Demis Hassabis

Demis Hassabis0:00

This is the BBC.

Presenter0:02

Hello, I'm Kirsty Young. Thank you for downloading this podcast of Desert Island Discs from BBC Radio 4. For rights reasons, the music choices are shorter than in the radio broadcast.

Presenter0:13

For more information about the programme, please visit bbc.co.uk slash radio four.

Presenter0:41

My castaway this week is the artificial intelligence expert Dema Sassabis, although to tie him to just one subject is to rather understate his achievements. He's also a neuroscientist, computer games designer, entrepreneur, and world-class chess player. So what's he currently working on? Simple really. A potentially meta solution to any problem.

Presenter1:03

Given his track record, it's clear he's A. entirely serious, and B probably up to the job. A child prodigy, he won a place at Cambridge, aged 16, going on to graduate with a double first. In his thirties, he and his partners sold their artificial intelligence company, Deep Mind, to Google for a reported £400 million. He says, well, a lot of interesting things, but here's a flavour: trying to build an artificial mind is, I think, the best way of investigating some of the mysteries of our own minds, like what consciousness is, what dreams are for, what creativity is. So welcome, Thenis Hassabis. Dreams, creativity, consciousness, they are.

Presenter1:47

really at the very core of what defines humanity. You uh it strikes me you're clearly not somebody who uh suffers from existential overwhelm. You you you like you like the big questions, do you?

Demis Hassabis1:58

Yes, I've always been drawn to the huge questions in life, you know, the meaning of life, why we're here, what's going on in the universe, and things like physics and science and artificial intelligence are ways of exploring those uh mysteries.

Presenter2:10

Now this notion of this potentially meta solution to any problem, the idea is that there is a key available that will unlock all understanding.

Presenter2:21

At best, that sounds unlikely, and at worst, it sounds preposterous. Put me right.

Demis Hassabis2:27

Well, if you think about how civilization has evolved and all the things that, you know, human society has built around us, it's all the product of intelligence. So intelligence is this kind of meta-solution to everything else that as a society we might want to achieve or understand. And I've always felt, again from a young age, that computers can be this kind of magical device that can extend the power of the mind. And I think artificial intelligence is the sort of the end point of that.

Presenter2:54

Um very intelligent people, highly intelligent people, are often very unusual people. What what's the most unusual thing about you?

Demis Hassabis3:02

I think I probably have some quite unusual habits. I generally sleep at about four in the morning. You know, I'll get into work around 10 a.m., do a full day's work in the office, come back for dinner, spend a bit of time with the family, and then start a second day's work at 10 p.m., 11 p.m. and go on to the small hours of the morning. Usually that's the time when I do my research. So I'll be reading about the latest academic papers. I like doing all of my kind of creative thinking in the small hours of the morning. And I've a lot of habits like that that I've learnt maximize the way that I think.

Presenter3:36

Can you give me an example of the most current exciting breakthrough in artificial intelligence?

Demis Hassabis3:43

The most exciting breakthrough, I think, from our perspective is combining two different areas of artificial intelligence together. There's one area called deep learning, which uses neural networks that are kind of mimic what the brain does.

Presenter3:54

So the other pathways.

Demis Hassabis3:55

Yeah, all the pathways that form, these big networks of pathways, and they're now got amazingly powerful. So you can do object recognition, you can do speech recognition.

Presenter4:05

So, for example, this is when, you know, thirty, forty years ago, a computer would not have been able to look at a cat and say that's a cat.

Demis Hassabis4:10

Exactly. Whereas now it can. Where now it's now a cat, exactly, from just being given a few hundred examples of what cats look like. And then the second part, reinforcement learning, which we specialize in, is about decision making. So given your model of the world out there, your experiences, you build a model of the world, how it works, and then how do you use that model to make decisions? So in a computer game, which is what we test our algorithms on, it will be getting the highest score you can get. And these two systems together, which is our big innovation, gives you, in some sense, the rudimentary beginnings of a full intelligence. So in essence, we're still at the first steps of this. That's sort of the whole cycle that you need for intelligence.

Presenter4:10

Uh

Presenter4:50

I don't know how much you had to dumb that down, but I actually understood it very clearly. Thank you so much. We're going to come back to the concerns, the opportunities, the worries that people have later. But for now, let's have your first piece of music then, Demos Hesalpis. Tell me about this. What are we going to hear?

