The Many Faces of AI: Sebastian Mallaby on Demis Hassabis and the Quest to Read God’s Mind

00:00:31 Andrew Keen: Hello, everybody. It is Thursday, April 9th, 2026. Year later than I thought. Interesting piece in the New Yorker this week on Sam Altman. He may control our future, but can he be trusted? And the image of Sam is of many faces of Sam. And, of course, this question of trust and the future in Silicon Valley is one that we often pursue on this show. In comparison with Sam Altman, another of the major players in our AI game is Demis Hassabis, the head of AI at Google. Perhaps the adult in the room, and whatever one thinks of Demis, I think it would be hard to argue that he carries as many faces in his pocket as Sam Altman. My guest today, has a new book out about not Sam Altman, but Demis Hassabis. Sebastian Mallaby is one of, Britain's most distinguished journalists, worked for many years for The Economist, written many books on venture capital, and now has this major biography of Demis Hassabis. He's talking to us from a hotel in Downtown New York. Sebastian, congratulations on the new book. Would it be fair to say that it would be hard to imagine Demis carrying lots of faces around in his pocket a la Sam Altman?

00:02:00 Sebastian Mallaby: I think Sam Altman is, sort of almost uniquely, duplicitous. I mean, Silicon Valley has quite a lot of fake it till you make it. But Sam Altman really does seem to have managed to annoy everybody he's worked with, by saying one thing and then doing the opposite. Demis Hassabis, on the other hand, you know, has had a much more stable career path. You know, he founded DeepMind back in 2010, sold it to Google in 2014, and he stayed there running DeepMind, continuously for sixteen years. Whereas, you know, Sam has jumped around a bit and considered things like running for governor of California. So I think, you know, certainly, Demis is more consistent. He's motivated by something a bit different. He doesn't just want power. He wants scientific enlightenment. And I say enlightenment advisedly. It really is a sort of a spiritual quest for him. And so, yeah, I think they are different.

00:02:56 Andrew Keen: You've written, as I said, a number of books, mostly on money, venture capital. You seem to have wrestled, morally at least, with this decision to write this book. It took you, what, three years. You had to get Demis' permission. What were the moral questions, for you personally, Sebastian, that you wrestled with, in terms of writing this biography of Demis?

00:03:23 Sebastian Mallaby: Well, I I kind of came in with the idea that creating really powerful AI was sort of the twenty first century version of the Manhattan Project, that, you know, this was dangerous stuff, that the AI inventors of today have their hands on the modern version of the nuclear material. And, you know, in some ways, it's an amazing scientific feat to bring this into the world. At the same time, it could be dangerous. And I wanted to portray, a thoughtful person wrestling with a reality that he could be doing something really bad. And what did it feel like? I mean, you know, sort of how do you sleep? How do you explain your motivations to the world and to yourself? And so those were some of the kind of wrestling, that that went on. It was less about me than about the subject I was grappling with.

00:04:12 Andrew Keen: So there's a bit of the Oppenheimer about Demis, do you think? This this guy who perhaps didn't sleep very well, or was that more in your mind? I mean, I know in, the movie The Thinking Game, which of course isn't Oppenheimer, it's a documentary in comparison to Oppenheimer, which was fiction, and was major Hollywood fiction. Demis is portrayed as a very down to earth kind of character, very ordinary, perhaps not just surprisingly ordinary, but in a weird way, disappointingly ordinary.

00:04:45 Sebastian Mallaby: Yeah. You know, he's a funny mixture because he's ordinary seeming in the way he holds himself, the way he delivers his message. You know, he he he might be up on the stage, and the tone of the delivery is sort of, hey. What about if we wash the dishes, you know, after lunch? You know, this kind of boyish figure in a in a round neck, shirt and sneakers. And, you know, it's not like Elon Musk where you can't count the number of children. There's a guy who married his Cambridge girlfriend and has two kids and, you know, lives in the same house continuously. And, you know, it's just it's just a very different vibe. But what is not normal and what is not at all disappointing is that in this kind of guy next door package, you get an outright genius. I mean, he he can riff on everything from neuroscience to computer science to the history of movies to philosophy to, you know, business theory and how you build companies. I mean, he can do it all. And, and the range of his thinking, as I sat there in a pub in North London talking to him for hours and hours and hours, was, like, unlike anything I'd ever experienced. I mean, I've

00:05:57 Andrew Keen: Which pub was it? Where did you sit? And where does he live?

00:06:00 Sebastian Mallaby: So he lives in North London in

00:06:02 Andrew Keen: Whereabouts? I mean, I don't have street, but what neighborhood?

00:06:05 Sebastian Mallaby: Highgate. He lives in Highgate in North London.

00:06:09 Andrew Keen: And Not the fanciest neighborhood?

