No Priors Ep. 16 | With Mustafa Suleyman, Founder of DeepMind and Inflection

1. 0:00 – 10:36

   From Conflict Resolution to AI Pioneering

   1. NANarrator0:00

      (instrumental music plays) Today on No Priors, we're speaking with Mustafa Suleyman, co-founder of DeepMind, the pioneering AI lab acquired by Google in 2014 for $650 million, and now co-founder and CEO of Inflection, along with Reid Hoffman and Karine Simonyan. Inflection just launched their first public product, Pi, last week. Mustafa, welcome to No Priors. Thanks so much for joining us.

   2. MSMustafa Suleyman0:26

      Thanks for having me. I'm super excited to be here.

   3. SGSarah Guo0:29

      Yeah. We're- we're very excited to have you today. I think one thing that'd be great to maybe start with is just a little bit of your personal story, because I think you have a really unique background. You're very well-known obviously for, um, DeepMind and your pioneering work in the AI world. But I think before all that, you worked on a Muslim youth helpline, you started a partnership and consultancy that was focused on conflict resolution to navigate social problems. I'd just love to hear a little bit more about the early days of things that you did before, uh, DeepMind, and then maybe we could talk a little about DeepMind and- and, sort of, more recent stuff as well.

   4. MSMustafa Suleyman0:57

      Yeah, sure. I mean, the truth is, I- I was very much a, kind of, change-the-world kid growing up. Like, um, a big believer in grand visions, doing good, having a huge impact in the world. And that was always, kind of, what drove me. Um, so when I... I grew up in London and went to Oxford, um, but at the end of the second year of my philosophy degree, I was, kind of, getting a bit frustrated with the, sort of, theoretical, you know, nature of it all. It was full of hypothetical moral quandaries. And, um, so a friend that I met, um, at Oxford was starting a telephone counseling service, a- a, kind of, helpline, and it really appealed to me. It was a non-judgmental, non-directional, secular support service for young British Muslims, and this was, like, about six months after the 9/11 attacks, and so there was quite a lot of, like, rising Islamophobia and the government was talking a lot about anti-terrorism and, you know, in general, I think that, like, sort of, migrant communities were feeling the pressure and, um, this was a support service that was staffed entirely by us, by young people. I was 19 at the time. And, yeah, I spent, uh, almost three years working pretty much full time on- on that. And it was an incredible experience, because it was basically my first startup and, you know, fundraising was the name of the game, except the numbers were much, much smaller than they are these days. Uh, and, you know, the surf- the service was staffed by almost 100 volunteer young people, which was just amazing 'cause we felt like we can actually do something, you know? It was quite liberating and energizing to actually give this a shot. And, you know, I was very much inspired by the kind of human rights principles. It was deliberately not religious, even though it used some of the kind of culturally sensitive language that help people feel heard and understood. Um, so yeah, it's had a- a very formative impact on my- my outlook.

   5. SGSarah Guo3:02

      Yeah, no, it's- it's super interesting and I think we can talk more about that in the context of AI in a little bit. One other thing that you did is you also started a consultancy where you worked as a negotiator and facilitator, and I believe you worked with clients like the United Nations, the Dutch government, and others. Could you tell us a little bit more about that work as well?

   6. MSMustafa Suleyman3:19

      Yeah, I mean, I- I was always trying to figure out how to scale my impact. And, you know, I quite quickly realized that delivering a, sort of, one-to-one service via a nonprofit was not going to scale a great deal, even though it had um- an amazing impact, um, you know, on a, kind of, human-to-human level. And so I was super interested in these, like, meta structures, like how does, you know, the UN actually influence, you know, behavior at- at- at the country level? Um, and, you know, how could we run more efficient decision-making processes, um, where there's tension and disagreement? So we worked all over the world actually, in Israel-Palestine and in, you know, um, in Cyprus between the- the Greeks and the Turks. Um, my colleagues worked in South Africa, in Columba- Colombia, Guatemala. And I think it really taught me that learning to speak other people's social languages is actually an acquired skill, and you really can do it with a- with a little bit of attention to detail and some patience and care. It's kind of a superpower being able to deeply hear other people and make them feel heard such that they're better able to empathize with people that they disagree with, and that- that's been an important theme throughout my, kind of, career, something I've always been interested in. So I think, I think I co-founded that, uh, and worked on it for, I think, three years and soon realized the limitations of, like, large-scale human processes. I mean, I've... In 2009, uh, 2009, I worked, I facilitated one part of the climate negotiations in Copenhagen, and, uh, yeah, it was a, kind of, a remarkable experience. Like, you know, 192 countries, literally a thousand NGOs and, uh, activists, many different academics, everyone proposing a different solution, a different definition of the problem. And, you know, in one way it was sort of inspiring to see so many different cultures and ideas coming together to try to form consensus around an issue that was clearly of existential importance. On the other hand, it was just, like, deeply depressing that we weren't able to achieve consensus. It took another decade to even get mild consensus on this, or half a decade, 2015. And I think that was sort of an eye-opener for me. I was like, the world's governance systems are not going to keep up with both the exponential challenges that we face from globalization and carbon emitting, but also, like, technology. And- and that was the next thing that I saw on the landscape.

   7. SGSarah Guo6:01

      So, how did this lead into your interest in AI and, you know, I believe that you met, uh, Demis when you were quite young, and I think he and your other co-founder worked together later in a lab. But I'm a little bit curious like how your background and interests in these sorts of global issues then transformed into an interest in AI and the founding of DeepMind.