Demis Hassabis5:03

So the film Blade Rhino, obviously it's about artificial intelligence and it's about robotics. And I just love everything about this film and it was very formative for me when I saw it as a teenager because it sort of brought artificial intelligence to life in this really beautiful visual world, sumptuous visual world, I would say. And I love that speech that he gives at the end. He's talking about a couple of very emotive memories that he's seen that no other humans have seen because he's been traveling the stars. And he talks about those moments being lost forever, like tears in rain. I also love the Van Gellish soundtrack. For me, it was one of the big inspirations, you know, to think about spending my career on artificial intelligence and, you know, trying to make some of those things reality.

Speaker 35:45

I've

Speaker 35:46

See the

Speaker 35:49

Wouldn't make it.

Speaker 35:53

Attack ships on fire off the shore of a light.

Speaker 36:00

I watched C D's.

Speaker 36:03

Glitter in the dark meeting ten hours a game.

Speaker 36:11

Oh no.

Speaker 36:13

Moments will be lost.

Speaker 36:16

In time.

Speaker 36:20

Bye.

Speaker 36:22

Yeah.

Presenter6:28

Tears in Rain. That was uh Van Gelis' tune from the soundtrack to a film Blade Runner. Demos Hassabas Whenever I read the things that you've said about your work into AI, y you make it sound a lot more

Presenter6:40

well, palatable and I think touchy feely and warm and round edged th than I'm used to. Just explain a bit more about that. You know, there seems to be a bit of a disconnect between those of you who are working

Demis Hassabis6:48

Yeah.

Speaker 46:48

Just to

Demis Hassabis6:48

Yeah.

Presenter6:53

In artificial intelligence, and those of us on the outside who think that scares the heck out of me.

Demis Hassabis6:59

Well, if you think about it, intelligence is this amazing power in the universe. You know, it's almost like a force. And I just think it's an incredible thing that we have these minds that can invent and create. So for me, as an engineer, trying to break that down and understand the processes and the mechanisms that are involved in that is the most fascinating thing you can work on.

Presenter7:20

And you think that this is all computable? You think that this is it is all possible to extrapolate this through, well, numbers, I guess.

Demis Hassabis7:27

Yeah, I have a very open mind about that. I mean, you know, different great scientists have said different things about this. You know, Roger Penrose, for example, the great mathematician, you know, he thinks there's some quantum effects in the brain. So in which case you wouldn't be able to compute it on a classical computing device. You'd need a quantum computer.

Presenter7:42

Where do you stand?

Demis Hassabis7:43

So the biology and the neuroscience shows that there isn't something mysterious like quantum mechanics going on in the brain. To do science properly you always have to keep an open mind that you're going to discover something new or someone else in a related field will discover something new that changes your opinion. But for now we're operating under the assumption that you can compute everything that's going on in the brain on a normal computer.

Presenter8:04

And the very name of the company, Deep Mind, of course, gives us access to this thing that can be tricky to understand, which is creativity and consciousness are.

Demis Hassabis8:09

No quit

Demis Hassabis8:12

Yeah.

Presenter8:13

Can I use the word hidden in those parts of the brain that until now we have no access to?

Demis Hassabis8:19

Yes, I think we can say hidden or maybe emergent. So these are deep mysteries, I think, that go to the core of what makes us human. And I think we are maybe on the cusp of starting to understand what some of these things might be.

Presenter8:32

Could you explain for me things that have been understood maybe even in the last year to two years that up until now we thought would be impossible?

Demis Hassabis8:34

Single

Demis Hassabis8:37

Yeah.

Demis Hassabis8:40

Definitely. So I think there are some theories, for example, on what dreaming is. Dreams have been a mystery for thousands of years, I think. And, you know, it turns out that dreaming may be your part of the brain called the hippocampus replaying memories that you've had recently that are very emotional, that have been recorded in your mind, and then replaying them back at much faster rates while you're asleep. So your brain, the rest of your brain, gets to learn about that event, you know, what to do or what not to do from hundreds of examples, even though you only experienced it once in real life.

Presenter9:11

Let's have some more music, uh Demosalvist. What are we gonna hear?