00:06:11 Sebastian Mallaby: Not the fanciest neighborhood. I mean, I think he has a nice house, but not a kind of you know, for somebody who presumably must be worth, you know, 500,000,000 or something, given the way that Google's stock price has performed and Demis' position in Google, you know, having sold his company to Google. You know, he he he's got plenty of cash, but he's not spending it, as you can as you can sort of infer from from where he lives and how he lives. I once said to him, you know, it's okay. How do you spend your money? Do you have a second home? He said, no. No. No. I don't want that. Well, what kind of car do you drive? Can't remember what it's called, but it's more than ten years old. My wife has it. Well, do you have any hobbies that cost a lot of money? Well, I did buy season tickets to Liverpool. How much does that cost? Well, I think it was sort of 2 and a half thousand pounds. You know, I mean, it it there there isn't a lot of, you know, taste for for luxury, in the package. So it this is the ordinary side, but the, less ordinary thing is just this intellectual range. And, you know, so we would meet in this pub. As I was saying, it's nearest to his house in Highgate. There was a dusty staircase at the back of the pub, which people didn't know about. So if we went up there to the upstairs room, there'd be nobody disturbing us, sit there for two hours. Cumulatively, it was more than thirty hours, of discussions. And, you know, he would just to give you an example, and going back slightly to Oppenheimer, you know, I asked him one day what it was what it felt like to set up DeepMind, raise the money from Peter Thiel. It's 2010, Demis. You've got your first office. It's in Russell Square in London. What did that feel like? And I know from experience that when you ask people to reconstruct the emotion of something that happened fifteen years before, the answer you normally get is, yeah. It was cool. I mean, people don't don't recall the feeling very well. But Demis is so articulate and vivid that he would say, yeah. You know? So we were in this attic office in Russell Square. It was kinda hot because the sun was beating down, and we it wasn't very well insulated. But, you know, not again, I get a breather, and so I would I would go down the stairs. Of course, no lift. It was a very old building. So ding, ding, ding, ding, ding. I go down the stairs and outside in front of me, there is the the greenery of Russell Square. So that's nice. But what's really good is on the right, three doors down, there was the London Mathematical Society where Turing Turing, the father of British computer science, he gave his lectures there. And so now we're completing his mission as by building AI. And beyond that, do you know what there is, Sebastian? There is a pedestrian crossing. It goes black, white, black, white, black, white. And as somebody in the 1930s do you know who it was? It was the Hungarian nuclear scientist, Szilard. He was crossing that pedestrian crossing. He had the idea for a nuclear chain reaction, which led to the Manhattan Project, Sebastian. And it's perfect because we're doing the modern version of the Manhattan Project. So his storytelling is off the charts.

00:09:10 Andrew Keen: Yeah. Maybe you should write a novel. The the Guardian review was a little I'm sure you've read it, Sebastian. Probably annoyed you.

00:09:19 Sebastian Mallaby: I did. Yeah.

00:09:20 Andrew Keen: It it it it it it does focus on this, and it suggests, that he's not as interesting as as you think. I mean, I don't know how the Guardian guy knows because I I don't suppose he spent hours in a Highgate pub with, Demis. But, how would you respond to that kind of criticism? I mean, obviously, the zeitgeist these days is a hostility towards, major techno technologists like Demis. What's your thoughts on this dismissal of, Demis as, a major intellect? I mean, obviously, he's very smart on the technology front. What what what what

00:09:59 Sebastian Mallaby: Yeah. Look. I think you said it yourself. I think people just are very, very skeptical of technology leaders. And, you know, if you look at that New Yorker profile of Sam Altman, you can see why they are skeptical of technology leaders. Right? It's not a crazy position to take. My point is truthful. You know, there can be differences between individuals. Just because some guy in Silicon Valley, may be a sociopath, it doesn't mean that his competitor in London is the same. No. They're different. Demis is different. You know, one has a Nobel Prize, and the other one dropped out of Stanford. One is a deep scientist. The other one isn't a scientist at all. You know? One it's just you can go on. I mean, it's it's so different that I think tarring Demis Hassabis with the prejudices one absorbs by reading about Elon Musk or Sam Altman and applying those to Demis' case would just be totally wrong. And so I didn't do that. I learned a lot about him because I spoke to a 100 people in his entourage, people who knew him when he was at Cambridge, people who knew him when he was building his first country you know, company. I talked to people who had fallen out with him, like Mustafa Suleiman, the cofounder of DeepMind. You know, I really did a three sixty. And And at the end of the three sixty, at the end of more than thirty hours with him and all these hours with other people, I can tell you he's a sincerely good person. And I I know this, from from so what am I supposed to do? Am I gonna say, well, he's just a tech leader, though he's a bad guy? I'm not gonna do that. And the and the serious other reason why I shouldn't do that is that look. There are two categories of stuff that goes wrong in the world. Right? There's one category where you have some idiot in control who makes a bad decision because they're either an idiot or they're just evil. And, you know, the Iran war is a very good example of an idiot making a stupid decision to go to war. Right? And so if you had switched out the person who happened to be the president, you wouldn't have had the Iran war, and we would be in much better shape. And there's a second category of screw up, which is much, much more interesting, where you have a good person who is smart and trying to do their best, and yet ineluctable larger forces lead them into a position where things go wrong. And that is the story with Demis Hassabis, that he is trying to make AI safe, but larger forces having to do with the race between China and The US and multiple labs competing against each other makes it very hard to deliver safety. And so, you know, the problem with the Guardian kind of position is not merely that it's wrong about Demis, but also that it obscures a much bigger structural thing that we should all be worried about, that this is larger than individuals. The reason why AI is going wrong right now is larger than individuals. It's about race dynamics and the failure of governments to intervene to prevent those race dynamics from causing a lot of danger.

