Using AI to Accelerate Scientific Discovery

ladies and gentlemen a very warm welcome to this evening's lecture here in the splendor of the sheldonian theatre it is hosted by oxford university's institute for ethics in ai and is part of the oberzee tanner lectures on artificial intelligence and human values my name is nigel shadbolt principal of jesus college i'm also a professor of computer science here in oxford and chair the institute steering group it was my privilege to help set up the institute which brings together world leading philosophers and other experts in the humanities with the researchers developers and users of ai the director of the institute is professor john tasoulis and its ultimate home will be the stephen a schwartzman center for the humanities whose construction is soon to start in recent years ai has gone from strength to strength it's now ubiquitous in our phones the games we play in our cars our drug discovery companies the search engines we use and the translation tools we depend on much of that is down to a new generation of ai methods and techniques that are powered by modern machine learning algorithms great swathes of data and the prodigious power of modern day computing hardware some of ai's most dramatic recent accomplishments owe a great deal to our speaker here with us this evening and the company he co-founded demis asabe ceo and co-founder of deepmind one of the world's leading ai research companies demise's own career and intellectual journey is an extraordinary one a chess prodigy hugely successful computer games developer with a double first in computer science from cambridge demis has always been fascinated by the human brain understanding how it gives rise to intelligence after the success of his games companies he went on to a phd in cognitive neuroscience at ucl followed by a henry welcome post-doctoral research fellowship at the gatsby computational neuroscience unit also at ucl his papers in cognitive neuroscience investigated imagination memory and amnesia and appeared in leading journals such as nature and science he combined his interest in computing and neuroscience with the formation of deep mind in 2010 its compelling ambition to solve intelligence and then use intelligence to solve everything else he and his team used games as the context in which to test new ideas about how to build ai systems using machine learning methods inspired by neuroscience first arcade games and then famously go a previous talk here in the sheldonian in february 2016 pre-figured alphago winning 4-1 against former world champion lisa dole just a month later games have proven to be a great training ground for developing and testing ai algorithms but the aim of deep mind has always been to build general learning systems ultimately capable of solving important problems in the real world deepmind's alpha fold system is a solution to the 50-year grand challenge of protein structure prediction culminating the release of the most accurate and complete picture of the human proteome a core aim for the institute for ethics at ai is to bring together world leading academics and the practitioners at the cutting edge of ai development tonight we will hear first-hand experience of ai's enormous potential to accelerate scientific discovery experience which will inform our research and thinking about the critical ethical considerations that must be considered by policy makers and technical developers of ai demis has predicted that artificial intelligence will be one of the most beneficial technologies ever but that significant ethical issues remain please join me in welcoming demis has to deliver tonight's tana lecture using ai to accelerate scientific discovery [Applause] thank you thank you sir nigel for such a great introduction and it's a real pleasure to be back here in oxford in the shedonian and giving the tana lecture it's a real honour so what i'm going to talk about today is using ai to accelerate scientific discovery and in fact as you'll see throughout my talk this was my original motivation and has always been my motivation behind spending my entire career and trying to make ai a reality and i'm going to talk a lot about some of our most recent advances are actually now coming to fruition and especially the last year or two of using ai to crack difficult scientific problems but i'm also going to talk about the lead up to there and how i think about the games work we did originally in the foundational work we did originally and last time i talked here was just before the alphago match um in in korea so that was kind of a major moment for us and how in the last even five six years things have progressed enormously so just uh to sort of talk a little bit about what our vision was behind deepmind back in 2010 it's quite hard to remember the state of ai back in 2010 because today as as nigel was saying that you know ai is ubiquitous all around us it's one of the biggest buzzwords in industry um it's sort of hard to remember just 12 years ago almost nobody was talking about ai i would say and it was almost impossible to actually get funding in the private sector for ai at all and we have many funny stories back in the day of trying to do some fundraising back in 2009 and 10 and most people thinking we were completely completely mad to be embarking on this on this uh on this journey but we we founded it with this in mind of trying to build one day an apollo program like effort to build agi artificial general intelligence and we use this term artificial general intelligence to distinguish it from sort of normal everyday ai where we're talking about a general system that can that can perform well on many tasks to at least human level um and that's the the sort of general aspect that we are always striving for in all the work that we do so we're still on this mission now and i think we've done um a pretty good job of of of um basically staying true to this original vision that we had in 2010 when we were just a few people in a small little office in an attic in russell square so as uh as nigel said our original mission statement was um step one solve intelligence step two use it to solve everything else um we have updated that uh mission statement a little bit still means the same thing but just to be a little bit more descriptive now uh in the last few years just to be a bit more clearer about what we mean by solving everything else what exactly we're talking about and so the way we discuss our mission now is solving intelligence to advance science and of course for the benefit of humanity and that's always been the cornerstone of what we think about when we think about what should we apply ai to now there are two broadly two ways that i think ai can be attempted to be built one is the sort of i guess more traditional way of building logic systems or expert systems and these are hard-coded systems that effectively teams of programmers solve the problem they then incorporate those solutions in sometimes very clever uh expert systems but the problem with them is that they are very limited in terms of what they can generalize to so they can't deal with the unexpected and they're basically limited to what the programmers foresaw uh the situations that the system might be in and of course this this line of work was inspired by mathematics and logic systems on the other hand the big renaissance in the last decade plus is the um uh is the sort of um progress of learning systems uh of course in the in the 80s there was a flurry of work done on neural networks uh then that died down um we now know that probably we didn't have enough computing power or data maybe not the right algorithms as well um but basically uh in essence the ideas were correct so an idea of a learning system is that you know it learns it for itself solutions for itself from first principles from directly from experience um and the the amazing thing about these systems and their huge promise is that they can maybe generalize to tasks uh and that is that it's not been programmed for explicitly and maybe solve problems that we ourselves as the designers or scientists behind those systems don't know how to solve so of course that's the huge potential and also the risk of these kinds of systems and um originally these kind of learning systems took a lot of inspiration and also could be validated some of the ideas like reinforcement learning and and neural networks by systems neuroscience and comparing uh what these systems do comparing them on a systems and algorithmic level to what we know about how the brain works now everything we do at deepmind of course is um on the learning system side and we've been lucky enough to be in the vanguard of this uh almost revolution or renaissance and the last decade of these types of approaches so um how do we think about what's our sort of i guess special take on on learning systems and how powerful they can be so there are kind of two component um algorithms or approaches one could say that we fuse together so of course there's deep learning or deep neural networks and uh the way i think about this is that the deep neural network system is there to build a model of the environment of the data and the experience and then what do you use that model for well you can use reinforcement learning um which is a sort of goal seeking and a reward maximizing system to you can use that model and use it to plan and basically plan and make take actions towards a goal a goal that may be specified by the designers of that system so you have the model and then you have the kind of the the action uh and goal solving element of the systems so we and one of our early innovations was to sort of fuse those two things together at scale we call it deep reinforcement learning now and um the the cool thing about these systems is that they can discover new knowledge from first principles through this process of trial and error using these models so the idea here on this diagram of the agent system is it gets observations um from the environment those observations go towards building and updating an internal model of how the environment works and the transition matrices of of the environment there's some goal it's trying to solve in the environment achieve and then after its thinking time has run out it has to select an action um from the from the action set available to it at that moment in time that will best get it incrementally towards its goal and then the action gets output it may or may not make a change to the environment that drives a new observation and then the model updates further so you can see with this type of system the agent is actually the ai system is actually an active learner it participates in its own learning so the decisions it makes in large part governs what experiences and what data it will get next to learn more from so um so although this is a pretty simple diagram and basically describes the whole of reinforcement learning the reinforcement learning problem there's huge complexities of course of theoretical and practical complexities underlying this diagram that need to be solved but we know that in the limit this must work because this is how mammalian brains work including humans this is one of the learning mechanisms that we have in our own brains reinforcement learning was found to be implemented by dopamine neurons in the brain in the late 90s so we know if we push this hard enough this should be one path towards uh general artificial intelligence so what do we famously use this for um alphago uh was uh you know the program that i think we did a lot of things