Demis Hassabis9:14

So I picked a piece by Enya called Watermark and I picked it because it's the first C D that I ever bought. I think I was around 12, 13 years old. And later when I was a little bit older, around 15 and I taught myself how to play piano as a hobby. This was one of the first pieces I taught myself. And the thing I like about it is it sounds really beautiful, but actually technically it's quite easy to play.

Presenter10:10

Enya and Watermark. Demas Hasabas, your parents produced three very interesting children. Y your sister is a composer and a musician. Uh your brother is a professional poker player and then, as we know, there is you. I mean, tell me a bit about the upbringing that produced these three fascinating kids.

Demis Hassabis10:27

Yeah, my parents are very interesting actually. I would describe them in one word probably as quite bohemian. And my father was a singer-songwriter when he was young, but he did a lot of different creative jobs. And at one point, they ran a toy shop, and they followed their interests. And that stuck with probably all three of us. You don't have to be constricted by social norms. You can actually follow your own creative path and really deeply explore that. And it's fine to do that. In fact, that's the way you should live life.

Presenter10:31

Rice

Presenter10:56

You said he ran at times a toy shop. Is it true that littering the house on occasion were were toys that weren't quite complete?

Demis Hassabis11:02

That's right, that's actually gave me my first sort of game design practice because yeah, we would used to get all the boxes and the sets of games that had pieces missing and they couldn't sell. So I would take them. Often they would have the rules missing, so I'd just invent something and then I would play it with my brother and sister, though who are younger than me, and they used to be my games testers. And then I would refine it. And I think that's how I learnt game design.

Presenter11:23

You started playing chess by the age of four, and you were competing by the age of six, and fascinatingly, it was at a tournament in, I think, Liechtenstein when you were eleven that something

Presenter11:36

Well, terribly progressive and impressive.

Demis Hassabis11:39

Occurred to you. Yeah, so I was at this international tournament in Liechtenstein, and we were in this huge hall, you know, hundreds of international chess players. And I was playing, I think at that point, he was the current Danish champion. So I was there, and I was 11 years old, and we were into our 10th hour of the game. And we were in this incredibly unusual ending where I had a king and queen, and he had a king, a rook, a bishop, and a knight. And he had the advantage, but it should have been a draw. And we spent literally four hours with him trying to outmaneuver me. And finally, he tried one last cheap trick, which was that he trapped my king. And all I had to do was give away my queen, and then it would be stalemate and it would be a draw. But I was so tired, I didn't think about giving my queen away. And so I resigned, unbelievably, on the final moment, because I thought it was inevitable I was going to be checkmated. And he just got up, stood up, and I remember this very well. And he went, Why have you resigned? It's a draw. And he immediately, with a flourish, sort of showed me the drawing move. And I was thinking later, as I got older, and even at the time, I was thinking that's a little bit harsh on an 11-year-old. You know, he was a grown man. It just felt sort of wrong somehow. And then it made me think the rest of that tournament: it's like, are we wasting our minds? You know, at that level of chess, they're all fantastically smart people. What if we used that brain power for something more useful, like solving cancer or curing some disease? Wouldn't that be a better use?

Presenter13:03

That's what I'm saying.

Demis Hassabis13:04

That's 11. Yes, I thought that. It was a really epiphany moment for me.

Presenter13:08

And your father by that point had you

Presenter13:11

Been taken out of school to be home schooled so you could concentrate on your chest.

Demis Hassabis13:14

So if you could

Presenter13:16

Did you have a conversation at eleven with your parents and say, This is not for me?

Demis Hassabis13:18

This is not for me. I did. They were pretty shocked because at the time I was the second highest rated chess player in the world for my age and everyone was assuming that's what I was going to do. And certainly until then, that's been my whole life. For example, I can't even remember me doing any schoolwork because it was all chess, although computers were in the background. So yeah, it was a big thing for me to say that at 11.

Presenter13:20

Yeah.

Presenter13:41

More music. Now, Demis, we're going to hear your third. Tell me about this.

Demis Hassabis13:45

So I picked this track from the KLF from this album, The White Room. I can't really say I was a rebellious teenager, but that was about as rebellious as I got as the KLF and I just love their sort of anarchic mentality. And I just remembered listening to The White Room on cassette tape, sort of non-stop, while I was studying for my A-level. So it was an important time for me.