00:12:58 Andrew Keen: The other thing The Guardian picked up on was in the in the book. You said, for every Donald Trump supporter in the tech industry, there's also a Bill Gates. I mean, I assume you wrote that before the whole Epstein thing broke. Are you a 100% confident that there's no backstory to Hassabis, that there are no scandals, hanging around that may come out later?

00:13:20 Sebastian Mallaby: Yeah. Look. I mean, I'm pretty confident. I, as I say, you know, I spoke to people who knew him on the chess circuit when he was a kid, people who knew him at Cambridge, people who said to me honestly that working with Demis was too much for them. He's too intense. And, you know, one guy had a nervous breakdown, had to go spend a year recuperating. So and and and then, you know, there are people, like Mustafa Suleiman who worked with him and got kicked out and had plenty of negative things to say about him.

00:13:54 Andrew Keen: What what is the the Mustafa Suleiman story? We had him on the show. He wrote a book a year or two ago, which was kind of critical of technology. It was a bit of a, I won't say a me too book because that brings the wrong connotations, but a book that wasn't particularly original. I know that Suleiman got thrown out of Google when, Hassabis and Suleiman came over. Now he runs AI at, at, Microsoft. These always seem to me as an outsider. You know these guys a million times better than I do. A couple of North London lads founding a a tech company. Why did they fall out? And and is Suleiman the the the evil face of, or maybe the real world face of, of Demis?

00:14:48 Sebastian Mallaby: You know, they had it's a very interesting story. They had backgrounds in North London, both of them. Suleyman is eight years younger than Demis, but he did go to the same school that Demis attended for a bit, a kind of good, government school. But it was competitive entry kind of grammar school.

00:15:05 Andrew Keen: And you, of course, Sebastian, went to Eton, which is anything but a government school.

00:15:09 Sebastian Mallaby: Totally different. Yep. That's right. So these people, I admire them because they had a lot more adversity in their background than I did. So they they both rise up, from modest circumstances, but Suleiman's is notably tougher than Demis's. Right? I mean, Mustafa Suleiman's father was an immigrant from Syria who drove a cab, struggled with English, was sort of semi literate. There was no books in the home, no music. You go to the mosque every Friday. It was devout, strict, and not enlightened. And then his parents split up, when Suleiman was 15, and they leave. One of them goes with a new partner to New Zealand, and then the dad, who's from Syria, goes back to Syria, leaving Mustafa at 15 to look after himself and his younger brother, who's 14, and survive in London. Right? This is seriously wild story. And, you know, despite that adversity, Mustafa, who is extremely smart, gets into Oxford. He hates Oxford because he feels out of sort because it's just culturally alien to him. So he drops out to to to to found a thing called the the Muslim youth helpline to prevent suicide amongst young Muslims in the time of Islamophobia on the back of the Iraq war. This is extraordinary background for somebody to then be a cofounder of a tech company. So, I mean, Demis is amazing, and, you know, Demis is the is the scientist of the two, And and Mustafa is on a on a different level in terms of his path and his and his sort of scientific foundation and training. But don't underestimate Mustafa's smartness and his drive and his determination.

00:17:01 Andrew Keen: Did Demis have anything to do with Mustafa getting fired at Google?

00:17:06 Sebastian Mallaby: It's a good question. I mean, Demis would and did argue to me at huge length that he didn't do it. It wasn't him. And, you know, basically, there was a inquiry into alleged bullying by Mustafa Suleiman, and the legal counsel took care of this, and it ended up in him leaving. This is nonsense, as I said to Demis. And nothing happened at DeepMind, without Demis being okay with it because, you know, he was the by far the dominant founder of the three founders by far. And he ran the whole thing. And if Mustafa was kicked out, it was because Demis was okay with him being kicked out. And so, yeah, I would say he fired him and Demis would protest. And I did hear that protest, and I kind of ignored it. So so the reason they parted ways is partly just this is a classic founder story. You know, you work intensively together for ten years or nine years. And in the end, there's a difference of emphasis on where you wanna go, and I think Mustafa had kind of grown a lot through the DeepMind experience and wanted to basically run something himself. And Demis was the boss of DeepMind, so there wasn't space with two people trying to run it. And they clashed increasingly. And in the end, you know, Mustafa's very hard charging style led to bullying allegations, and that was kind of the pretext or the trigger, to to push him out. And, Demis was kind of okay with that because he was sick of fighting with his cofounder, and he was happy to move on.

00:18:46 Andrew Keen: Do you bring much of your own personal baggage to this? You you're very much part of the British elite. Your father was sir Christopher Mallaby, the British ambassador to Germany. You went to the top British school, Eton, then you got a first at Oxford. Demis came up a very different way as parents were. I mean, in in the way it's presented, and maybe you present it slightly different, the way it's presented in The Thinking Game, Demis' parents are kind of alternative hippie style people who didn't really fit in in any way. They're very unusual characters. But, certainly, his parents weren't British ambassadors to Germany, didn't go to Eton. What did you learn from, the Hassabis story about yourself? Did did you have some revelations about British class system and all that sort of thing?