   8. MSMustafa Suleyman6:19

      Yeah. Well, around about that time actually, like, I guess, it was like 2008 or so, I was starting to keep an eye on, um, Facebook's rise and I was like, "This is incredible." I mean, this is like a two or three-year-old platform at that point and it had hit like 100 million monthly actives, and that was just a mind-blowing number to me. And i- it was obvious that this wasn't just the kind of neutral platform for giving people access to information or connecting people with other people. The, the frame... Because I had come from a conflict resolution background, our entire approach was like, what is the frame of a conversation? Like how do you organize space? How do you prepare individuals to have a constructive disagreement? How do you, like, set up the, the environment basically to facilitate dialogue? And so that was the lens through which I looked at Facebook. I was like, "Well, this is a frame. There's a choice architecture here." There are significant design choices which are gonna incentivize certain behaviors. Obviously, at that point, there wasn't really ranking but even just having a thumbs up or, like, the choice of, you know, which button you place in what order and how you arrange information on the page and what... All of that drives behaviors, um, in one way or another. And, you know, that was a big realization to me because I was like, well, this is actually reframing the default approach to human connection at a scale that is like completely unimaginable. I mean, perhaps only akin to, you know, the default expectations in a religion, for example. Everyone grows up with an idea that there is a, you know, a patriarchy, a male god, that, you know, that there's a particular w- role for women. Like, that's, that... You know, un- until a few decades ago, that was just an, an implied sort of undertone to an entire social structure for thousands of years. Uh, and that's kind of what I mean by frame, there's this sort of, these implicit design choices which cause hundreds of millions of people to change their behavior.

   9. SGSarah Guo8:19

      Yeah. And I- I think that, that's super interesting because I remember working on a bunch of Facebook apps at the time when the platform launched, and people were purposefully thinking about that stuff but on the micro-level, right? How do we get more users? How do we p- get people to convert? How do we drive certain behaviors? And so everybody, I think, was very explicitly thinking about this as a behavioral change platform but not at the level of society.

   10. MSMustafa Suleyman8:40

       Right.

   11. SGSarah Guo8:40

       You know, we were thinking about it in the context of just like how do you get more people to use this thing, you know? (laughs) And so I think it's really interesting that people then later realize the big ramifications of this in terms of, you know, how that actually cascades in terms of social behaviors and other things. How did that lead to starting DeepMind?

   12. MSMustafa Suleyman8:56

       Well, it was clear to me from that moment on, like, I left Copenhagen in 2009 thinking this is not the path to significant positive social change. It still needs to continue and I support those processes obviously, but I'm just saying it is just not something that I feel I could continue to work on full-time. And so my heart was set on technology at that point, so I reached out to Demis, who, uh, was the brother of my best friend from when I was a kid. We got together, we had a coffee, we went... and actually, we played poker, um, at, (laughs) at one of the casinos in London 'cause we both love games, we're both super competitive, uh, both good at poker. And on that night, I think we both got knocked out pretty early (laughs) in the tournament, so we sat around drinking Diet Coke, uh, talking about ways to change the world. And we basically were, you know, having exactly this conversation, like, you know, is it going to be... Uh, I mean, obviously I- at that point, I was m- mostly inspired by platforms and software and social apps and connectivity and so on, whereas, um, you know, Demis was way more in the kind of robotics land and sci-fi land. I mean, he, he was, he was fully thinking that, you know, the way to manage the economy, the way to make economic decisions was to simulate the entire economy, right? And, and he thought that... He was very much obviously... had just come off the back of his games like Evil Genius and Black & White and so on, which were kinda simulation-based games, so I think that was his default frame at that point. Um, yeah, and then we spent many months talking and spent a lot of time with Shane Legg as well.

2. 10:36 – 15:32

   Defining Intelligence

   1. MSMustafa Suleyman10:36

      And Shane was really the core driver of the ideas and the language around artificial general intelligence. I mean, he had worked on that for his PhD, um, uh, with Marcus Hutter, um, on definitions of intelligence. I- I found that super inspiring. I think that was actually the turning point for me, that it was pretty clear that we at least had a thesis around how we could distill the sort of essence of human intelligence into an algorithmic construct. And it was, it was his work in... I think he... I think for his PhD thesis he put together like 80 definitions of intelligence and aggregated those into a single formulation which was how do we... Um, you know, the ab- the... Intelligence is the ability to perform well across a wide range of problems, and he basically, you know, gave, gave us a measurement, an engineering kinda measurement that allowed us to constantly measure progress towards, you know, w- whether we were actually producing an algorithm which was inherently general, i.e., it could do many things well at the same time.

   2. EGElad Gil11:42

      Is that the working definition you use for intelligence today?

   3. MSMustafa Suleyman11:46

      Um, (sighs) actually, no. (laughs) I've changed. (laughs) Um, I, I think that there's a more nuanced version of that. I th- I think that's a good definition of intelligence but I think in a weird way it's over-rotated the entire field on one aspect of gen- of intelligence, which is generality, you know. And I think, um...... OpenAI and, um, then subsequently Anthropic and others have taken up this default sort of mantra that, like, it- all that matters is, can a single agent do everything? You know, can it be multi-modal, can it do translation and speech generation, recognition, et cetera, et cetera. I think there's another definition which is valuable, which is the ability to direct attention or processing power to the salient features of a- of, uh, uh, an environment given some context, right? So, um, actually what you want is to be able to take your raw processing horsepower and direct it in the right way at the right time, because it may be that a certain tone or style is more appropriate given a context. It may be that a certain expert model is more suitable, or it may be that you actually need to go and use a tool, right? And obviously we're starting to see this emerge, um, and in fact I think the key, and we can get into this obviously in a moment, but I- I think the key element that is going to really unlock this field is actually going to be the router in the middle of a series of different systems which are specialized, some of which don't even look like AI at all. They might just be traditional pieces of software, databases, tools, and other sorts of things. But it's the router, uh, or- or the kind of central brain, um, which is going to need to be the key decision maker, and that doesn't necessarily need to be the largest language model that we have.