Speaker 414:03

They're justified and they're ancient And they like to roam the land They're justified and they're Asian I hope you understand

Speaker 414:22

Up in Tennessee, next at Tenny, stand by the jams. But if you don't like what they're going to do, better not stop them, cause they're coming through.

Presenter14:40

That was the KLF featuring Tammy Wynnette and Justified and Ancient. Demos Hassabas, you used your chess winnings. I imagine you'd got a treasure chest, as it were, from those. When you were a kid to buy your first PC, it was a ZX Spectrum. You were about eight? Yeah, eight years old, yeah. And you yourself have quoted the physicist Richard Feynman saying, What I cannot build, I cannot understand. So is that what propelled you to try to understand code?

Demis Hassabis15:06

Yeah, and I bought some books on programming and I started programming this thing and I just realized that this was an incredible tool. And I've always loved the fact that you can set your computer off, you know, late at night, doing some calculation, you can go to sleep, and then this tool carries on working for you. And then you get up and then it's solved. And for me, that's just magical. It's almost like offloading, enhancing your own mind. And then later on, my next computer, which was a Commodore Miga 500, I had a couple of school friends, and we used to have, you know, I guess you could call it a hacking club where we used to try and outdo each other, writing an assembler code, making demos of cool visuals, maybe replicating films we'd seen or little games. And all my spare time was figuring out how to get better at programming these computers.

Presenter15:54

Did you feel like you were in on the big secret?

Demis Hassabis15:56

I did, actually. It did feel like that. There there was this fascinating netherworld that almost no one knew about. So it was it had that element to it.

Presenter16:04

You sat your A levels a couple of years early. You did uh it was double maths, physics, chemistry. You were offered a place at Cambridge, but they told you you're you're too early to take the place, but you have the place. And you ended up in that period working for a computer games firm. What what did you do for them?

Demis Hassabis16:13

Yeah.

Demis Hassabis16:20

Yeah, so actually I won a sort of national competition, which was the prize was that you won a job at this company called a company called Bullfog Productions, which at the time was run by a very famous game designer called Peter Molyneux. And one of the games that I obsessed with when I was 12, 13 is a game called Populous, which is this huge simulation game. And they'd made that game. So I went down there the day after my A-levels and they were pretty surprised because obviously they were expecting a graduate and they couldn't even legally employ me. So I was ended up being paid in, you know, in cash to pay for my YMCA hostel that I was staying in. But we ended up writing some really famous games. So a game called Theme Park, I ended up being the lead programmer and co-designer of, which went on to sell millions of copies.

Presenter16:28

Who's company?

Presenter17:05

And you you say you're living in a s a little hostel and your parents were cool with all of that.

Demis Hassabis17:08

Yeah, they didn't really know what was going on. So so by this time it was just so alien to them I just, you know, disappeared off to Guildford on the train and I don't think they had any idea what I was doing. Tell me about your next piece of music, Then's Alvis. What are we gonna

Presenter17:20

I got

Demis Hassabis17:20

Uh

Presenter17:20

Here.

Demis Hassabis17:21

Prodigy. So, yes, when I got to Cambridge, I had an amazing three years. Honestly, it was the best three years of my life because I think for me, it was like a holiday camp because I'd basically been working my whole life, you know, professional chess. I never had a summer holiday because we were always traveling around the world, which was great, but I never had a normal summer holiday. And then, in my years out, instead of doing the normal travel around South America, you know, on a backpack trip, which I do regret not doing that, I was working literally from the day after my A-look exam to the day I got to Cambridge. So, I was determined at Cambridge to have, you know, a great social time as well as learning a lot of things. And for me, the Prodigy album came out in 1994, and I was listening to that all the time in my room, lying on my bed with the morning light streaming in after a whole night of going out and listening to these tracks.

Presenter18:42

That was the prodigy and skylined. I'm delighted to hear, Demisisabis, that you had uh you had fun at university. I mean, even though you got a double first, y you really you you partied. Yeah. Did you ever endanger the brain? I mean that's pretty trippy.

Demis Hassabis18:54

I did actually. So that was a good life lesson for me. So in my first year at Cambridge, I really did let loose. And I had an amazing group of friends that are still my best friends today.

Presenter19:04

Um you joined another uh computer games company after you graduated and you set up your own design uh studios to market your own creations. And I I hate to say this, I almost blush. I have never played a computer game in my life. What am I missing?