before this like atari games and other other proof-of-point proof points but alphago was really our first attempt at doing this at huge scale on uh to crack a big problem that was unsolved in ai kind of one of the holy grails of ai research which is a program to beat the world champion at the game of go and i want to talk a little bit about this um in hindsight now knowing what i know now how i've reinterpreted what we did with alphago and i think i can explain it in a much more simple and general way than perhaps you know how i was explaining it back five six years ago when we were in the midst of building this system so just for those of you who don't know um i don't know why that's not updating there we go so this is the game of go oops this is the um the game of go the board game and um it's a phenomenal game uh and it's um it's much more esoteric game and uh artistic game one could say than chess so occupies the same intellectual echelon chest stars in in in the west in china and japan and korea and other asian countries they play go and go has resisted a sort of old-fashioned logic system and expert system approaches whereas chess was solved by those things because of various factors one is the search space is truly enormous in go it's roughly 10 to the power 170 possible board positions um which is way more than there are atoms in the universe so there's no way one could exhaustively search all of the possible uh board positions in order to find the right path through even bigger problem actually is that it's impossible or thought was it was impossible to write down an evaluation function to sort of hand code an evaluation function which is what most modern day chess programs use so um and the reason is because go is such an esoteric game right it doesn't have materiality um in chess you know as a first approximation one can add up the piece values on both sides and that will tell you very crudely like who which side is winning in that position and obviously you need to know that in order to make decisions about what to do next so many people are attempted to over 20 years since deep blue attempted to write to construct um these evaluation functions for go and one of the issues is is that go players themselves do not know consciously at least what that information is so because it's so complex the game they actually use their intuition uh rather than explicit calculation in order to deal with the complexity of go where chess players if you ask them you know how why did they make a decision a chess grandmaster will tell you will be able to tell you explicitly the various factors involved a go player generally won't do that they'll just say things like it felt right um this felt like the right move which is what i think also makes go an incredible game but of course intuition is not something one would associate with computer programs especially logic systems um and maybe in the q a we can discuss a little bit more about what intuition may be but i don't think it's it's sort of i don't think it's my conclusion now after doing all these games uh and indeed some of the science things we've done is that it's not some mysterious thing it's actually information that our brain knows about and has learned through experience of course i mean there's no other way one can learn information but it's just it's in the association cortices so it's not actually consciously available to a high level cortex so it seems mysterious to us you know how we ride a bike how we swim these sort of motor sensory motor things we're able to do because our conscious part of our brain cannot access those representations so and if we can't do that then we definitely can't explicitly code it in in some logic code right which is why traditionally those tasks including things like computer vision have been quite hard for logic systems to solve even over the last 50 years so a lot about what we were doing was trying to approximate this kind of intuition in these learning systems so how did we work and i'm actually going to describe not just alphago here but the whole series of alpha x programs so alphago the original one that beat lisa dole in in 2016 and then alphago zero that then didn't need human data to learn from just learn for itself and then finally alpha zero which uh uh could play any two player game so i'm going to sort of describe them all roughly speaking with this uh with this sort of demonstrative diagram so the way you can think of all of these systems is we're initially training a neural network through self-play for it so the system plays against itself and and it learns to evaluate positions and to pick the most likely moves that are most you know useful for it to look at right so that's what it's got to do now initially it starts with no knowledge right so you have an initialized neural network it starts with zero knowledge so it literally is moving randomly right so that's we can call that version one right that's the neural network and what it does is it plays roughly a hundred thousand games against itself okay and so that then becomes a data set so that hundred thousand gains we take that as a data set and what we try to do with it is train a version two of that network a new neural network but we try and train it on this version one data set to predict in the middle of a position in the middle of the game from a position in the middle of the game which side is going to win right so uh so kind of predict ahead of time and also what sorts of moves does the v1 system choose in a particular position right so that's trying to do is trying to be better at both those two things and then what happens is um we train that v2 system and then we have a little mini mini tournament between v1 and v2 so it's roughly 100 games and they have a little match off and basically if there's we the v2 system hits a particular threshold win rate 55 in this case then we say it's significantly better than v1 right and if that's true then what we do is we replace v1 with version 2 network this new network in purple and that of course plays another 100 000 games against itself right and now it creates a new data set but this data set now in purple in the middle is slightly better quality than that first data set right because the player is slightly better and to begin with almost imperceptibly better so it's just slightly better than random now right but that's enough signal to then train you know of course we train a version three system and that plays off against version two now if um the version if there if if you don't reach this 55 win rate what you do instead is you take back the version two and you continue to generate more data without another hundred thousand games so then you have two hundred thousand to train your next version three right and eventually that version three will be better than version two so after one does this um around 17 or 18 times you go from random to better than world champion that's it right and and you can do this with any two-player game perfect information game right so the same network can do that get to better to world champion within you know 20 to 30 generations of doing this so you literally and we got to the point where so fast you literally set it off in the morning you could play chess about it at lunch time and maybe just beat it and then by tea time you know you no chance literally in the day you could actually see the evolution in one day it's kind of incredible to watch as a as a chess player so what it what is it doing then uh in terms of um thinking about this enormous search base so what's happening is and and the and the sort of i think advance of alphago one of the advances was combining this neural network system or model uh with a a a kind of more classical tree search algorithm in this case we use monte carlo tree search um and you can think of the the tree of possibilities looking a bit like this in go where each node here is a positioning in go obviously shown by these little mini go boards and you can imagine if you're you're some middle game position you know there's just this countless 10 to 170 possibilities in the limit how is one supposed to find the needle in the haystack right the good moves that could be world champion or better level decisions so what the what the neural network does is it constrains that model constrains the search to things to make it tractable right to things that are reasonably likely to work reasonably effective and it can evaluate that at each node level with its evaluation function and so instead of having to do you know 10 to the hundreds of of of possibilities one can just zoom into you know mere thousands 10 000 or so um searches and so therefore instead of that searching the entire gray tree of all possibilities one just looks at this far more limited you know uh search tree in blue here and then when you run out of thinking time of course you select the best path that you found so far in pink here so you know we did this back in 2015 and then in the subsequent years we still work on this now there's a system called mu0 which is our latest version of this that can do not only do two-player perfect information board games but can also build models of it for of its environment so it can actually also do things like atari games and video games where you actually don't have the the rules of the game given to you it has to actually figure that out for itself through observation as well so it's one step even more general than alpha zero and what we did with alphago of course now is um as shown mentioned is we we took it to seoul in 2016 in this million dollar challenge match with lisa dole and some of you may remember this um but we won 4-1 you know it's a huge thing in especially in asia and in korea i mean the country almost came to stand still and there's over 200 million people uh watch the games and it we won 4-1 and and experts in both ai and in go proclaimed this advance to be you know a decade before they would have predicted but the important thing in the end was actually not just the fact that alphago won the match but how it won was it was i think really instructive so i'm just gonna give one example of this but actually alphago i think has in the end changed the way that we as human beings view the game of go um but this is the most famous uh game of that set of five there are actually some amazing different games including the one that lisa dole won with a genius move in game four and um but move 37 in game two i think will go down and go history and this was the board position at that time and um haven't got time into to go into why this was so amazing but suffice to say alphago here was black and lisa doll is the white stones and um this is very early on in the game move 37 you know go games last for a few hundred moves generally and uh alphago played this move 37 stone with on the right hand side here uh marked in red and the the amazing thing about this was the position of the stone was on the fifth line from the edge of the board and that if you're an expert go player is unthinkable it's like you would be told off by your go master that you should never do um make a move like that um because it gives white too much space on the side of the board but alphago decided to do it never seen before in master play would be recommended against and then a hundred moves or so later it turned out this stone this move 37 stone was in the perfect position to decide the battle that spread out from the bottom left all the way across the board and it was just in the right place to decide that battle which decided the whole