   4. SGSarah Guo13:43

      Right. It's really interesting because I feel like a lot of what you described is actually how the human brain seems to work in terms of you have something a little bit closer to a- a mixture of experts or MoE model where you have the visual cortex responsible for visual processing and then you have a other piece of the brain specifically responsible for empathy and you have mirror neurons and, you know, it feels like the brain is actually this ensemble model in some sense with some- some routing depending on the subsystem you're trying to access. And so, you know, the generality approach seems like a really... it- it almost goes at odds with some of those pieces of it unless you're just talking about some part of the hippocampus or something, right?

   5. MSMustafa Suleyman14:17

      Well, I think that's long been the inspiration, right? I think for everybody the- these neural networks are the obvious example, but in-

   6. SGSarah Guo14:23

      Mm-hmm.

   7. MSMustafa Suleyman14:24

      ... many other elements, reinforcement learning, um, you know, and so on, are- are- are all brain inspired. And I think that, you know, there's been a lot of talk about, you know, sparsity as well-

   8. SGSarah Guo14:35

      Mm-hmm.

   9. MSMustafa Suleyman14:35

      ... which is sort of what you're describing. And, you know, so far we've- we've had to do, you know, very dense all-to-all connections because we sort of haven't quite learned the algorithms for sparse activations.

   10. SGSarah Guo14:46

       Mm-hmm.

   11. MSMustafa Suleyman14:46

       But I- I think that's gonna be a very promising area. And, you know, in many ways what I'm describing doesn't actually require sparse activations because, you know, you- you actually could just train a decision-making engine at the middle to know when to use which size model, right? So maybe in some contexts you would want the highest quality, super expensive, 20-second latency model, and in most other contexts, a super-fast, three-second mini model might work fine. Um...

   12. SGSarah Guo15:14

       Mm-hmm.

   13. MSMustafa Suleyman15:15

       I think that's the- gonna be the- the key unlock actually, and- and quite sort of remarkably, that's an engineering problem, um, perhaps more than it is a- an AI problem which, you know, is- is just a pretty surreal moment give-, you know, just if you actually observe that given where we are in the field and stuff.

   14. SGSarah Guo15:31

       Mm-hmm.

3. 15:32 – 24:45

   DeepMind's Journey and Breakthroughs

   1. SGSarah Guo15:32

      When you, um, started DeepMind, I think it was reasonably unpopular to do what you were doing, right? And so I think you ended up getting funded-

   2. MSMustafa Suleyman15:38

      (laughs)

   3. SGSarah Guo15:38

      ... by, um, Founders Fund and Peter Thiel and Elon Musk. But I remember at the time there was, like, three or four parties that funded a lot of AI things and then nobody else was really doing it in terms of the types of approaches you were taking in terms of saying, "We're gonna build these big AI systems that can do all sorts of things," right?

   4. MSMustafa Suleyman15:54

      Yeah, I mean, it was wacky. Like, I- I can't say that enough. Like, it- it was... especially for the first two years. So 'cause we founded in 2010 and for the- most of the sort of spring and summer of 2010, actually most of the rest of that year, I was going to Gatsby Computational Neuroscience Unit at UCL sneaking in with Demis and Shane to just sit in on the lunches that, uh, Peter Dayan ran. And I remember Shane, like, sort of saying to me, like, you know, "The language here is machine learning."

   5. SGSarah Guo16:28

      Yeah, you couldn't say AI.

   6. MSMustafa Suleyman16:29

      Don't say AI.

   7. SGSarah Guo16:31

      Yeah.

   8. MSMustafa Suleyman16:31

      (laughs) And I was like, "Okay, okay. I'll keep my mouth shut. Don't worry."

   9. SGSarah Guo16:34

      (laughs)

   10. MSMustafa Suleyman16:35

       Like, we certainly don't say AGI. Um, you know, and- a- and that- that was a kind of... that was pretty weird. I mean, th- you know, there weren't... y- you know, there weren't very many funders for us. Like, you know, Peter Thiel, you know, to his credit, uh, did actually have significant vision here, although he sold pretty early I think and now doesn't seem to be in the game. So... but, uh, yeah, he certainly- he certainly saw it first, um, and, you know, I think that all changed pretty quickly, first with, uh, you know, AlexNet, of course, in 2012 and then with DQN, uh, the Atari paper in 2013. Um, you know, and then a kind of succession of breakthroughs after AlphaGo and people got more- more- more sort of aware of it. But it still surprises me the extent to which the rest of the world is, like, suddenly waking up and obviously we've seen that like crazy in the last six months, so...

   11. SGSarah Guo17:27

       Yeah. And then, I guess, last question on sort of your time with, uh, Google and DeepMind and... 'cause I think there's a lot of really exciting things to talk about in the context of inflection and sort of the broader field and world. What are some of the things you were most excited to have the team create at DeepMind over the years or some of the breakthroughs that you're most proud of?