Demis Hassabis19:18

Well, I think computer games, the best ones, are a great microcosm of some aspect of life. I think you can learn a lot from them. You know, in business or in life, there are not many times you can practice in a safe way ideas or strategies or, you know, develop your own mind. So I find games, it's a little bit like a gym for the mind. So I've used games, board games and computer games, to train my own mind and train myself into particular directions. And I pick games that would accentuate that.

Presenter19:45

Let's talk then on that subject about something called uh the Mind Sports Olympiad. Now you have won that a record five times between nineteen ninety eight and two thousand three. And it's a sort of festival for for mental games, really. We see Olympians, we see people use their bodies. And we can although we know we're never going to do it ourselves, we understand it. We see the strain, we see the sinews pulsing, we see the sweat. What's the sensation?

Demis Hassabis20:12

Yeah.

Presenter20:12

of doing Olympic

Demis Hassabis20:14

Pick feats with the mind. I think it probably has a very similar feeling, I'd imagine, in the sense: okay, it's not physical, but you have this same joy of intense mastery and reveling in using your talents and being able to demonstrate that. And it gives you a really meta-understanding about what it takes to be the best at something. And that's one of the many lessons I've taken away from my games career: that feeling and being able to assess very quickly where you measure up and what would be required to go further. Certainly, in my job now, skills to have when I assess projects or people or the difficulty of something. I have all of those things in mind. And, you know, the other thing about games is, and what I've used, I think games like poker and chess and other things is you learn a lot about yourself and how you operate in the best way. And I think all of those things are very transferable.

Presenter21:09

Let's have some more music, Adams. Tell me about what we're going to hear now. We are on your fifth disc.

Demis Hassabis21:14

So I've chosen a piece by Michael Nyman who's one of my favourite sort of modern classical composers I guess you could call him and this particular piece that I listened to a lot when I was programming my first big game and really for me Theme Park was my gateway into my professional games career and I listened to this particular track a lot during the very creative sort of couple of months when we were sort of designing the game.

Presenter22:14

Michael Nyman with The Garden is Becoming a Rob Room from the film The Draftsman's contract. Demis Assabis. By twenty ten you had founded your company. You did that with your partners Mustafa Suleiman and Shane Legg. And I need to leap forward because there's so much to fit in. But three years later, after a lot of hard work,

Presenter22:32

You presented what seemed to the world this hugely significant step. You announced that the company had developed a machine that could become an expert gamer with and here's the key the instinct to score highly. You had not taught it to play the game, but you had given it the tools with which to work out for itself how to play the game. How had the team cracked it?

Demis Hassabis22:57

Yeah, these were classic Atari games from the 70s and 80s and all we gave the system was the pixels on the screen and then the goal to maximize the score and everything else it learns for itself from first principles through trying out the game and experiencing the game itself. So it was learning for itself and that was our big breakthrough was really showing that a system like that could master something that was challenging even for a human expert.

Presenter23:21

And that's what makes somebody like Stephen Hawking say AI could be the biggest event in human history. Unfortunately, it may also be the last. Because when we give the computers the instinct to not just think what we might be thinking, but to think ahead of us, then surely they're in control.

Demis Hassabis23:38

Well, I think there's a lot of research left to go, but we have to think about what goals we give these systems, what values we give these systems, and how we make sure that they stick to the goals that we give them. And the way I think about AI is as this amazing tool that we can use to enhance our own goals as humans.

Presenter23:58

But you yourself have even been surprised. You had come up with this AlphaGo computer programme. You watched it as it beat Lisa Doll, one of the world's leading players of this ancient game. Now, just to be clear, I mean a typical chess game would have sort of thirty-five possible moves. A typical Go game has how many? Has two hundred. And so as you watched it beat him, you were astonished at what this machine was doing.

Demis Hassabis24:15

Possible.

Demis Hassabis24:19

And so

Demis Hassabis24:24

Yes, we were astonished by the capability the machine had and our program we created, it actually came up with its own ideas and its own motifs that even stunned the Go world. But we weren't surprised that it could play Go well, we were surprised that it could create its own moves and we were delighted by what it created within the domain of what we'd built it to do, which was to play Go.

Presenter24:43

Right, which is why you believe that there will be a benevolent intent because you will build that in.