game and almost as if it had presently sort of seen that influence ahead of time so now people play on the fifth line all the time i'm told so so this has changed this has changed everything and there's you know multiple books now written about alphago's strategies and you know this is an original strategy because um because uh this is not something that alphago could have learnt from human play in fact it would have learned the opposite it would have learned not to do this kind of move so if you're interested in more about alphago i recommend you um you know this amazing award-winning documentary that was done by an independent filmmaker it's on youtube now um if you want to see the sort of ins and outs of it it was you know very emotional as an experience for us from all sides especially me being an ex-games player i could really understand it from lisa dole's point of view too so as i said we then took this to alpha zero a couple of years ago two three years ago now and generalized this to all two-player games and these graphs show how alpha zero did against the best machines at the time in the specialized games of chess it beat the best version of stock fish which is this incredible hand crafted system the the descendant of uh deep blue and it was able to beat stockfish eight which was the best stock fish at the time in four hours of training um it could be alphago alpha zero beat alphago in eight hours right at go um and then and we just tried it with one other game the japanese chess shogi actually which is a really interesting variation on chess and um it could be the best uh handcrafted program called elmo within two hours of training um the same system all three games so um so that was generalized and then of course because i'm a chess player i play a little bit of go but i'm not very strong but so chess is my game and so for me this was the most exciting part of applying alpha xero because i actually had a discussion with murray campbell who some of you will know was one of the project leaders behind deep blue at ibm back in the 90s and uh we just i think we just um were about to play the lisa doll match or maybe we just finished and i was giving a lecture at a conference and murray campbell was there as well in the audience and he came up to me afterwards and we were discussing you know i said to him i'm thinking about maybe we should try this with chess and see what happens and i wanted to know what his prediction would be you know do you think um these these incredibly powerful handcrafted systems like stockfish could be beaten um was there any more headroom in chess you know we we we chess is probably the oldest application of ai right i mean turing and shannon and people like that have have all tried their hand every ai researcher at some point has tried their hand on on a chess program back to the 40s and 50s even if turing had to you know run the program by hand on a piece of paper and a pen um but and then of course in the last 25 years or so you know world champions have been studying with their chess programs and mapping out all of chess opening theory all of these things so it was a legitimate question actually to ask is was there any more head room left and what sort of chess would alpha zero play if if if if we were to train it from first principles and play it against these uh amazing you know um hand engineered uh you know monsters in some sense of a machine you know incredible calculating machines and so of course we couldn't actually come to an agreement on that and that you know as a scientist in the audience we'll know that's the sign of a good question i think where either answer would be interesting right if we were to win and there was some new style out there that would be incredibly interesting and also be interesting if there were if these handcrafter systems at least in one domain chess um had reached the limit so we got off and started doing that and i'm pleased to say that alpha zero not only played stronger but it did come up with um a completely new style of chess which i think and my chess friends tell me um is more aesthetically pleasing as well as a as a chess program obviously subjectively from a human expert's point of view and the reason it is is because what it does and it does many innovations and um but the main one is that it favors mobility over materiality so traditionally handcrafted chess programs have always favored materiality you know the joke within the chess circles is a you know chess computer sees a pawn and then grabs the pawn because it loves material because it gets plus one in its evaluation function and then it tries to hang on for dear life in a you know really ugly position but it wins because it never makes any tactical mistakes so it's sort of very effective but it's a little bit sort of aesthetically you know um unsatisfying one would say as a style um but instead of that actually alpha zero does the opposite it loves sacrificing pieces and material um to get mobility to get um more mobility for it for its remaining pieces and this this so this is a game from we did a you know 100 match between alpha zero and stop fish and this is one of the very and then we gave it to some the british chess champion actually to analyze and uh he picked out like the the coolest positions and this is my favorite it's sometimes called the immortal zugswang game zugswang is a phrase in chess german phrase that means any move that one makes in that position makes your your position worse so it's a special type of position where you're you're in zugzwang which means anything you do it's going to make it worse which is very unusual and it's super unusual in this kind of position for those of you know chess wear black which has got more pieces the two rooks and the queen so it's got its big material advantage very powerful pieces and the most powerful pieces remaining in chess but they're all stuck in the corner and alpha zero has sort of sealed them up you know with cement with its pieces and basically none of those pieces can move right so this is kind of an incredible position so almost anything blacked out in fact anything black does in this position it's black to move will make its position worse even though it's got all of these very powerful pieces so um so that was one invention innovation there were lots of interesting properties about alpha zero that i won't go into um but one can think about well why is it that alpha zero plays like this and um traditional chesh engines didn't um nowadays actually interestingly they've updated stockfish to include some of these ideas by hand in stock fish and actually now it's very powerful it's even more powerful so it's kind of interesting hybrid system but um my feeling is that uh it's better evaluating positions than chess engines so that's one thing so it's got a better evaluation function and the main thing is it doesn't have to overcome these inbuilt rules that's why it like sacrificing pieces because if you think about it a hardcoded chess engine would have to calculate in each search tree that if it was going to sacrifice a rook for a bishop you know that's minus two points is it going to get back those two points of value right within its search tree horizon alpha zero doesn't have to worry about that because there's no rules like that in there it can evaluate things contextually based on the particular situation at hand and the patterns involved there and also the other big thing is stock fishing programs like that they have thousands of handcrafted rules so one problem is generating those rules but an even bigger problem in my opinion is balancing those factors together right that's a huge sort of hand crafted juggling act and instead of that obviously alpha zero learns itself how to balance out the factors that it's learned um and to uh and to do that automatically so um one can actually see how efficient this system is based on the amount of search that traditional search engines have to do per each move they make um and a human ground master makes only the order of like looks at about 100 moves per decision so incredibly efficient with our models um and the state-of-the-art chess engine like stockfish would make tens of millions of um evaluations per move and alpha zero was um you know sort of in the middle here in terms of orders of magnitude uh uh tens of thousands of moves so not as efficient as as human uh top human players but far more efficient than the search one would get in um in the these search engines so again uh if you're interested in the details about all your chess planes and the details about what this changed um the the the british champion and uh and natasha reagan uh wrote an amazing book called game changer when we gave them behind the scenes access to to alpha zero uh and what new motifs they found at least a dozen new motifs they found in chess and the cool thing is that was very gratifying for me is that uh people like magnus carlsen who's the current world champion incredible player um you know he said a few years back he was one of the first people to read the book and who we sent it to and you know i've been influenced by my heroes recently one of which is alpha zero which is really cool to say and he actually incorporated he's so talented he was able to quite quickly quicker than all the other chess players incorporate some of these ideas into his play and then gary kasparov he used to be a hero of mine when he was world champion when i was growing up and playing chess um you know he worked the forward for the book and he said programs usually reflect priorities and prejudices or programmers but alpha zero it learns for itself and i would say its style reflects the truth which is a you know i think a beautiful quote so we've been lucky enough to have several of these sort of fundamental breakthroughs in games we started with atari um and our program called dqn being able to play atari games from directly from pixels and maximize the score from just from pixels not being told the rules of the game alphago and alpha zero i just mentioned and then we went further with programs like alpha star which played the most complex video game called starcraft 2 which is a very complicated real-time strategy game with huge other challenges it's only partially observable it's not perfect information there's an economy system to it and you have generally thousands of possible actions you can take any for any choice not not a few dozen um and and we managed to also get to grand master level at that so that was all of our games work but really it was leading up to this moment which um in the last couple of years it's been just so exciting and um and and so gratifying for us to make progress with which is that the games and i love games always will love games playing them designing them and using them as testing grounds they were the perfect testing ground for developing ai but ultimately their aim was not to play games to world championship level it was to build general systems that could generalize and solve real world problems and the one that's particularly passionate for me is using ai for scientific discovery and there are three things that i look for when currently when we want to select a scientific problem that we believe our systems could be good at so number one is we actually search out massive combinatorial search bases or state spaces so the bigger the better actually why is that well because we know then traditional methods and exhaustive brute force methods won't work right so we're in a raisin where something else is needed and and we think that we're