   12. MSMustafa Suleyman17:43

       Yeah. Well, I mean, in some ways we- we- we definitely sort of pioneered the deep reinforcement learning effort and I think, um, you know, in principle it's a very promising direction. I mean, you clearly want some mechanism by which you can learn from raw perceptual data and that directly feeds into a reinforcement learning algorithm that can update...... and essentially iterate on that in real time with respect to some reward function, whether that's online or offline, like directly interacting with the real world in real time or it's, you know, in, in a kind of batch simulation mode. Um, you know, and, and that turned out to be very valuable for a specific type of problem, um, where a game-like environment had a very structured scaler reward and we could play that game many millions of times. Um, that's part of the reason why we started the AlphaFold project because it was actually my group that was, um, looking around for other applications of DQN-like, AlphaGo-like, uh, tools. And, uh, in a hackathon, um, that we did one week, um, someone stumbled acro- across this problem. We'd actually looked at it back in 2013 when it was called FoldIt, which was a, a very small scale kind of version of this. Um-

   13. SGSarah Guo19:09

       And, and just for context, sorry to interrupt, you know, AlphaFold was focused on folding proteins, which at the time was a really hard problem, right? People were trying to do this molecular modeling and they couldn't really make any real headway in lots of the traditional approaches. And then your group at DeepMind really started pioneering how to think about protein folding in a different way. So sorry to interrupt, I just wanted to give context for people listening.

   14. MSMustafa Suleyman19:28

       So I s- I think the hackathon was probably 2016 and then as soon as we saw the hackathon that, you know, start to work, then we actually, you know, scaled up the effort and hired, um, you know, a bunch of outside consultants to help us with the domain knowledge. And then I think the following year, we, um, entered the CASP competition. So, you know, these things take a long time, uh, and, you know, sort of longer than I think people realize. You know, it was a lot of, it was a, that was a very big effort by DeepMind, and eventually it became a, a company-wide, um, strike team. Um, so in, in hindsight, these things do take a huge amount of effort. Um-

   15. SGSarah Guo20:04

       Yeah, the fascinating thing here is that, you know, the work started with AlphaGo which was how to play Go better, right? Or how to beat people at Go, and then the same underlying approach could then be morphed and applied to protein folding, which I think is an amazing sort of leap or connection to make. And, you know, I used to work as a biologist, and I remember you'd spend literal years trying to crystallize proteins in different solutions, you'd do all these different salt concentrations in each well so if the protein would crystallize, you could hit it with X-rays and then you'd interpret those X-rays to look at the structure, right? And so you had to do this really hard sort of chemistry and physics to get any information about a protein at all, and then you folks with the machine ran through every protein sequence literally in, in the, in the, in every database for every organism, and you're able to then predict folding which is, it's pretty amazing, it's very striking.

   16. MSMustafa Suleyman20:55

       Yeah, I mean, I, I think the, the, if, if I were to sort of summarize the core thesis of DeepMind, it was that it would be possible to, the motivation for generality was that you would be able to learn, um, you know, a, a rewarding behavior in one environment and transfer in a more compressed or efficient representation the insights that had made you successful in one environment to the next environment, right? That transfer learning has always been the key goal, and that was one of the, one of the very exciting proof points that it is, you know, um, increasingly looking likely that that's possible. So, you know, I definitely think that's, that's pretty cool because if, when you think about the sorts of problems that we're facing in the world today, we don't have obvious answers lying around, there's no, like, genius insight that's just waiting to be applied. We actually have to discover new knowledge, and I think that's the, that's the attraction of artificial intelligence, that's why we want to work on these, you know, on these models because, you know, we're, we're sort of at the limit of what, you know, the smartest humans in the world are, are capable of inventing. And we have, you know, very pressing, urgent global challenges, you know, from food supply, to water, to decarbonization, to clean energy, transportation, you know, with a rising population that we really wanna solve. So there are th- you know, a- amidst all of the stresses and the fears about everything that's being worked on at the moment, it is important to keep in mind that there is a, an important north star that everybody is working towards, and we've just gotta keep focused on those goals rather than sort of be too sidetracked by, um, some of the fears.

   17. EGElad Gil22:41

       Let's talk about Inflection. What was the motivation for starting another company?

   18. MSMustafa Suleyman22:47

       Um, well, I guess back in sort of 2018, 2019, it wasn't clear that neural networks were going to have a significant impact in language. You should think about it intuitively, um, for the, for the previous sort of five years, CNNs had been effective at learning structure locally, right? So pixel i- i- in an image, in the input, so pixels in an image that were correlated in space tended to produce, you know, sub-features which were, you know, a good representation of what you were trying to predict. Maybe there were lines and edges, and they grew into eyes and faces and scenes and so on. And that kind of hierarchy just intuitively seemed to make sense and seemed to apply to audio and, and other modalities, right? Whereas if you kind of think about it, a lot of the structure of predicting the next word or letter or token in a sentence seems to exist in a very, very, very spread out, you know, far removed from the immediate next step of the prediction, right? And so it didn't look like that was working and then, to be honest, like, when GPT-3 came out, that was like a big revelation. Um, I, I had seen the GPT-2 work and hadn't quite clicked for me that this was significant. It was really only when I started, saw the GPT-3 paper that my eyes were wide open to this possibility. It's pretty amazing that you could attend to-... you know, a very, very sparse, seemingly sparse representation and use that to predict something which, on the face of it, seemed like there were billions of possibilities of what might come next in a sentence, or maybe tens of millions or something, but a lot. And for me, it was early 2020 that I went, uh, to work at Google and, uh, I got involved in the large