00:19:46 Sebastian Mallaby: You know, look. I am who I am. I and I don't think it I mean, you know, it's easy for me to say this, but I don't feel like, you know, it's it's influencing how I judge people like Demis who came up a totally different way. I mean, if anything, it would make me more impressed, right, because I'm aware of my own privilege, and I'm, I think properly you know, I I acknowledge and and celebrate people who do incredibly well despite not having the advantages of going to a private school. So I I I, you know, I'm not sure what else to say about that. I mean, there there are other things that I think talking to Demis made me think about it. Right? Which, you know, one would be, for example, here you have this person who wants to make AI safe. He's not managing to make AI safe because it's a race dynamic, and it's kind of above him. It's bigger than him. He can't prevent other labs from being unsafe with AI, so he's sort of stuck. And so it makes me think about why does he go on? And I came to the view that, look, you could just condemn him and say you should stop. But then if he stopped and he took a professorship at Princeton, it wouldn't make the world safer. You know, Google would carry on building AI. All these other labs would carry on building AI. It would just remove from the scene one actor who would like ultimately to try and make it safe and use it for science and for medicine and good stuff. So I'm not sure it helps the world if he takes a kind of morally purist position and says, I'm out of here. I'm not gonna build it because it's not safe. Furthermore, I think you could argue that, you know, this you know, he sees a technology. He wants to develop it. It's both exciting, and it's frightening. And he takes that trade, and he moves on with it. Isn't that kind of like all of human history that we as humans have always confronted new technologies? They've always been both exciting and disruptive. We've taken that trade. We've moved on. We've gone forward with them. And this is, like, what it means to be human. We are not just, you know, I think, therefore, I am. We're also, like, we I invent, therefore, I am. And so I think there's something maybe Demis is sort of an enlarged version of all of us. So I I had those kinds of thoughts confronting Demis. If you wanna, you know, push me and and ask more specifically where the class thing matters, Happy to talk about it.

00:22:27 Andrew Keen: I had Geoffrey Hinton on the show a few months ago. Another Nobel Prize winner, another genius, maybe not quite in the Demis class, a slightly more mercurial character, perhaps. But I was really struck you talk about making AI safe. I was really struck by Hinton's intellectual naivety. He said to me that the way to make AI safe was to program it so it thinks like a woman, which to me, you know, not all women are ideal. We we all live through mark and missus Thatcher in England and blah blah blah. Does Demis bring you know, I take your point that he's incredibly smart and that he knows all about nuclear science, and he can walk around Russell Square and point to crossings and streets where one kind of scientific, invention was made or discussed. Does he have a sophisticated moral antenna? Has he read philosophy?

00:23:22 Sebastian Mallaby: Yeah. He has read philosophy. He he will he will talk to you about Spinoza. He will talk to you about Kant. He will talk to you about, I guess there's actually two that that came up the most, but, I he's he's pretty much of a polymath. You know, I read a book called Range once, pointing out that a lot of people who do great stuff do it because they, in their formative years, tried out a lot of different things. And so it's not actually true that to be a great violinist, you should only play the violin. You should play other instruments too. If you wanna be a great sportsperson, trying several different sports is a good idea. And Demis is the ultimate magpie. He's tried lots of things. You know, first degree in computer science, graduate degree in neuroscience. He did something different. Played chess to a almost professional level, then kinda quit and let go and a bunch of other games. You know, he's just spanned a lot of things. And, you know, that's why he's so interesting to me because he's got this range of intellectual curiosity. And I so I think trying to say, oh, no. No. He doesn't have a formed and sophisticated view of philosophy or for that matter, almost anything would be to underestimate him. He wouldn't Yeah.

00:24:38 Andrew Keen: I certainly wouldn't suggest that. I mean, I haven't met a guy. Although I've I hear about him a lot because my wife is senior executive at Google, so I hear all the stories about him. Mhmm. So he's more of a fox than a hedgehog in Berlin's terms. You you mentioned that he talks about, Spinoza. He talked to you about Spinoza and Kant, two very ethical philosophers or at least philosophers of ethics. How did he turn Kant or Spinoza into today's questions or dilemmas over AI? What how did he present his deep reading of philosophy in in terms of the dilemmas that he, as the head of AI at Google, has today?

00:25:27 Sebastian Mallaby: Well, Spinoza came into the conversation because Spinoza had this formulation, that I mean, as I I I did a bit of reading about Spinoza to try and get on Demis' wavelength. And as I recall, the interesting thing about this moment in the seventeenth century was that, you know, he was trained both in classical Greek philosophy, which was sort of non Christian, and at the same time in Christian thought. And he wasn't a conventional Christian, and in fact, the Christian church, condemned him. But he had a spiritual language because of that side of his training. And he talked about getting closer to God through scientific inquiry. Once you understand the way nature works, the way science works, the way physics works, that creation that you're understanding better is God's creation, and you're getting closer to God by understanding his creation. That was kind of the God the God that Spinoza describes in his writing. And for Demis, this God of Spinoza is also his religion. I mean, he said to me that for him, scientific inquiry is precisely getting closer to understanding, the fabric of reality. And this reality seems to be designed in a way that makes sense. And so there must be some sort of intelligent design behind it, or at least there might be.