good at that something else number two is that we want to have um you know we like problems that have a clear objective function or metric that one can specify so that you can optimize and he'll climb against it with your learning system and then number three is we look for problems that either have a lot of data available to to learn and train from or and ideally it's and or an accurate and efficient simulator that one can use to generate more data and that simulator doesn't have to be perfect it just has to be good enough that you can extract some signal from the data that it generates now it turns out that when you when you look at a lot of problems with this prism then actually a lot of surprising number of problems can be made to fit these criteria and of course the number one thing we were looking at was protein folding which i want to talk a bit about now and um and we look for problems not only they just fit those three criteria but of course there's always an opportunity cost when you embark on applying ai to something major it's going to take you many years depending on how hard that problem is and we look for something that will have really huge impact perhaps you know we sometimes talk about root nodes that can open up whole new branches of um scientific discovery if they were to be solved and protein folding ticked all of those boxes so for those who don't know what protein folding is it's this classic problem of can one go from a one-dimensional amino acid sequence you can think of it as the genetic sequence for a protein that describes a protein coded by the genome and can you predict from that directly the 3d structure of the protein in your body the 3d form that it takes and the reason this is important is that proteins are basically essential for everything in life every function in your body and it's thought that the 3d structure of the protein in at least in the large part governs its function so if one can understand the structure then one can get closer to the um the function of the protein now until alpha fold came along the way you would do this is experimentally and it's extremely painstaking expert work that needs to be done uh and using you know x-ray crystallography and electron microscopy and the rule of thumb is generally that it takes one phd student their whole phd to do one protein right and that's if you get lucky you can be unlucky so it's it's it's hard and really painstaking and difficult and what happened is that the nobel prize winner christian anfinsen uh in part of his nobel lecture in 1972 so 50 years ago exactly now um he conjectured that the 3d structure of protein should be fully determined by the amino acid sequence i.e this should be possible this mapping um and it's a bit like you know sometimes this problem is called fermat's like fermat's last theorem equivalent in biology because it's a bit like saying this is possible but the margin is too small can't give you the answer and so what happened instead is obviously it set off a 50-year quest in biology and computational biology to try and um solve this problem and uh and it's been going you know ongoing ever since uh uh like the 1970s so the big question is is can um protein structure prediction uh the protein structure prediction problem which is the specific part of protein folding that we're interested in um be solved computationally um just just computationally and levantal who is another famous contemporary of anfinsen uh he in this in the 60s and 70s as well he calculated back of envelope that there would be roughly 10 to the 300 possible confirmations shapes of an average size protein that it could take right so 10 to 300 so that's a good number that's ones we like because you know it's bigger than go and obviously that means exhaustively sampling this is totally intractable but of course the the the the the of light is that in nature in our bodies uh physics solves this right so it's it can if proteins spontaneously fold in a matter of seconds sometimes milliseconds in the body so there's obviously some energy path uh through this so how do we get to this problem well actually it's quite a long winding road for me personally for others in the team less so but for me i actually came across the protein folding problem in the in the 90s as an undergrad in cambridge because um one of my friends in our sort of group of of of colleagues was um obsessed with this problem and he would talk about it i remember this very clearly every opportunity in the bar playing pool whatever it was you know if we can crack this that would open up you know you know all sorts of things in biology and i sort of listened to him and i was thinking about this i was fascinated by the problem as a problem and i felt it was actually very well suited to potentially to ai although obviously at the time i didn't know how it could be tackled but i filed that away as an interesting thing and then it came up again in the late 2000s when i was doing my postdoc over at mit and this game called foldit came out from david baker's lab um who who works on proteins and it was a citizen science game you can see it on the left here and what they've done really interestingly is turn protein folding into a puzzle game right and they actually got you know a couple hundred gamers to fold proteins bit like you know playing tetris or something and um and then some of them actually became really good uh and and i remember so of course i was fascinated this just from games design perspective you know wouldn't it be amazing if we could design more games where people played them they were actually doing useful science while they're having fun that would be amazing and i think this is still the best example of that and um but also you know again protein folding uh was coming up and in fact it turned out that a couple of you know more a few really important proteins structures were found this way by gamers and published in you know nature and nature structural biology and so this actually really worked and that when we then got to you know the third piece of the puzzle was doing go and and and trying to sort of think about what we've done with intuition and other things as i mentioned earlier and i felt that actually you know if we'd managed to mimic uh in some sense the intuition of go players master go players who spent their entire life studying go um you know maybe one could mimic the intuition of these gamers who were only by the way of course amateur biologists right but somehow some of them were able to make counterintuitive falls of the backbone that were if you just followed an energy landscape in a greedy fashion uh one would not you know reach a local a local minima or local maxima and and you would not um be able to find the right structure so um so it's almost the day after we got back from korea we then um you know i instigated the alpha fault project and and uh i thought it was the right time to basically start uh working on this problem the other important piece of the puzzle was this competition called casp um which is uh sometimes thought of as like the olympics for protein folding and it's sort of run every two years an external benchmarks an amazing thing actually that i think more areas of science should do and it's been run sort of uh religiously for every two years for nearly 30 years so you know huge kudos to the uh organizers john mull and and his team for doing this and organizing it so professionally for every two years without fail uh for 30 years and the cool thing about it is it's a blind prediction assessment so there's no way you can accidentally sort of train on test data or any of these kinds of pitfalls because at the time when the competition runs over summer usually every two years the experimentalists globally agree to hold back a few of their structures that they've just found but at that point in time they're the only ones who know what that structure looks like they held back the publication for a couple of months and they give it to john mull and his colleagues to put it into the competition and then you get those it's quite fun um tournament because uh because then you know it's quite exciting you get the email and then there's a new structure that amino acid sequence nobody has ever seen you know knows the structure of and then you have a week to sort of get it back to the competition organizers um before it's published and then at the end of that three four month period they obviously score your predictions against the ground truth which at that point is published obviously in peer review journals that the experimental ground truth and then you get a kind of distance measure between your predictions and the molecules in that prediction and where they really are in 3d coordinate space so when we started um getting involved in this area post 2016 um you know we looked at casp and the history of it and actually there'd been very little progress for over a decade it's sort of the field had stalled and um this uh graph here shows you the um the scores of the winning team on the hardest category of protein where you don't have any evolutionary similar template proteins to sort of rely on so it's called free modeling and this is a percentage accuracy it's called gdt it's a slight nuance of the of the measure but you can think of it as the the number of molecules the percentage number of molecules you've got roughly in the right place to a certain tolerance distance tolerance and you can see they were hovering around 40 or less which is useless for experimentation right basically it's pretty much random and so that was the average and it hadn't really moved and so um what we did in 2018 is that we came along with alpha fold one as our first entry after a couple of years of working on this and we sort of um you know i think we revolutionized the field in a way is that we for the first time we bought cutting-edge machine learning techniques the sort of techniques we developed in in alphago and other new ones for this domain and we um as the core part of the system and we improved the winning scores by 50 you know got to close to 60 gdt here um and then of course we didn't stop there we then re-architected based on that knowledge we actually tried to push that system further and it turned out it hit a brick wall so we had to go back to the drawing board with the knowledge that we had we architected with a brand new system and then that finally reached in cast 14 in 2020 atomic accuracy so accuracy within the width of an atom right for all the molecules so when we look at the the scores and the results of of of cast 14 what you see here is that alpha fold two uh this is the root mean squared error um is is is less than one angstrom error uh on average uh and um you know from the hundred or so proteins that we're supposed to predict so and one angstrom is the you know the width of a basically a carbon atom so um so that's finally that was the magic threshold that um john mull and others of the organizers said that they always said outcast to do because that would make you competitive with experimental techniques which are roughly you know the best ones are at that kind of error rate so if one could do that computationally then suddenly you have a technique that could be you relied on in tandem with experiments or instead of and so alpha fall 2 got a got an error of 0.96 angstroms which was three times more accurate than the next best system in cast 14 even though those systems had obviously incorporated the alpha fold one um techniques that we'd already published by then so this led to the cast organizers