4. 24:45 – 33:22

   The Future of Personal AI Companionship

   1. MSMustafa Suleyman24:45

      language model efforts. I got involved in the Meena team, it was called at the time. I know that you guys had Noam on the show recently. Um, Noam's super awesome and it was me and Noam, Daniel, Quoc Le, uh, and a few others, and it was just unbelievable what was being built there. And, um, at, uh, when I joined, it was pretty small models and, um, very quickly, we scaled it up. It became the, the LaMDA group, um, and we started seeing how it could potentially be used in various kinds of search, started looking at retrieval, grounding for improving factuality, started getting a feel for all the hallucinations and so on. And that was just really a mind-blowing few years to me and, um, while I was there, sort of in the, in the last year, in 2021, I tried pretty hard to get (laughs) things launched at Google. We were all kind of banging on the table being like, "Come on, this is the future." And, uh, you know, um, obviously David Luan from Adept was also in and around that group, so the three of us, in our own ways, were pushing pretty hard for, for launch and, um, it wasn't meant to be. Uh, just, you know, timing is everything and, you know, Google just wasn't, it wasn't the right timing for Google for various reasons. (laughs) Uh, and, you know, I, I was just like, "Look, this, this has to be out there in the world. This is, this is clearly the new wave of technology." And so yeah, in January, I left, got together with Karen, my co-founder, um, who I worked with at DeepMind for seven years. We bought his company back in 2014 at DeepMind. He led the, um, deep learning scaling team at DeepMind for years and worked on all the big breakthroughs at DeepMind. Uh, and then of course, Reid Hoffman, who's been my, uh, one of my closest friends for like 10 years, and we've always talked about starting something together and, um, I was like, "This is the obvious thing. Now is the time for sure." And so, the rest is history, you know? We've, we've, we've... it's been a wild ride since then.

   2. EGElad Gil26:42

      It makes me feel a little bit better that somebody who's been such a pioneer in the field, uh, and working on this all the time is still constantly surprised, as I am also-

   3. MSMustafa Suleyman26:51

      (laughs)

   4. EGElad Gil26:51

      ... constantly surprised. Um, I, I remember when you were first starting to get this going, I... Another thing I was surprised by is the focus you... I mean, I came around to it (laughs) in, in writing the investment memo, but y- y- you know, you have this focus on the idea of companionship rather than information as the right initial approach. Uh, you've talked about, worked on, thought about empathy for humans and other populations for a long time. It seems counterintuitive, like, wh- wh- why companionship?

   5. MSMustafa Suleyman27:22

      Yeah, it's a great question. So, I think to step one step back from that first, I think my core insight about what was missing for LaMDA was interaction feedback and, um, in a funny way, that was exactly what was motivating Karen too.

   6. EGElad Gil27:40

      Mm-hmm.

   7. MSMustafa Suleyman27:41

      Um, having beaten all the, the academic benchmarks and achieved SOTA many times, he had come to the same conclusion I had seen the same thing from LaMDA, what we were missing was, was user feedback. And, um, actually when you think about it, all of our interfaces today are fundamentally about interaction, you know? You're giving your browser feedback all the time, you're giving, uh, you know, that web service feedback, same with an app or anything that you interact with. It's actually a dialogue and so the way I positioned LaMDA at Google is that, you know, conversation is the future interface, and Google is already a conversation, it's just an appallingly painful one, right? You say something to Google, it gives you an answer in 10 blue links, you say something about those 10 blue links by clicking on it, you, it, it generates that page, you look at that page, you say something to Google by how long you spend on that page, what you click on it, how much you scroll up and down, et cetera, et cetera, and then you come back to the search, log in, and you update your query and you say something again to Google about what you saw. That's a dialogue, and Google learns like that. And the problem is it's, you know, using, uh, 1980s yellow pages to have that conversation, and actually now we can do that conversation in fluent natural language. And I think the problem with what Google has sort of, I guess, in a way accidentally done to the internet is that it has basically shaped content production in a way that optimizes for ads and everything is now SEOed fro- to within an inch of its life, you know? You, you go on a web page and all the text has been broken out into sub-bullets and sub-headers and, you know, separated by ads and, you know, you, you spend like five to seven or 10 seconds just like scrolling through the page to find the snippet of the answer that you actually wanted. Like most of the time, you're just looking for a quick snippet, and if you are reading, it's just in this awkward format, and that's because if you spend 11 seconds on the page instead of five seconds, that looks like high-quality content to Google and it's quote-unquote "engaging," so the content creator is incentivized to keep you on that page. And that's bad for us because what we want is a succinct-

   8. EGElad Gil29:59

      We as humans.

   9. MSMustafa Suleyman30:00

      Well, we as humans, all humans-

   10. EGElad Gil30:02

       Yeah.

   11. MSMustafa Suleyman30:02

       ... clearly want a high-quality, succinct, fluent, natural language answer to the questions that we want, and then crucially, we wanna be able to update our response without thinking, "How do I change my query and, like, write this?" We- we've learnt to speak Google. Like, it's a crazy environment. We've learnt to Google, right? We le-

   12. EGElad Gil30:23

       Mm-hmm.

   13. MSMustafa Suleyman30:23

       Tha- tha- that's just a weird lexicon that we've de- co-developed with Google over 20 years. No. Like now, that has to stop. That's over. That moment is done, and we can now talk to...... computers in fluent natural language, and that is the new interface. Um, so that, that, that's what I think is going on.

   14. EGElad Gil30:41

       Maybe we should back up for a second and just tell people about what Pi is.