Demis Hassabis24:48

Yes. In the course of the next few decades, as these systems develop, we will understand better what kind of control systems are required, how to check and interpret what these systems are doing. And I think we will solve those problems on the way to building AI.

Presenter25:01

I mean what about these programmes being developed by people who have not our best but our worst interests at heart?

Demis Hassabis25:08

Yeah, I think one good thing about it at the moment is that these are incredibly tough systems to build and incredibly complicated. There's only a few hundred people in the world that can build these things. And most of them know each other as well. So we're quite a small community. And I couldn't do it now on my own. We have 400 people, 250 PhDs from the top places in the world coming together. And even then, we're only making small progress towards this. That's one defense. The other thing is, at some point, what we hope is that once we understand these systems better and understand what we would be legislating for, there would be some stronger form of governance that would be agreed by world governments.

Presenter25:46

I can breathe out then. Demos, tell me about your next piece of music. What are we going to hear? What's this?

Demis Hassabis25:49

So I've chosen a track from Queen Who Wants to Live Forever. I think Freddie Mercury is one of the greatest vocalists of all time. And this particular track I love because it just speaks to the briefness of life and the fragility of it. And I really feel that, and I've always felt that, and how you've got to make the best use of your time because it's just so short.

Speaker 426:09

Who wants to live forever?

Speaker 426:40

Must I

Presenter26:42

That was Queen Anne who wants to live forever. And Demas Hasabis, you were pondering on uh

Presenter26:48

There.

Presenter26:49

On mortality and the searing proximity of death. I'm wondering if that might be why when Google came a knocking in 2014 and said we want to buy your company, you said yes, because of course to be a founder of a company is a precious thing. To have seen it from its infancy, to being as it has been a pioneering and significant world presence in artificial intelligence. Why did you say yes to Google?

Demis Hassabis27:12

I said yes, actually, precisely because of that reason. You know, I did a year of due diligence on them and how it was going to work post-the acquisition. And for me, it was nothing to do with the money. In fact, my investors, you know, mostly didn't want to sell, even though it was a lot of money. It's because I was convinced that by joining forces with Google, we could accelerate the progress of the mission. You know, by using Google's power and all their resources, we could explore this space of algorithms much faster. I could hire more brilliant research scientists, and the whole kind of mission and the whole research program would accelerate. And that's what's happened over the last three years. So it's worked.

Presenter27:46

Artificial intelligence and all its associated technology does of course have the ability to threaten lots of traditional jobs. Once AI has got it covered, then we won't need human beings to do it. You've said you think that there is a huge potentiality that is offered by that, whereas a lot of us just see a lot of unemployment. Can you give me an example of where you think inroads are significantly being made?

Demis Hassabis27:54

Holmes

Demis Hassabis28:09

Yeah, so I can tell you right now. So, for example, we are using variants of the algorithms we use to crack the game of Go. One in the medical area is we're looking at the problem of protein folding. So, in biology, one of the biggest issues is how do proteins in the body fold into the 3D structure? Because the 3D structure determines how the protein is going to react and act in your body. And things like Alzheimer's, for example, may be because proteins have folded in the wrong way. Now, what biologists dream of in pharmaceutical companies is if you could take an amino acid sequence and just tell straight away from the genetic sequence how the 3D structure is going to look, just from that, then you would accelerate drug design by five or ten years. So, at the moment, it's laborious and it takes years to find out the structure of these things. And we're applying AlphaGo type algorithms to a simulation of these proteins, and it's working out for itself how to fold the protein. Now, we're not there yet, we haven't cracked it yet, but we've got some promising initial results. And I think in another year or two, maybe we'll be at the level where it would be useful for drug discovery companies to use it as a virtual way of testing out different designs of drugs to target certain proteins. And that could then help with all sorts of diseases.

Presenter28:26

It's

Presenter28:47

Just

Presenter29:24

What a great description. Thank you for that. We're going to go to your penultimate track then.

Demis Hassabis29:28

So I've chosen a track by Duff Punk and I really like their most recent album, Random Access Memories. I think it's about an AI or about computers or a robot maybe gaining consciousness or something like that. And talking about how quick life is going, I can't really believe, you know, I was 40 just last year. And look.

Presenter29:45

What? If done

Demis Hassabis29:46

Was it life?

Presenter29:46

Was it nice?