and john mull declaring that the structure prediction problem had essentially been solved after all of these years and this is what the predictions look like so um the ground truth is in green uh and you can see the prediction from alpha fold two in blue and you can see firstly proteins are exquisitely beautiful it's one thing to know that i've learned over over the many years i've been working on this now they're like exquisite little um nano machines and um and you can see how accurate the overlays are and we were standard of course when we first got these results back um and then you know there are many uh this is the this is the architecture for alpha fault which doesn't have time to go into the details of today but there were a huge number of innovations that were required to to make this work and the key technical advances were basically first of all i should say there was no silver bullet um it needed uh actually 32 component algorithms uh described in 60 pages of supplemental information actually in the in the paper um and that was required and every single part of that was required so we did these ablation analyses which sort of took out components to see if we could get away without having them and the result of that was everything was required um and the three key sort of takeaways of why alpha fall ii was an improvement over alpha fold one is we made the system fully end to end um so and with it you can think of it as sort of going end to end with a with a recycling iterative stage so over time it sort of jigs the the protein structure nearer and closer and closer to the final structure that it's going to predict and our alpha fold one system didn't do that it went from the amino acid sequence to this intermediate representation called a distagram which is a pairwise uh distagram of all the protein molecules and their distance to each of the other molecules the other end molecules and then for map we used a different method to create the 3d structure but with alpha fall 2 we actually made this end to end so we went straight for predicting the 3d structure and those of you work at machine learning will know that generally speaking if you can make something end to end and optimize directly for the thing that you're after usually uh your system will be uh will have better performance we use an attention-based neural network to infer this implicit graph structure of the of the of the of the residues of the amino acid sequences um in alpha fold one we use the convolutional neural net which was sort of borrowed from computer vision and if you think about it that was introducing the wrong bias into protein folding because with computer vision you know pixels next to each other are obviously going to be correlated in an image right in some sense so convolutions make sense but actually for a protein um the amino acid sequence you know residues that are next to each other or close to each other on the string of letters may not end up being near each other once you get the full 3d fold or things very far away could end up folding over near each other so in a way we were giving it the wrong biases so we actually had to remove that and then finally we built in um some biological and evolutionary and physics constraints into the system without impacting the learning um and again usually so you can think of it as a little bit of a hybrid system that um usually if you put in constraints that impacts the learning and we managed to do do that without that so this was a huge research effort over sort of five years took about 20 people at its maximum and it was a truly multi-disciplinary effort so we needed biologists and physicists and chemists as well as machine learners and i think that's an interesting lesson maybe to learn about cross-disciplinary work in ai for sciences is you need the experts also from from the domain and then the final maybe interesting point to note on this is that normally we're always after generality so you can see that from the the journey from alphago to alpha zero was we increasingly made things general right you start with performance then you start throwing things out of that system to try and make it simpler and more elegant and that usually makes it more general as you understand what it is that you're doing but that's because um go and chess and those things were were test beds for what we wanted to do if you are trying to solve a real-world problem that really matters to other scientists or or or health or in this case you know biology then actually um you need to you know you might as well throw the kitchen sink at it right because you actually are really after the the output itself in this case protein structures and that's what we did here we really threw everything we had at it and it's i think the most complex system that we've ever built other things to note about this system is that it's also um alpha fold one was relatively slow took a few weeks to to of compute time to do to do a protein alpha fall 2 took 2 weeks to train the whole system on a relatively modest setup of 8 tpus or 150 gpus which by modern day machine learning standards is quite small and then the inference the predictions are can be done lightening fast you know order of minutes sometimes seconds on for an average protein on a single gpu so when we did this alpha fall 2 we announced the results published the methods over christmas that christmas this is back uh uh 2020. um we were thinking okay how should we give access to the system to biologists around the world and normally what you do is that you set up a server people biologists send you their their amino acid sequences and then you give back you know a few days later you might give them back the prediction but actually what we realized because alpha fall 2 was so fast we could actually just fold everything ourselves in one go right so we just fold all proteins um and we start with you know the human proteome which is like the human genome equivalent but in protein protein space um and so that's what we did over the christmas uh we folded the whole human proteome um and so it's another thing i love about ai and computing is you know you can have your christmas lunch and while you're doing that of our ai is doing something useful for the world so the human proteome so we published that as well in the summer of 21 last summer so alpha fold two we predicted that every protein in in the human body it's around 20 000 proteins represented obviously expressed by the human genome and at the point where we did this experiment experiments 30 years of experiments 34 years experiments had covered about 17 of the human proteome right um and we more than doubled that overnight in terms of very high accuracy structures obviously we folded all of them but very high accuracy so that's less than one angstrom error you know they're sort of up to experimental uh quality uh we went to 36 percent uh and 58 at high accuracy where we call hierarchy when the backbone is is mostly you know where you can be confident in um but the side chains may be slightly out um and then of course the question is what about the rest um the other 42 and and and it may be that some of those um you know alpha fold two is just bad at but increasingly and this is an open research question when we look at it with biologists and biologists often send us in results it's like i look at this one folded really well or this one didn't fold well we often find that the ones that didn't fold well were actually what's called unstructured in isolation so they're disordered intrinsically disordered proteins which means that until you know what they interact with they're basically squiggly bits of string and then presumably when they interact with something in the body they then another protein usually they'll then form a shape but we don't know what that shape is in isolation right we may not even know what it interacts with at this stage so actually people have turned this around now to use it as a as a disorder protein predictor so where alpha fall doesn't do well perhaps that's pretty good evidence that it's a disordered protein which of course is very important in things like disease you know alzheimer's other things are thought to be to do with um badly folded or disordered proteins one of the other things we did which was a nice innovation for for alpha fold was was have the system predict its own confidence in its own predictions and the reason we did this is we want a biologist to use this who maybe would not care about the machine learning teaks or not understand them or frankly it would be irrelevant to them they would just be interested in the structure and we wanted to make sure that they were easily able to evaluate um the quality of that prediction and what parts of it they could rely on right and which other parts they maybe need to check experimentally and um so what we did is alpha fall basically we put we produced predictions they were split into three uh uh uh um uh three thresholds over 90 was what we call very high accuracy so less than one angstrom error experimental quality greater than 70 was the backbone's correct and then less than 50 maybe these red regions so you can see in the database that's what they look like it's something that should not be trusted we did a further 20 model organisms covering all of the critical um model organisms used in research and some also some important other ones in disease like tuberculosis and also agriculture like wheat and rice and a lot of these these gene sorry proteomes are much less covered than the human proteome right of course the human one has been where the most effort's been that's 17 for some of these organisms it's like less than one percent so for the researchers in those plant scientists and other things you know this is a huge boon for them because they would never have the resources to spend that time to crystallize the proteins they're interested in we then teamed up with embol ebi the european bioinformatics institute at cambridge and they're amazing as a partnership team they host a lot of the biggest databases around the world already and we thought what the best way to host all this data is to just give it to them and allow them to host it and plug it into the mainstream of biology tools and so we had a great collaboration with them and then we basically released all this data for free and unrestricted access for any use industrial or academic because so completely free and and you know it's amazing to see the impact of that and we've tried to sort of maximize the scientific impact of this um by releasing it in that way the other thing we did do and i want to touch this on this at the end is think about the safety and ethics of this and um we consulted with you know over 30 experts in various areas of biology bio bioinformatics biosecurity and pharma to check that what that this was going to be okay to release this type of uh information and they all came back with that they were not worried about this but they were potentially worried about future things so that's something that we we bear in mind they're now a million predictions in the database today we i just want to call out one thing is we especially um ourselves we especially prioritize neglected tropical diseases because those are the ones that affect the developing world the poorest people in the world the most and they're the least researched because of course there's no money in it for pharma companies so they're often it's ngos and and non-profits that have to do the work there so for them it's amazing to get all the structures