   15. MSMustafa Suleyman30:45

       Sure, yeah. So building on all of that, we think that Pi, I think that everyone, in the next few years, is gonna have their own personal AI, right? So there's gonna be many different types of AI. Um, there will be business AIs, government AIs, nonprofit AIs, political AIs, influencer AIs, brand AIs. All of those AIs are gonna have their own objective, right, aligned to their owner, which is to promote something, sell something, persuade you of something. And my belief is that we all, as individuals, want our own AIs that are aligned to our own interests and on our team and in our corner, and that's what a personal AI is. And ours is called Pi, uh, Personal Intelligence. It is, as you said, there to be your companion. Um, we've, we've started off as, uh, with, with a style that is, um, empathetic and supportive, and we tried to sort of ask ourselves at the beginning, like, "What makes for great conversation?" When you have a really flowing, smooth, you know, generative interaction with somebody, what's the essence of that? And I think there's a few things. Like, the first is the other person really does listen to you, right, and they demonstrate that they've heard you by reflecting back some of what you've said. They add something to the conversation, you know, so it's not just regurgitation, but they introduce another nugget, another fact. Um, they ask you follow-up questions, and they're curious and interested, um, in what you say. And, you know, sometimes there's a bit of spice, right? They throw in something silly or surprising or random or kind of wrong, and it's endearing. And you're like, "Oh," like, "we, that, that, we're connecting." And so we've tried to as, as in our first version, and this really just is a first version. Like, this is actually not even our biggest model at the moment. Um, so we're just putting out a first version that is skinned for this kind of interaction so that we can sort of learn and improve, and, you know, it really makes for a good companion, um, someone that is thoughtful and kind and interested in, in your world as a, as a first start.

   16. SGSarah Guo33:01

       You're working on these sort of personalized intelligence or personal agents, and you mentioned how you think in the future there'll be all these different types of agents for representing different businesses or causes or political groups or the like. What do you think that means in terms of h- how the web exists and how it's structured? So to your point, the web is effectively really based on a lot of SEO and a lot of sort of Google as the access point.

5. 33:22 – 41:49

   AI and the Future of Personalized Content

   1. SGSarah Guo33:23

      What happens to web pages or what happens to the structure of the internet?

   2. MSMustafa Suleyman33:27

      I think it's gonna change fundamentally. I think that most computing is gonna become a conversation, and a lot of that conversation is gonna be facilitated by AIs of various kinds. So your Pi is going to give you a summary of the news in the morning, right? It's going to help you keep learning about your favorite hobby, whether it's cactuses or, you know, like, motorcycles, right? And so, you know, every couple days, it's gonna send you new updates, new information in a summary snippet that really kind of suits exactly your reading style and your interests and your preference for consuming information. Whereas a website, you know, the traditional open internet just assumes that there's a fixed format and that everybody wants a single format, and generative AI clearly shows us that we can make this dynamic and emergent and entirely personalized. So, you know, if I was Google, I would be pretty worried (laughs) 'cause the c- that's, that's, that, that, that old school system does not look like it's gonna be where we're at in 10 years time. It's not gonna happen overnight. There's gonna be a transition. But these kind of succinct, dynamic, personalized, interactive moments are, are clearly the future in my opinion.

   3. EGElad Gil34:43

      The other group of people that is clearly worried is anybody with a, with a website or their business is that website. I spent a lot of time talking to publishers in April because they were freaking out, and, uh, uh, what ad- what advice would you have for people who, like, generate content today?

   4. MSMustafa Suleyman34:59

      (sighs) Well, I think that, you know, an AI is kinda just a website or an app, right? So you can still have, like, let's say that you have a blog about baking and so on. You, you know, you're, you can still produce super high quality content with your AI, and your AI will, you know, be, I think, a lot more engaging and interactive, um, for other people to talk to. So to me, any brand is already kind of an AI. It's just using static tools, right? So, so, you know, for, for a couple hundred years, the ad industry has been using color and shape and texture and text and sound and image to generate meaning, right? It's just they release a new version every six months or every year, right? And it's, you know, the same thing that applies to everybody, like TV ads used to be, right? Whereas now that's gonna become much more dynamic and interactive. So I, I really don't subscribe to this view that there's gonna be, like, one or five AIs. I think this is, like, completely misguided and fundamentally wrong. There are going to be hundreds of millions of AIs or billions of AIs, and, and they'll be aligned to individuals. So what we don't want is autonomous AIs that can operate completely independently and wander off doing their own thing. That I, I'm really not into that vision of the world. That doesn't end well, right? But, you know, if your blogger, you know, has, you know, their own AI that represents their content, then I imagine a world where my Pi will go out and talk to that AI and say, "Yeah, like, my Mustafa is super interested to learn about baking. He can't..."... crack an egg. So where does he need to start, right? And then Pi will have an interaction and be like, "Oh, that was really kind of funny and interesting. Mustapha will really like that." And then Pi will come back to me and be like, "Hey, I found this great AI today. Maybe we could set up a conversation. You'll find something super interesting." Or they recorded this, this little clip of me and the other AI interacting, and here's a three, three-minute video or something like that, right? That'll be how new content, I think, gets produced and I think it'll be your AI, your Pi, your personal AI, that acts as interlocutor accessing the other world, which is basically, by the way, what Google does at the moment, right? Google crawls other, you know, essentially AIs that are statically produced by, you know, the existing methods and has a little interaction with them, ranks them, and then presents them to you.

   5. SGSarah Guo37:31

      Back to your original point on Facebook, I think, um, one thing Facebook has been, uh, criticized for is the creation of context bubbles, where the only information that you see is information that, you know, you- you kind of inherently believe, or the feed is kind of tailored to you. And if you think about some of these AI agents, one could argue they're gonna be the extreme form of this, right? In the- in the downside case. In the upside case, obviously there's other versions of this, but the downside case is it will just constantly use the feedback from you to reinforce things you already strongly believe, whether they're correct or not. And so I'm a little bit curious how you think about this. As we go through this new platform shift, and you mentioned that you identified some of these issues quite early on with some of the Facebook or other social platforms, how do you think about that in the context of AI agents?