00:27:02 Andrew Keen: Yeah. So you you wrote in the book doing science or he said doing science is like reading the mind of God.

00:27:08 Sebastian Mallaby: Yeah. And this comes from Spinoza, and it comes also via via Einstein and other scientists who invoked the god of Spinoza. And the funny thing is that I think once Demis started talking about it, some other tech people, notably Elon Musk, have said that they believe in the god of Spinoza. So it's it's kind of passed around amongst these tech pros. But, but I think Demis came to it before Elon. And, yeah. I mean, so that's an example of how philosophy figured.

00:27:41 Andrew Keen: So but some people might say if doing science is like, as, Demis said, doing science is like reading the mind of God, maybe science is a kind of warning. I mean, did he bring up counter enlightenment thinkers, people who questioned that faith in science? I mean, there's a great tradition of that in Western thought too and non Western thought.

00:28:10 Sebastian Mallaby: He didn't bring that up, but, I know that he, you know, he would occasionally say, the problem with such and such is that, you know, it's a challenge to the enlightenment. And he I think he shudders at that because the scientific method is something that he identifies closely with enlightenment thinking, the kind of faith in reason, the faith in accumulating knowledge. We all build on the shoulders of previous intellectual giants. That's how humans make progress. I mean, that kind of vision, which is an enlightenment vision, is sort of, I think, also his vision. And he sees himself as building AI to do better science. It's kind of the modern version of the telescope. In the past, scientists built new tools, microscopes, telescopes, etcetera, to understand the fabric of reality better. And now his way of building on that is is a machine cognition.

00:29:03 Andrew Keen: You mentioned, Demis could be teaching at Princeton rather than working at Google. Of course, Einstein worked at Princeton. Many other great physicists too, I think, including Oppenheimer at some point. Maybe I'm wrong on that. Yeah. You're right. Why I mean, if the way you're presenting him as this highly moral, highly intellectually sophisticated figure who cares about the future of the world and sees in science the mind of God. Why is he a senior executive of Google? Why not just take that job at at Princeton? He could choose his own job, his own title. He could probably choose to create his own university, call it AI University or Demis Hassabis College. Why is he spending his time with other senior executives at a 3 or $4,000,000,000,000 company based in Silicon Valley? Why waste his time on that front?

00:29:57 Sebastian Mallaby: Yeah. I don't think he would agree that he's wasting his time. I mean, he's

00:30:00 Andrew Keen: I don't mean wasting his I mean, he's obviously not wasting his time, but it's just one company. I mean, there's OpenAI. There's there's Microsoft. There's, there's Anthropic. I mean, he's not controlling the whole future of AI.

00:30:14 Sebastian Mallaby: No. He's not. But he, he is one of the people who is a leading architect of the future of AI, and possibly the most important one. Because probably OpenAI will go bust, because its valuation is insane.

00:30:29 Andrew Keen: Yeah. You wrote an interesting piece, back in January, for the New York Times, and I'll pay this is what convinced me OpenAI will run out of money. So you're definitely, very bullish on, I'm sorry. Yeah. Very bearish on on OpenAI. But sorry. Go on.

00:30:46 Sebastian Mallaby: Yeah. So and, I mean, look. I think, you know, Sam Altman, as I said before, not a scientist, a very good fundraiser, but so good at fundraising that he's overdone it and pushed the valuation of OpenAI up too much. And they need to radically focus the company and cut their cost and their burn rate if they're gonna survive. I think it's fifty fifty if they can pull that off. But, anyway, I I I I I think, you know, it's safe to say that Demis has more influence on the future of AI than anybody at OpenAI, not least because there is no single person at OpenAI who is both a scientist and a leader. So that's that's one comparison. You know, Daria is closer to being like Demis than, Sam Altman is.

00:31:33 Andrew Keen: Dario Amodei, of course, the CEO of Anthropic has got a lot more press these days because he's been willing to take on the DOD.

00:31:40 Sebastian Mallaby: Yep. Exactly. And, you know, he is a proper scientist, and he's the founder of a company, and he's building serious artificial intelligence, a very good model, particularly the one that came out in January, which is agentic. So he is definitely an important player. Whether he's more important than Demis, I sort of doubt because, I think if you look at the full range of innovations that's come out of Demis's shop, particularly on the science side with AlphaFold, which which was the system that won the Nobel Prize, I would say that Demis is more important than Dario, but anyway, we could we could debate that. But I mean, you know, so he so he's not the only AI leader. I'd say he's the most important AI leader. And, you know, see, this notion that he's at some, you know, mega company wasting his time is not really fair. He he's writing No.

00:32:31 Andrew Keen: I didn't say that. I I I it's exactly what I didn't say. I

00:32:35 Sebastian Mallaby: Oh, okay.

00:32:35 Andrew Keen: My question was and I don't say waste his time. I mean, you you say he doesn't really care about money. The only thing he spends money on are his Liverpool season tickets, which are probably very cheap these days. Why work at a large corporation when he could set up his own nonprofit when he could create his own department at Princeton or Harvard or Oxford or wherever he chose.