because they can go straight to drug discovery without having to go to the intermediate step of of finding these structures so we prioritized all these uh diseases and including ones that we've got being given from the who about potential future pathogens and what's the community done with alpha fold already we've seen just in nine months or ten months incredible amount of work has been done um this is really cool on the left here with some colleagues at emble they they they were they used alpha fold and experiment to combine with their experimental data to put together what's called the nuclear pore complex which is one of the biggest proteins in the body it's massive for a protein and what it is is it's a little gateway into the nucleus of your cell and it opens and closes to let things in and they were able to you know it's beautiful if you look at it able to put it all together and then visualize it um i talked about this disorder predictor who top 30 pathogens and actually interestingly uh it's it's helped experimentally are the ones that benefited first from this because they can combine this with their maybe some low resolution uh images they have and if they have two sources information they can then make a sharp prediction from their maybe their slightly lower resolution experimental data and then a computational prediction so um so it's you know been really gratifying to see hundreds of papers now and and applications already with uh being used for alpha fold also in industry too for drug discovery so what is the impact been so you know we already have 500 000 researchers have used the database we think that's almost every biologist in the world has probably looked up their proteins they're interested in 190 countries 1.5 million structures viewed and already over 3 000 citations and we've had some nice accolades along the way from science and nature on on on the method and then over the next year we plan to fold every protein you know in known to science which is in uniprot which is the massive database that has all the genetic sequences and there's over 100 million proteins known to science and we're steadily sort of progressing through that right now and we'll be releasing that over time so stepping back then what does this mean i think that maybe you know we're entering a new era of what i would like to call digital biology so i think the way i think about biology is that at the most fundamental level it's an information processing system albeit an exquisitely complex and emergent one and i think of it as maybe the potentially the perfect sort of regime for ai to be useful in because you know one thing i think of analogous to is in physics you know we use mathematics to describe physical phenomena and it's been extraordinarily successful in doing that of course mathematics can also be applied to biology and has been applied successfully in many domains but i think a lot of these emerging and complex phenomena are just too complicated to be described with a few equations right i just don't really see how you can say come up with you know kepler's laws of motion just from uh of a cell right how would one do that you know just a few differential equations doesn't seem to me likely and i think maybe a learned model is a better way to approach that um and i think and i hope that alpha fold is a proof of concept that this may be possible and uh maybe i should help usher in this new dawn of digital biology and our attempts to go further in that space is obviously we're researching further at deep mind and the science team we sort of doubled down on all these things within the biology team at deepmind and we've also spun out a new company isomorphic labs to specifically build on this work and other related work but specifically for drug discovery to accelerate drug discovery which we hope using computational ai methods can maybe be an order of magnitude quicker currently you know it takes an average of 10 years to go from identifying a target to a candidate drug so just to start closing then i just um you know there isn't time to go into this but it's for us it's been like a renaissance year in some sense i've been having so much fun ticking off all of the my sort of childhood dream projects uh infusion and quantum chemistry in conjectures in maths uh material science weather prediction um this has all become reality now in the last year of applying it to important problems in each of these domains and you know publishing nice uh and important work in each of these areas um in in applications of course there are lots of amazing industrial applications um that we've been doing and that we have an applied team at deepmind that that works with google product teams to to incorporate all of our research into hundreds of products now at google pretty much every product you use of google's will have some deepmind technology in it some of the ones i just want to call out are our data center work and energy optimization of data centers and the energy they use and the cooling systems they use and we're looking at applying that to grid scale now uh wavenet which is the best text to speech system in the world so any any um device that you talk to that talks back to you will you be using wavenet to have super you know really realistic voices um even you know interesting things like better video compression for youtube we can say four percent of the bitrate that um is used to but whilst maintaining video quality and also things like recommendation systems but there's just too many to to to mention actually and then of course very in vogue now and we have a ton of work on this area but the whole talk in itself is large models and we have our own really cool large models that alpha code that can program from a text description and write write code still amazing to me in competitive programming level chinchilla which is our large language model that uses is compute efficient flamingo that's our vision language combined model that can describe images and then gato our latest model that is super general can do robotics video games all sorts of things language just with one model so this is all very exciting but i just want to end my last couple of slides with a bit about ethics because obviously you know this is hosted by the institute of ethics and um and uh it's very important topic and uh not just because of that but uh and and it's also what the tanner lectures are about too and uh so we think a lot about pioneering responsibly this is actually two of our values at uh at deepmind combined you know pioneering and being responsible and um you know i hope i've convinced you and you hope you will realize that ai is this incredible potential to help um with some of humanity's greatest challenges you know i think disease climate um all of these things could be in scope but obviously ai has to be built responsibly and safely and we have to make sure the people who are building these things it's used for the benefit of everyone so we've had this sort of front of mind from the beginning of deep mind and as with any powerful technology and i think ai is no different although it may be more general more powerful than any that's gone before whether or not it's beneficial or harmful to us in society it depends on how we deploy it and how we use it and what sorts of things we decide to use it for and i think it's important that we have a really wide debate about that at places like this and and the institute of ethics i'm very excited to see that being set up and for us to um to interact with the new institute and here just one mention is that dni has been really critical and i think we've been pushing very hard on this the last few years and i think is critical to this um to make sure we get the broadest possible input into the design and deployment decisions of these of these systems especially for the people that this affects the most that each these systems affect the most and that's something we've been pushing very hard on there's still a lot more work to do but we've been making some good progress at deepmind and we've been also doing that with all of our sponsorship that we do we've now done nearly 50 million dollars worth of sponsorship of scholarships diversity scholarships chairs and academic institutions and projects and also funding things like the deep learning in darpa which is africa's biggest uh conference on machine learning i'm really proud to say that a lot of deep minders helped set that up and so there's many many things that we're doing across the industry that we hope is also can act as a role model for the rest of industry so then on ethics and safety this has always been central to our mission because you saw our audacious mission at the start and we even back in 2010 in our little you know attic room we were planning for success and of course what you know we had to think through as scientists what does success mean what will the world look like and obviously if one thinks that through and it's becoming obvious now in 2022 that but it was obvious to us then in 2010 that this we would have to be critical that it would be really important questions that would have to be addressed and part of that so we've been doing this in the background all along and we'll be talking more about this work probably in future we were instrumental in drafting google's ai principles which are now publicly available and they were partly based on our original ethics charter that we've had from the very beginning of deep mind and the aim of these principles and you know you can look them up later if you want to to look at what they say is obviously to help realize the far-ranging benefits that clearly ai could have for everyone whilst identifying a mitigating potential risks and harms ahead of time and we continue to try and act as thought leadership for the ai community on many of these topics strategy risk ethics and safety so what should we do then and i just want to end with this last slide here is what i think we should not do is move fast and break things which is you know sort of the silicon valley trope right and i think we've seen the consequence of that playing out right this can be very extremely effective to get you know powerful systems and growth and other things but i do not think it's the right way to address really powerful dual use potential dual use technologies like ai and um the problem with it is is that you know one one of the things that falls out of moving fast and break things is actually doing live a b testing in the world right with your minimum viable products and other things and of course the question is if one does that where does the you know and then you know option b turns out to be a terrible option right well where does the harm of that happen well it resides in society doesn't it they pay that that pays the cost of your learning because you've done it in the world and it's probably fine if you're just you know doing a little gaming app or photo app or something but we already see with social networks it's not fine when you're a billion user scale and you know things really matter right in terms of like your a b testing i don't think it's um uh responsible to to to do that so what should we do instead well fortunately we already have um uh another method which i think would be better the scientific method which i do think is probably maybe humanity's greatest idea ever and i think it can apply here and i think we should use the scientific method when we're approaching how to deal with these very powerful incredible potential technologies um and what does the scientific