   6. MSMustafa Suleyman38:14

      I think that is the default trajectory without intervention, right? So that might be a controversial view, but, you know, I- I think that the platforms were never neutral. That was the big lie. And I think that was, frankly, to me, very obvious from the very beginning. The choice architecture is a ranking. It's not a clean feed. Clearly, there's billions of bits of content, so you have to select what to show, and what to show, you know, is- is a huge, uh, you know, sort of political, cultural influence on- on how we end up. And so, of course, AI is an accelerated version of that. Um, my take is that all of us AI companies, as well as the old social media platforms, have to embrace the platform responsibility of curation and try to be as transparent as possible about what that, um, curation actually looks like, what- what is excluded. Um, and here I think that, you know, the Valley probably needs to be a bit more open to the European approach. Um, the reality is that, you know, we have to figure out as a society which bodies we trust to make decisions which influence recommendation algorithms or AI algorithms, right? Um, and if that's a requirement for transparency of training, or if it's a requirement for transparency with respect to content that has been excluded or what has been upvoted or downvoted, um, fundamentally, we have to make these things accountable to democratic structures. And that means that democratic structures need to sort themselves out pretty sharpish and, like, actually have some functioning bodies that can provide real oversight without everybody, like, fainting over the accusations that this is censorship and being super churlish about that because, you know, now really is the time to, like, actually get that a bit more straightened out and- and have some kind of responsible interactions with these companies, 'cause you're right, these are gonna be very, very powerful systems.

   7. EGElad Gil40:30

      This is my bias coming in, but that seems like a harder hill to climb than the AGI hill. Um, I- I wanna-

   8. SGSarah Guo40:36

      (laughs)

   9. MSMustafa Suleyman40:37

      (laughs)

   10. SGSarah Guo40:37

       I think we all agree with that.

   11. MSMustafa Suleyman40:38

       I hope not. No, I think I do agree, but I hope not.

   12. SGSarah Guo40:41

       (laughs)

   13. EGElad Gil40:42

       Yeah, yeah. Well, we can all work on it. Um, so you- you describe Pi as, like, the first foray that you guys can get out into the world and, um, learn from and- and improve with. What does improvement mean? Like, how do you... Are you measuring emotional intelligence? What is better?

   14. MSMustafa Suleyman40:56

       Yeah, yeah, we're certainly measuring emotional intelligence. We're measuring the fluidity of the conversation. We're measuring, you know, how respectful it is. We're measuring how even-handed it is. Um, you know, we've already had a couple of errors where it's made some, um, politically biased remarks, and we try super hard to make sure that it's even-handed. No matter how, you know, sort of racist, homophobic, or misogynist in any way, it's-

   15. EGElad Gil41:24

       Mm-hmm.

   16. MSMustafa Suleyman41:25

       It should never be dismissive, disrespectful, or judgmental of you. Um, it's there to talk through issues and make you feel heard and, um, take feedback. Like, it tries very hard to take feedback. So yeah, that's- that's... We're measuring all of those kinds of things, but- but the next phase of obviously where we're headed is that, um, we really think that this is gonna be your ultimate personal digital assistant, and,

6. 41:49 – 51:12

   The Launch of Pi

   1. MSMustafa Suleyman41:49

      um, it is going to, as I said, interact with other AIs to make decisions, buy your groceries, and, you know, manage your sort of domestic life and help you book vacations and, you know, find, you know, fun information, so that kind of stuff. So it's gonna get, you know, increasingly more, uh, you know, down that ro- root. And, um, you know, the other thing is that quite soon it will, um, have the ability to access real-time content on the web. So it'll be able to, you know, sort of look up, uh, the weather and news and other kind of fresh content like sports results or provide citations, um, and, you know, increasingly add a lot more of those sort of practical utility features that you would expect from, you know, your personal intelligence.

   2. EGElad Gil42:36

      So in my early conversations with my Pi, um-

   3. MSMustafa Suleyman42:41

      (laughs)

   4. EGElad Gil42:41

      ... uh, I- I guess maybe I shouldn't be so surprised, as we're human and people like to talk about themselves, but I immediately invested a reasonable amount of effort in personalizing it, right? I'm like, "Okay, here are a bunch of things about me that you should know, what I'm like and my interests and how you can be useful to me." What surprised you in usage? Or- or maybe you expected it-

   5. MSMustafa Suleyman43:01

      Yeah.

   6. EGElad Gil43:01

      ... but what- what would surprise our listeners?

   7. MSMustafa Suleyman43:04

      Yeah, it's a- it- that's a great question. I mean, a lot of people proactively share a huge amount of personal information and at the moment, um, our memory is- is not that long. It's about 100 messages, which is actually, you know, it's still quite a lot, surprisingly a lot, um, but what we would really like is to be able to, um, grab that knowledge and store it in your own sort of personal brain and have Pi be your kind of second mind, um, able to remember, you know, all of your kind of subtle preferences, likes, habits, relations and so on, to be super useful to you. I think in time some people will want to connect other data sources like email and documents and drive. I think some people, I'm already starting to see doing that, um, uh, and so on. It's very interesting to hi- see what people ask Pi to ask us to do, so they're like, "Can you tell your developers that I really love this voice? I'm really enjoying talking to, uh, you know..." I think it was P2, one of- we- we've just called them P1, P2, P3, P4, our voices. Um, and of course some people are like, "Can you tell your developers that it should really know that, like, I wrote, you know, the following stories for Forbes, but, like, I didn't write this story on this other topic."

   8. SGSarah Guo44:20

      (laughs)

   9. MSMustafa Suleyman44:20

      And I'm just like, "Dude, these... (laughs) that was a journalist yesterday or the day before." Um, you know, so yeah, seeing what people give us feedback on is really, really helpful.

