrolab through the centuries but when you think about a machine that can model almost anything let’s say it can’t even model quantum systems like you pointed out when does it stop becoming a tool and will that ever happen well my my my feeling strong feeling is we should in this sort of mission to and journey to build AGI u those of us on that journey many of people in this room you know I feel like it would be best to build a tool first, an incredibly intelligent and useful and precise tool and then cross the the next sort of rubric that’s already profound enough and um has you know of course the tool could start becoming more and more autonomous and agentike that we’re all uh seeing. We’re in the midst of that the agent era now. Um but then there’s a further step of like you know does it have agency? Is it conscious? These sorts of questions which are also going to be questions we’re going to need to address. But I would um I would recommend we do that as a second step perhaps using the tool in the first step to help us uh uh with those next profound questions. And ideally also we could understand uh our own brain and and minds better and define things like consciousness a lot more precisely than we can today. Do you have estimations of what that definition of consciousness might look like? No, I mean I haven’t got much to add beyond that thousands of years of philosophy hasn’t said already. But I mean it’s very clear to me that um it’s obvious some components are going to be needed. They’re probably necessary but not sufficient. things like self-awareness uh and you know the idea of self and other um some kind of continuity over time. So some of these things are clearly needed uh for anything that might look like consciousness but um I mean obviously it’s an open question as to as to uh uh what the full definition is and I’ve talked to many of the great philosophers about that. Daniel Dennett uh obviously sadly passed away recently but we had a long conversation a few years back about this and I think um you know one of the issues is how does a system behave does it behave like a conscious system so that’s uh uh you know you could argue some of the AI systems might end up being able to do that as they get close to AGI um but then there’s still the question of why do you you know why do we think each other are conscious one is the way we’re behaving we’re behaving like conscious beings but the other thing is we’re we’re running on the same substrate So I think if those things are true then it’s parimonious to imagine you’re experiencing the same thing I’m experiencing which is why we don’t have that debate about you know normally about as are each other conscious but I think we’ll obviously we’ll never have the substrate equivalence with an artificial system um so there’ll always be a I think it’ll be hard to to completely close that gap so you can look at it behaviorally but but what about experentially um there are probably some ways to do that post AGI but it’s a bit out of scope today even for AI for science discussion. Yeah. Brilliant. Uh so we’re going to open the room to questions in just a moment. Get your questions ready. Uh but you brought up philosophers. You’ve mentioned Kant and Spinosa as two as your two of your favorite philosophers. Kant is this you know deonttological highly duty-driven uh philosopher. Spinosa almost has this deterministic view of the universe. How do you kind of connect those two beliefs and where where is your thinking of how the world works? Well, the reason I like those two, they they stuck out for me is that I think Kant when when I was doing my PhD in neuroscience, you know, his his sort of statements about the mind creates reality, right? I think that’s basically true. And um and so another reason to study the mind, right? And how the brain works. Um and I’m interested ultimately in the nature of reality. So we have to understand how the mind is interpreting that. Um and so I think that’s for me what I took from Kant and then Spinosa it’s more about the um you could almost call spiritual dimension of like well if you’re trying to understand the universe using science in my case as the tool you’re sort of understanding some deep mystery about how the universe works right in a in a really uh uh uh kind of deep way. And um that’s what I feel I’m we’re doing and I’m doing when you know I do my science and we work on AI and we’re building these tools is somehow we’re kind of uh reading the language of of the universe. Beautiful. Uh what a beautiful way to say what you do every day. Deis scientist or philosopher. Um we will before we wrap do a couple of rapid fire questions. Okay. Thank you for uh he’s not seen these yet. Sure. Um, o over under on distribution, year of AGI. Oh, wow. Um, or reject premise of question. Uh, no. 2030. I’ve been pretty consistent about that. Okay. 2030. Um, must readad book, poem, or paper for when we achieve AGI. Oh, wow. Um, for when we achieve once we achieve it. Um, well, my favorite book is The Fabric of Reality by David Deutsch. So, I think that still holds. I’d hope to answer the questions in that book with with the AGI. That’s my post AGI work. Brilliant. Yeah. Uh, proudest moments so far in Deep Mind. Uh, oh wow. Um, we’ve been lucky to have a lot. I mean, probably Alpha Fold. Okay. Yeah. Now, a couple games questions. If you were engaged in a highstake strategy game, turnbased strategy game. Okay. Uh, Civ, Polytopia games, uh, and you could select one scientist from history. We’re thinking the the Einsteins, the Turings, Newton’s. Who would you select to be on your team? Uh on my team? On your team? Oh gosh. Uh um probably Vonoyman, I think. I mean, he’s You won a game theorist, I think. And I think I think I think I think he’s the best. Makes sense. That’s clearly how you’d be a good teammate. Yeah. All right. Yeah. Well, Dennis, you do it all. Thank you so much for being with us. Please join me in thanking