method involve here in this domain well sort of thoughtful deliberation and thought ahead of time and foresight ahead of time uh where you um hypothesis generation on what might happen if one were to be successful with what you're trying to do right so how about we think about that ahead of time not afterwards um then there's rigorous you know and careful and controlled testing i think that's one of the main things i learned from my phd apart from all the neuroscience was also the value of control tests i don't think you can really understand like in in a way i think when i started my phd at least i was all about the you know what's that what's the condition of interest and that's the thing that you're going to make you know your new advance with but actually you can't conclude anything of course unless you have good controls and i think that's something i don't think engineers get first time around actually but you know scientists and researchers of course do get that because that's one of the things that you learn from doing a phd a research phd so control testing and controlled environments not out you know in the world until you better understand what it is that you're doing so you know course one updates on empirical data obviously ideally with peer review so you get critique from the outside and people who are independent from your work all of these things are standard in the scientific method right but are not standard in engineering and that all of this is in service of getting a better understanding of the system before one deploys it at scale right and then maybe you find out something so my view is that as we approach um artificial general intelligence and it's a super exciting moment in time as you can hopefully you know get from my talk and and my excitement over that but we need to treat it with the respect and precaution that and sort of humbleness i would say that the technology of this magnitude demands um and i think that's what we are trying to be at the forefront on and and i think i'll be talking a lot more about this in future so i'll just end by on on the on the sort of going back to the science question um i think if we get ai right it could potentially be the greatest and most beneficial technology humanities ever invented and i think of ai as this ultimate general purpose tool to help us as scientists understand the universe better and perhaps our place in it thank you [Applause] [Music] well thank you dennis for that extraordinary tour de force um we do have a little time for uh questions um but we wanted to give you the the chance to kind of give us that sense of your vision um now we've got an opportunity to have questions from the audience um got to wait for the microphone to be handed to them and to stand up if possible when asking questions but i'm afraid there is a kind of discrimination it's only those on the ground floor that could ask a question due to health and safety policies in the theater so um please if you have a question please raise your hand and i'm happy to take questions at this point so um could i yes john perhaps i'll start with john i'll give you the provision there is a roving microphone and uh just declare who you are john and perhaps stand up and just ask a question thank you uh interesting to begin an ethics talk with some discrimination nigel but um i'm john tesules i'm the director of the institute for ethics and ai um thanks so much for a really fascinating and inspirational talk guess i want to ask two questions one is a very general question about the nature of the project you're embarked on so the objective is to generate a powerful all-purpose tool that will help create new scientific understanding and the nature of this tool is artificial general intelligence so that is a tool that can replicate or outperform human beings across a wide range of cognitive tasks the worry is is there a tension there if you had something that could outperform human beings across a wide range of cognitive tasks could we still regard that as a tool or would it become a colleague so you talked about respecting ai at the end but it looks like something with that level of capacity would demand a different form of respect that would preclude the original objective of now treating it as a tool so that's one question the second question is you've talked about what will benefit humanity and so i guess one question i have is along these lines how do you make that determination so you might say look some people have the view that ai applied to military applications will benefit humanity others don't how do you make that determination and i guess there's also this further dimension there's a division of labor in making that assessment do you think too much has been placed on the shoulders of developers researchers corporations and that really government should step in and resolve some of these issues thanks john you know great questions so i think um you know with your first question the reason human capabilities are an interesting mapping is because um the human brain is the only evidence of general intelligence we have in the universe as far as we know so so i think you know there's always the question is how do you know you got there and so um and you can approximate it with millions of tasks potentially so that's one approach the more tasks you have in your grab bag and it can do all of them compare it against human performance you might have done it but you need there's always the possibility that my one might have missed out a particular type of of cognitive ability like creativity or something so that's why i think um and also i think ai can be applied back to neuroscience as well by the way that's one of our scientific areas that we apply ai to is neuroscience itself and better understanding our own minds so i have this view that um as a neuroscientist that this journey we're embarked on with ai will be is the most fascinating journey one can ever take scientifically because there's not only the artifactual building it's it's then comparing that to the human mind and then seeing i think uncovering the mysteries of our own minds you know what's dreaming what is creativity what emotions all of these questions that we have free will potentially even consciousness um the big questions uh i think um building ai and intelligent artifacts and then seeing what is missing in them is a good way to explore that scientifically and so then i don't know the answer to your question i think that's part of this journey is at what point would these things not become just tools and it may even be that it's a design question because um to whether we should build you know what is consciousness we don't know and that would be a whole obviously debate in itself but should we build it to the extent of what it is should we build them in our systems um i would say no to begin with if we have that choice until we better understand them as tools and then we can bring in that extra complexity of free will and you know what who where did they get their goals from why initially it will be designers but if they could be self-generated so i think we're still a long way away from those things but i think we should that's one of the things i think we should inch towards very cautiously and with precautions because uh also it will get to the heart of what it means to be human so and i think that should exactly be done multi-disciplinary with philosophers and and ethicists and um theologians and and and the wider you know humanities i think this is where the humanities comes in um as well as the science so yeah so i think i think that's uh that's uh you know that's all to come okay question thank you so much for for a great presentation karina proclaim research fellow at the institute so you mentioned at various points the potential for dual use and in particular malicious dual use so i'm curious to hear how you approach this topic at deepmind so what precautions or how do you address the potential for dual use yeah great um yes so we have a lot of different um mechanisms now at deepmind that have been built up over time so uh one is the institutional review committee we have which is formed so it's chaired by leila ibrahim our ceo and it's formed with different people from across the company uh you know legal we have ethicists and philosophers as well it's also rotating boards some senior researchers and they get involved early with research projects and try to assess them from all aspects and they will draw on outside experts so they bring in biologists for example for alpha fold bioethicists so things we might not have in house and then they work with the research teams to you know either say no that project should not proceed okay it can with caveats or why don't you build or do it in a different way with these safeguards so that's our prototype um i would say committee that does these things and we're kind of exercising our muscle when the stakes are relatively low currently um so that we can learn from what works and as effective as we um get more powerful systems and obviously over time i think at some point um they've got to be outside bodies that get involved um but the problem is is that is and we've experimented with that too is that a lot of these things are very specific to the technology itself so one has to sort of um almost you know understand the technology is a deep level maybe even have access to it somehow but in a controlled way because one can't just you know open sourcing is not just a penalty either because if it's a dangerous system open sourcing it means any bad actor can use it too for anything um so there's a lot of complicated i think uh ethical questions around this i don't think there's an easy answer so anyone who thinks there is one i think is kidding themselves i'm not i don't you know i i hope everyone realizes the complexity involved but we're you know i think it's pretty i'm very happy with our internal system but i appreciate more is going to be needed than that as the systems get more powerful and impact more of the world okay a question just behind you i think if you just passed the microphone literally behind you thank you hi my name is ulrich i'm a postdoc at the computer science department in the human centered computer group um so deepmind looks like it's this great example of how you can take the best from science and then sort of bring it together with a commercial company and then make very rapid progress and you mentioned in the end here how he thought that the scientific process should totally inspire the the commercial world as were i'm curious about what you think about the other way around so what have you learned by being sort of embedded in google that you think we as researchers should to learn from in order to make more rapid progress yeah you're absolutely right that was the that was the thinking the original vision behind the or so i spoke about the original vision of the of the company this apollo program but um the original vision behind the organizational setup and processes was to be a hybrid like the best of both worlds startups and the energy and creativity and pace that they have and nimbleness and the best from reese from academic research you know the blue sky thinking ambitious thinking that happens there but sometimes with a lot of bureaucracy right so i think that we did actually successfully combine those two things and then when we um agreed to get acquired we combined it with a third thing which is scale and resources of a you know large very successful company like google and i think that's the main lesson