Demis Hassabis29:47

And I actually used it as the soundtrack to my fortieth birthday.

Presenter30:24

That was doved punk and motherboard. Demosabis, you mentioned there, and you gave a beautiful description of how AI might help in something as important as Alzheimer's research. And it suddenly occurred to me, of course, that y your wife is involved. She is a fellow scientist, she's involved in Alzheimer research. Do you discuss work together?

Demis Hassabis30:41

Yes, we do. We often talk about how could that process of discovery be made more efficient? And what part could AI play in that to help us accelerate the discoveries which are badly needed with aging population and the amount it costs to look after these people and the tragedy of that.

Presenter30:58

I wonder if you have advice for the parent you're you're a parent yourself, if you have advice for the parents of very uh uh prodigious, smart, advanced kids, because you should be a pain in the neck, and you're not.

Presenter31:14

I sit here with you today, and you're such a sort of warm, open, thoughtful, you lack any kind of condescension, you are very open to the idea of communicating with the rest of us.

Presenter31:24

How did that come about and what do you think parents should do to make very bright young kids fit into the world?

Demis Hassabis31:29

I think that what I would do is encourage them to explore heavily when they're young and really get a wide range of experiences. So, encourage deepness and expertise in things, but not at the expense of everything else. Life's so rich, you should partake in all of it, right? So, I think the kind of skills I would teach children today are the ability to learn rather than the specific thing that you're learning. One of the first things you should learn about, which is not taught at schools, is about yourself. How do you work best? What do you want? What's your dreams? What are you excited about? And I think you have to explore that and find that out for yourself.

Presenter32:02

And do you have concerns, as indeed we heard President Obama say in his final speech about this individualization of culture, that as our computers and as things like AI begin to allow us to live the lives that are entirely adapted and defined by us, what we lose is the connection.

Demis Hassabis32:20

I agree. And I think that's been the case always with any kind of technology. Maybe it's accentuated today with our current technology, mobile, internet, you know, AI, things like this. But I think it's important to be, again, very broad in your upbringing so that you're used to understanding different points of view, different things, and actually going and seeking that, right? And enjoying that and learning how to deal with that sort of disagreement in a constructive way.

Presenter32:46

Damn us, if we could only talk all day, but we've got to

Demis Hassabis32:48

Uh

Presenter32:48

But in it's your final disc? Tell me about this final eighth disc.

Demis Hassabis32:48

It's your favorite.

Demis Hassabis32:51

So, I've picked the soundtrack from the film Interstellar, which I think is an amazing film. And, you know, I love Chris Nolan as a director and a friend of mine now. And he works with Hansimmer a lot. And I just love all of Hans Zimmer's soundtracks. And I picked Interstellar because I think it relates to ultimately what I'm trying to do in this world, which is to understand the universe around us. And obviously, Interstellar was all about that, understanding time and black holes and really our place in that universe. And you know, that's what I hope to use AI for in the future.

Presenter33:54

That was Hans Zimmer with part of the soundtrack to the film Interstellar. Demis, it's time now for me to give you some books. I give you the Bible and the complete works of Shakespeare. And what else are you going to take?

Demis Hassabis34:05

It's a very difficult one actually, but I think I'll take The Lord of the Rings because I read that several times when I was nine or something like that. And I think it's such a beautiful world he created. I think being alone on the island, it would be maybe nice to have a very rich, deep, fantastical world to lose yourself in.

Presenter34:23

Yeah.

Demis Hassabis34:23

It's yours then. A a luxury too. I was gonna bring a solar powered chess computer so that I could play lots of games of chess and maybe if I ever got rescued from the island, uh, you know, maybe I'd be Grand Master level when I left.

Presenter34:35

Don't get any sand in it then. You can have that. Which of the eight tracks would you like to save?

Demis Hassabis34:41

I'd probably save the Vangelis track because I could mile over those lines. I think they're they're beautiful poetry for for a very long time and sort of find renewed significance in it.

Presenter34:51

It's you, Estemo Sasabis. Thank you very much for letting us hear your desert island discs. Thanks.

Demis Hassabis34:55

For having me.

Presenter34:56

Yeah.

Presenter35:40

You've been listening to a download from the BBC. You'll find more information on the Radio 4 website: bbc.co.uk slash Radio 4.

Demis Hassabis36:13

This is the BBC.