00:32:58 Sebastian Mallaby: Right. So so the answer is look. He he did fantasize to me several times about, you know, the delights of retiring to Princeton and going to the Institute of Advanced Studies, which is where one of those intellectual heroes of his spent time from Oppenheimer to Einstein to Kurt Godel and others. So that definitely is appealing to him, and, you know, he he doesn't dismiss that as a possibility at one point. But for now, to do what he's trying to do, build super powerful intelligence, he needs a ton of money. He needs enormous amounts of computing power and all the energy infrastructure that's needed to power those data centers. You can't do that even if you, you know, are a famous Nobel Prize winner. You're not gonna get foundations at an academic institution to back you to the tune of $100,000,000,000 a year, which is the current research budget at Google. So so to to do large scale cutting edge AI, you you need to be in a corporation. And the proof of this is partly OpenAI where they started that as a nonprofit, and they had to give that up because as a nonprofit, you couldn't raise the cash to to build Frontier Systems.

00:34:12 Andrew Keen: You, in the book, you spend quite a lot of time in the Infinity Machine. You talk a little bit about, the initial ethical battles Demis and Mustafa had with Google about the terms of acquisition and their role within the company. Perhaps talk a little bit about that. I mean, has is one of his is one of Demis' ambitions to transform Google to make it reflect his own values? After all, Google is a very controversial company on lots of fronts. They they initially started with the with the claim that they weren't gonna do any evil. Most people would disagree with that these days. Is he a moral crusader in your view within Google? Is that his history there?

00:35:05 Sebastian Mallaby: No. He's not a moral crusader. And let me explain a bit about that. I think I think he combines good values with pragmatism. And he believes in picking his shots. So he's not going to do a Dario, which is to take a principal position and tell the Pentagon, you can't use my technology either for domestic mass surveillance nor for autonomous lethal weapons, which is what Dario Amodei did about a month ago. And the reason Demis isn't gonna do that is because he thinks if he does that, it won't make a difference to the outcome. And in fact, it hasn't made a difference to what the pentagon does. What Demis does do is figure out where the door is a bit ajar and where you can push and actually get it open. So an example would be after ChatGPT came out, you know, politicians were suddenly waking up to AI and worried about safety. He suggested to, Rishi Sunak, The UK prime minister at the time, why don't you do a global AI safety summit, invite everybody, including the Chinese, and we'll do this at Bletchley, and we can start a conversation going about how to make this technology good for humanity. And Sunak did that. And then there was a follow on conference in Korea, South Korea. And so, you know, that was a little bit of progress. And then The UK set up the UK AI Safety Institute, and the scientists from DeepMind went over to that institute to be the, chief scientist working for the government on AI safety, and they did it with Demis' blessing. So, I mean, there are things where Demis sees an opportunity to actually make a difference, and I think he pushes on those. What he doesn't do and, you know, people can have legitimate debate about whether he should do more of this. He doesn't make a stand for the sake of making a stand.

00:37:02 Andrew Keen: I like Dario. I mean, you mentioned that the well read in Spinoza and, Kant, it sounds to me in the way he behaves, maybe in contrast to the way he thinks or believes that he's been reading Hobbes. He's a realist.

00:37:18 Sebastian Mallaby: Fair point. I mean,

00:37:19 Andrew Keen: there's not a lot of Kant and Spinoza in that kind of behavior, is there?

00:37:23 Sebastian Mallaby: No. You're right. But, yes, I take the point. I mean, Kant figures in his thinking in a totally different way, which is to do with But isn't

00:37:32 Andrew Keen: that in all seriousness, Sebastian, isn't perhaps I'm not suggesting that, Demis has many faces like Sam Altman. Very few people do. But is he able to compartmentalize technology and morality to suit himself, to make himself comfortable, to enable the remarkable life he's having? It must be very exciting managing AI at Google.

00:37:59 Sebastian Mallaby: Yeah. Look. You're you're doing, an intellectually stimulating job of of pushing on these points where he may just be self interested.

00:38:09 Andrew Keen: I'm not suggesting he's self interested or evil. I mean, there's a middle ground between being Yeah. A saint and a sinner, isn't there?

00:38:17 Sebastian Mallaby: Okay. Great. Now now we're talking the same language. I completely agree with that. And I think that there are lots of things where he has a position that's convenient. And, you know, you know, they're kind of you can debate whether they are true. Like, for example, he says that building Gemini, involves no taking away of focus from other bits of his mission such as AI for science. Now he is pursuing AI for science in parallel. You know, the AlphaFold two system has now been replaced by AlphaFold four. So the drug discovery agenda is advancing. But, surely, building Gemini takes up a massive amount of, resources and time and compute and energy and his psychic psychic focus and all that. And so there must be some displacement of progress in other fields. Then he also would argue, well, it's okay because if you look at the way that, large language models have developed since 2022, they've made a lot of progress, which is useful for all fields of AI. They're not just not it's not just chatbots that we're building here. We're actually advancing the science, which is generalizable. And, you know, I think that probably is true, actually, because there's been reasoning models and, you know, multimodal and all these different innovations, which are probably helpful for other types of AI that one might want to build in the future. But I think, you know, this is the interesting debate. This this, like, you know, everybody is somewhere in the middle between a saint and a sinner.