is to make sure you do things that huge impact and have um the ambition and realize that you know you can scale things to that and the consequences that come with that but also the potential of that so i think we've done that now um and very well like marry all three of those aspects together it's a daily challenge because as we get bigger one tends to get slower as an organization so i have to you know we have to fight against that all the time um but it's pretty unique i would say the you know the organizational cultural feel of the mind but it could be a blueprint for other i would say grand projects um could be organized in a similar way okay i'm going to just switch to this side now we've got a question there and a question about that so tim yeah so move fast and break things there's a quote from people who built a social network if the mind was to build a social uh social network using the deep mind way of doing things then what metrics would you use would you optimize for your to judge the quality of your social network and the second question that comes with it with do you in fact have a moral obligation to build that social network wow okay so two thanks tim i mean two two complicated questions there i haven't it's it's actually just generally so let's see i have to be careful what i say but i think um you know social networks have never really been my thing first of all so i haven't i haven't really thought a lot about it relative to scientific advances and the sorts of things that are my personal passion uh i would question actually the premise of you of your question which is that do we how much value does weak ties like that give like a sort of superficial connections like that versus deeper ties that you get in real life with your real family and friends i think is an interesting uh thing to understand like are we sacrificing deeper more meaningful moments for hundreds of more superficial moments it's not entirely clear to me that the metric of you know it sounds seductive connect the world right like why would that be bad but this is the thing i'm talking about with the scientific method is to try and think through the full consequences of what that would mean echo chambers you know manipulation all the rest of it that we all know very well i don't need to go into so i think if i was to do something like that i would i would you know use the scientific method again to try and really think through ahead of time um you know what do you want as the outcomes and metrics in fact often trying to find the right metrics that actually drive the right behavior that you think is good in the world is half the challenge it's like asking the right question in science everybody's doing science knows that asking the question is the hardest thing what is the right question and it says and it's especially hard oh you want an answer well i can't i don't i wouldn't i wouldn't want to give you an answer on the spot but we can you know talk about it over dinner but i i i at least one should attempt to start with serious thinking about the question first right that's the first part i don't know what the answer is because i've not given it enough thought but um one should at least understand the meta level of like that's how one should start uh including whether one should do that thing at all um potentially it could be the answer of of that hypothesis generation okay i'm going to try and get uh uh three more questions in we're right up against the clock got about seven and a half minutes uh there's a question here from helen helen and then thank you amelian landomar from yale university and visiting fellow at the research center for ethics in ai thank you for a brilliant talk so you showed us how ai can help us figure out the truth of the universe pretty much how about the moral world how about the political universe philosophy starts with plato's republic which is an attempt to figure out the best constitution surely unless one is a complete moral relativist there are some invariants we're trying to figure out about the moral world could ai help us map that out could it figure out like the best social organization you know borrowing from i don't know all the the things we've tried capitalism socialism libertarianism egalitarianism would it help expand our imagination and perhaps assuming you have an objective function like um satisfying majority and preferences subject to constraints to protect minority rights or something like that what do you see in the future we took two thousand years and we haven't made much progress good question i mean look i think the morality and political um science i think is one of the hardest things that ai you know i think it can contribute in some way but i would say it's far harder than the physical sciences right or the life sciences because the most complex things in the in the world are humans but human beings to understand and to to model and to and to understand people's motivations especially in aggregate i think one way it could help is um there's also the question of even if an ai theoretical ai could come up with a better political construct would humans beings and society accept that or even care or understand it so there's all those questions to try and and would it be implemented correctly obviously there's obviously implementation problems um i think more interesting maybe would be in and i've talked to economists about this is um and we did quite a lot of research on multi-agent systems so again having a little sandbox or simulation of millions of agents with interacting with each other um with motivations and some goal-seeking things and i think we're missing that experimental test bed actually from political science and economics quite a lot because again economics is one of those things where and political science where you sort of have to test it live a be tested in the world it's like are we going to go for this political system or not should we raise inflation or not well you've got models but then you actually just have to do it and then see oh it's you know cause a recession or something where maybe we shouldn't do that next time and and so we better if i think if we had a a simulation or a sandbox perhaps populated with ai systems that are approximate to you know uh idealized forms of humans and then we can maybe make some interesting uh uh we can do some experimental work in that uh much lower stakes so i think that could be really fascinating exploration area for things like market dynamics and setting the environmental settings to create you know more cooperation or something i i would be if i was economist i would be trying to to use all those things i used to be fascinated when i was a kid with um santa fe institute and they used to do lots of really cool models of agent-based systems and little grid worlds and i loved you know artificial growing artificial societies i think by axelrod i love those kind of work actually what was used to dream about going to santa fe to to work on something like that um i still think that would be pretty cool uh to to to have some sort of system like that let's see if we can squeeze just a few more there's a question uh chap in there who caught my eye there yes you just very and try and squeeze them in because there's two more questions over here and i'm not going to get everest questions that you're going to have super yeah i just have a quick question to be honest so i think you at the end you mentioned to kind of creating ai in the image of scientific method and and the title of your lecture is advancement of science through ai but in what sense do you think that neural networks are the little or the limited understanding i have of ai is in what sense do they follow the notion of scientific method we have do they also is there any sense of talking about hypothesis and then testing because it doesn't seem that neural networks work in that way they're opaque for most practical purposes and if they do outperform us should we just get rid of the scientific method thank you so so by the way it's not in the image of scientific method just to be clear right it's the it's using the approach of the scientific method i'm not sure what image inside of method means but um and yes today that is true that a lot of the systems we have are kind of black box like but i think that's exactly what we should be doing more work on is is making them less opaque there's no reason why they should be the way i say it to my neuroscience team is look we understand quite a lot about the brain now the ultimate black boxes we have mri machines and amazing tools and single cell recording so it's amazing and that's why i got into neuroscience in the mid 2000s so we can actually look into we don't have to do um philosophy of mind necessarily although we should know about that but we can actually you know empirically look at this not just to uh introspection and um so as a minimum we in in the field of artificial minds we should know as much about them as we do with uh uh the real the real brain and we don't know everything about the real brain obviously there's tons still we don't know but still there's a lot more that we do know than we do about these artificial systems and it should be the other way around that should be the that should be the minimum we understand because we have access to every neuron you know neuron artificial neuron in the artificial brain and we can completely control the experimental conditions so as a minimum what you know so i sometimes say this is a challenge to the team what's the equivalent of fmri for for a neural network right what's the equivalent of single cell recording we do ablation studies so we have a whole neuroscience team that's thinking about this and bringing neuroscience techniques analysis techniques over to uh ai now in in the in the defense of the engineers one of the reasons that this has happened is because the brain's obviously uh a static system we're all fascinated by of course right but artificial systems change over time like alphago is now in ancient history of ai right although it was very meaningful at the time so and it takes years to study a system right takes years to build it and then it years to study it so should you use that researcher time on studying a system that itself will be out of date by the time you come to any conclusions about it so i think only now are we reaching the point where we have systems that are interesting enough do enough interesting things in the world like you know large models and and and alpha fold type things that probably it's worth spending the resource researcher time on that and i so i think over the next decade we're going to see a lot more understanding of what these systems do i don't think there's some weird reason why that can't happen okay there are so many more questions i am literally in the red now i'm going to have to call this to a close i do apologize there is so much pent up i think interest and questions for you demis all i can say at this point is absolutely uh a wonderful lecture uh we're five minutes later than we should have been michelle daniel runs to a strict regime when it comes to timekeeping you gave us the most fascinating insights and you have given i think to the world uh with your company and your own talents a quite wonderful vision of a future in which ai can help us flourish empower us and not oppress us so thank you very much [Music] [Applause] you

Channel: Institute for Ethics in AI Oxford

Views: 119,166

Keywords:

Id: jocWJiztxYA

Length: 92min 30sec (5550 seconds)

Published: Wed Jul 13 2022