00:39:56 Andrew Keen: I think Elon, sure. I think Peter Thiel, who considered himself the antichrist.

00:40:06 Sebastian Mallaby: But I think I think dealing in, caricatures, of of these powerful people is always tempting, but not really true. But I'll take

00:40:17 Andrew Keen: your point. He was in the news recently. He's always in the news about possible Pentagon AI deals within Google. Did you do a lot of digging in Google? I mean, what's his reputation there amongst his staff and amongst other senior people? Is he the irritating moralist who always brings up good and evil in senior meetings, or is he the realist who gets stuff done?

00:40:43 Sebastian Mallaby: So there was a period between 2016 and 2019 when he conducted a secret with Mountain View. But people leaked me documents, so I knew all about it. And the point of the fight was that he wanted there to be an a sort of safety oversight board governing the deployment of artificial intelligence so that it wasn't just the Google corporate board that had a final say. But you you this technology is

00:41:24 Andrew Keen: Continue, Sebastian.

00:41:26 Sebastian Mallaby: Okay. So his position was this technology is is too powerful for just Larry and Sergei to be deciding what happens. So we need to get a board with somebody like Barack Obama or some other independent, reputable person who will be a counterweight. And Google kind of pretended to be open to that but wasn't really going to cross the line. So there was this long fight involving big teams of lawyers. People leaked me documents, which were the term sheets describing a DeepMind independence, spin out sort of structure, and it'd be covered with red lines where the other side's lawyers had suggested, you know, different language. So this is a very involved thing, and I think it you know, at that time, the people in Mountain View, at least at the top of the company who were on the other side of negotiation, view Demis as, you know, a very difficult character to handle. He was pushing on safety and making a big fuss about it. So, I mean, Sundar and David Drummond — the GC. Of course. And David Drummond, who was the sort of chief legal officer.

00:42:39 Andrew Keen: At the time, he's no longer there. He got thrown out too for bad behavior.

00:42:43 Sebastian Mallaby: True. And so, they, yeah, they they had a long fight. And, Demis was quite annoyed with me when I said I was gonna write about this, and I had to endure, you know, pressure from the DeepMind general counsel. But since he didn't have a leg to stand on, I I I I published it anyway. But, yeah, that's the best example of where he was viewed as an irritant.

00:43:11 Andrew Keen: I'm sure you're familiar with Parmy Olson's book, Supremacy, AI, GPT, and the Race That Would Change the World. It won the FT Book of the Year, I think, the year before last. Palmer's been on the show lots of times. Mhmm. She argues that Demis will end up as CEO of, of Google. That that I want and she hasn't written your deep psychological analysis. I mean, she I'm not sure how much time she spent with Demis. But when I used when I used, Gemini, Google's AI to when I asked Gemini about senior Google executives, it was interesting. Demis was about eighth on the list even after Neil Mohan, the, CEO of of of YouTube, and Kent Walker, who took over from David Drummond. Kent Walker is my wife's boss. Is he ambitious, do you think? I mean, is he seen as ambitious to wanna run Google? Sundar has done a great job, really. A lot of people thought he wasn't gonna last. The the Google stock today is up to about $3.20. At some point, he's gonna quit and go and spend the money, buy his own island. There are other very senior people who are influential. Philip Schindler, the SVP of business, the guy who runs the cloud, Thomas Kurian, TK is also very powerful. Do you think that is your sense that sometimes when, Demis can't go to sleep, he wonders whether he's gonna be CEO of Google? He plans. He plots.

00:44:50 Sebastian Mallaby: You know, I think this is a very interesting question because it goes to the dichotomy in his personality between, you know, the guy who would like actually to go to Princeton and be a professor and think big thoughts and have that freedom, to think, versus the sort of super competitive person who grew up as a chess player and simply loves to win. And so I think, you know, one part of him can't resist capitalist competition, and being CEO would be tough to turn down. But the other part of him would genuinely prefer not to do it and would rather do science, would rather at least being in the current position, he claims, that he still has time because he works until four in the morning every day to spend sort of between midnight and four, reading scientific papers and thinking about science. And, you know, he to his sort of emotional health and functioning, you know, I think he feels that's quite important. And he's worried that if he was the CEO of the whole thing, he couldn't do that. So I think he is genuinely torn. If I had to bet, I would say he does not want to be chief executive of, Google. But as I say, there is also this massively competitive side. So

00:46:07 Andrew Keen: Yeah. I wonder whether I mean, when you when I asked Gemini about Google priorities, the thing that came up first is an AI focus. One would guess I mean, he must be on pretty good terms with, Sunda. But one would guess that he wouldn't be particularly comfortable if somebody like Philip Schindler or TK or Neil Mohan took over Google with their very commercial instincts. So he might be sort of he might be dragged, if you like, morally to become CEO because in his moral mind, at least, or his semimoral mind, he's a better alternative to any of the other sales guys.

00:46:47 Sebastian Mallaby: Maybe. I mean, I I think, I think what's certainly true is that the relationship with Sundar Pichai is very strong, and they have this sort of buddy act where Sundar is managing the high politics of Google and delivering enormous resources so that Demis can do this extremely expensive research. And he's giving Demis the space, to to develop the AI as he would like to develop it. And, you know, Demis needs Sundar Pichai as his patron. And Sundar Pichai needs Demis as his scientific kind of, icebreaker. And, that buddy act is possibly the single most important buddy act in global business.

00:47:32 Andrew Keen: And that buddy act, I I assume, also exists in Demis' head. I mean, there's a little bit of Sundar in him. What do

00:47:39 Sebastian Mallaby: you mean by that?

00:47:41 Andrew Keen: Well, the realist versus the idealist, the guy who wants to read scientific papers until four in the morning in contrast to running a a company of several 100,000 people worth several trillion dollars, transferring to a board, having a stock price, dealing with the reality of sales, having to deal with cloud sales and lawsuits and all the rest of it?

00:48:05 Sebastian Mallaby: During these negotiations I described between DeepMind and Google, when Demis wanted to spin the company out or get a safety board, he kept on, you know, the latest legal draft of a potential independent DeepMind structure. And these things are, you know, fat, dense legal documents. And he would groan and say, why are you showing me this? I don't want this part of my brain to grow. And so I think that's an insight into his appetite for being CEO.

00:48:39 Andrew Keen: And finally, Sebastian, as we enter this AI age where we're supposed to have smart machines, you mentioned Turing earlier. Of course, there's the Turing test for mimicking AI. What does your story about Demis in the Infinity Machine tell us about what it means to be human? Often people come on the show and say, oh, AI is gonna destroy humanity or the human, whatever that means. I'm always wary of the h word. You haven't used it so far. But it does does Demis Hassabis himself, independent of Google and AI, does he offer, a lesson, a a slideshow, shall we say, in in what it means to be human as a model?

00:49:29 Sebastian Mallaby: I think well, first thing to say is that he has a model of superabundance. You know, some point in the future, maybe ten years from now, when the AI really has bedded in and it's generated enormous progress in science and productivity gains in the economy and all that. And, you know, we have abundance as a result. And in this environment, maybe companies don't really matter anymore because if you think about an economy where supply is massively increased by AI, the abundance means that prices fall. And and what is money and what I mean, you sort of he's Demis would've the,

00:50:09 Andrew Keen: the the, the Musk line as well, although I'm sure most of the people at Google wouldn't be thrilled if there were no longer companies.

00:50:17 Sebastian Mallaby: Right. I mean but Demis is quite happy to, riff on this super abundant future, and he says it derives partly from the science fiction he read when he was young. There's a series by Iain Banks called the Culture Series, which describes a future where there are lots of intelligent machines going about, you know, doing their business, having their own political intrigues, not bothering humans, living in parallel to humans. And humans, you know, can kind of travel to different planets because there are amazing rockets, and you can sample different ways of living on different planets. And, you know, to to Demis, this is a whole, a cornucopia of opportunity. And if you have abundant curiosity, then superabundance may be a wonderful thing. Now, you know, Demis is not a normal person in in the range of his thinking and the sort of privilege that that his intelligence buys him. And so for many people, this may be an entirely different prospect. So I think I'm, you know, I'm I'm telling you what he thinks. I'm not necessarily endorsing it because I find it somewhat naive, that that bit of it. Right.

00:51:19 Andrew Keen: So what you're saying, Sebastian Mallaby, author of the Infinity Machine, is that, Demis' idea of abundance is wrong.

00:51:31 Sebastian Mallaby: Yeah. Yeah. Certainly too simple. Yeah. Yeah. I mean, I I think I think to to to find for me, the bit in the story where one confronts a future with AGI, super powerful intelligence, is among the Go players who were defeated by AlphaGo. Because this is super intelligence in one domain. You had suddenly a machine that could beat the best players in the world at Go. And so for these players who had spent their entire lives drawing on a human tradition that went back centuries of how you find patterns in Go, how you play the game, the deep beauty of the strategy and the tactics and the patterns and the aesthetics, you know, there was a sense that all of this accumulated human understanding of the game had been devalued by an algorithm that could do it all better and discover even more patterns that humans hadn't thought of. And some players, professional Go players reacted by saying, I quit. I no longer see any point playing this. This you know, my life has been rendered meaningless. Others said, well, this is actually beautiful because the machine has discovered new patterns that we didn't know about, and I can find more depth in this game that I'm curious about. I wanna understand it, and the machine will be my tutor. And in fact, both in chess and in Go, all the top players these days train with a computer system that's better than them, that teaches them new strategies. And so I think that's two visions of how humans can react to incredibly strong machine cognition.

00:53:10 Andrew Keen: You didn't answer my question, Sebastian. I mean, the question was a simple one.

00:53:14 Sebastian Mallaby: Okay.

00:53:15 Andrew Keen: So Demis has a vision of abundance in the future where there's no companies, no money. It's a kind of Musk-ian utopianism. Do you believe that's realistic or not? Yes or no? No. Well, that's clear. Excellent conversation. Wonderful, story of Demis Hassabis, DeepMind, and the quest for superintelligence, the Infinity Machine by one of our best journalists. Thank you so much, Sebastian. Very interesting conversation. Much appreciated. We'll have to get you back on when Demis runs Google to see, what he's gonna do with

00:53:46 Sebastian Mallaby: the company. Okay. I enjoyed it. Thank you.