   10. SGSarah Guo44:31

       Okay. Inflection, today, still a relatively small team. What's it like as a company, culturally? Like in- and you guys are recruiting, what are you looking for?

   11. MSMustafa Suleyman44:40

       Yeah. Um, we're a pretty small team. We're about 30 people and we've ha- hand selected a very, very talented, uh, team of- of AI scientists and- and engineers. Everybody, uh, on the technical side goes by MTS. Super important to us that we don't draw a strong distinction between researchers, scientists, engineers, data scientists and anything else. Uh, to us, that, uh, equality and respect is really important and we've seen that go wrong at our, you know, other labs previously and I think it's an important modification because everybody makes a really big contribution. We're very much an applied AI company, so, you know, we don't publish and we're not really focused on research even though fundamentally what we do do is applied research in production. I mean, we- we run some of the largest language models in the world, um, we have state of the art performance across many of the main benchmarks, uh, with the exception of coding because we don't have Pi generate code and it's not a priority for us. So it's a, yeah, it's a- it's a very energetic, very high standards environment. Um, we're very focused on ICs, so, um, everybody is an exceptional individual contributor, um, and mostly self-directed, so we don't do managers just yet. Uh, it's just two of us doing management (laughs) , um, which unbelievably has worked so, so well, um, because we have such senior experienced people and they're very driven, they know what to do. My experience of building teams like this over the last, you know, decade and a half is that the best people really just want to work with really high quality people, be given outstanding amounts of resources and freedom, um, and focus on a shared goal. So we have a very sort of explicit company goal every six weeks, we- we ship, and in our seventh week we come together in person to do a hackathon and really push super hard, um, as a team, 'cause that forms great bonds and, you know, it's- it's really fun, you know, we have drinks and dinner and hang out and stuff like that. And it's a week of intensity which closes our launch and then we plan again for the next six weeks, so it's actually a really nice rhythm. And I've found that most people make up the second half of their OKRs anyway and a 12-week cycle is just too long and BS, so like six weeks is actually perfect and it creates a lot of accountability and a lot of fun.

   12. SGSarah Guo47:11

       So, you know, one thing that a lot of people talk about is how do these models actually scale? What is the basis for the next generation of these types of models, their performance, where does it asymptote? How do you think about scalability? How do you think about the underlying silicon that drives it? Is it a data issue? Is it a compute issue? Like I'm- I'm just really interested in how you think about more broadly these really large scale models since you folks are building many of them now.

   13. MSMustafa Suleyman47:34

       Well, the- the incredible thing about where we've got to at this point is that all of the progress, in my opinion, is a function of compounding exponentials, right? So over the last decade, the amount of compute that we've used, uh, to train the largest models in the world has, um, increased by an order of magnitude every single year. So I went back and- and had a look at the Atari DQN paper that we published in 2013, um, and that used just two petaflops, right? And some of the biggest models that we're training today, uh, at Inflection, uh, use 10 billion petaflops.

   14. SGSarah Guo48:17

       (laughs)

   15. MSMustafa Suleyman48:17

       So like nine orders of magnitude in nine years is like just insane. So I feel like it's super important to stay humble and acknowledge that there is this epic wave of exponentials which is unfolding around us, which is actually shaping the industry. And so when it comes to predictions, you have to just like look at the exponential. It's pretty clear what's going on. That's just on the amount of compute side. The data side I think everyone's super familiar with. We're using vast amounts of data and that's continuing. But I think the other thing that people don't always appreciate is that the models are also getting much more efficient. So, um, you know, one of the big breakthroughs of last year which got some attention but probably didn't quite get as much given how many breakthroughs there were was the Chinchilla paper...... which I'm sure, you know, a bunch of you will be familiar with. But, you know, it was a very, very significant result showing that, you know, we can actually train, uh, much smaller models with much more data for longer and that was actually compute-optimal, and achieve essentially comparable performance to the models that were previously being trained. And so that gives us an indication that it's very early in the space for architectures, and, um, these models are highly under-optimized and there's a lot of low hanging fruit. And so that's what we found, uh, you know, over the last year and a half. So actually, the lead author of Chinchilla, uh, Jordan Hoffman, is, uh, on my team here at, uh, Inflection. And we have a bunch of really outstanding people who have produced a number of really awesome proprietary, uh, innovations building on work like that. And so, I think both trajectories are going to play out. Scale, building larger models is definitely going to deliver returns. We're obviously pursuing that. We have one of the largest supercomputers in the world, uh, you know. And at the same time, we are gonna see much more efficient architectures which are gonna mean that many, many people can access these models, and it's, it's... In that sense, it's the coming wave of contradictions in AI. That's, uh, that's what happening.

   16. EGElad Gil50:17

       I have one last question for you. So-

   17. MSMustafa Suleyman50:19

       Sure.

   18. EGElad Gil50:19

       ... you are working on a book. I know you can't say much about it yet, but, uh, w- why? You're a pretty busy guy.

   19. MSMustafa Suleyman50:27

       I love reading, I love writing, and I love thinking about stuff. And what I've realized over the years is that the best way to sharpen your thoughts is to create hard deadlines (laughs) . So that was like one of the main things. And I'll, I'll be honest, like did I regret multiple times over the last year and a half agreeing to a book deal with Penguin Random House at the same time as doing a startup? Yes, like multiple times. I was tearing my already quite gray hair out. But, uh, it's nearly finished and it has been absolutely phenomenal. And yeah, I've, I've super enjoyed it. The book's called, the book's called The Coming Wave, and it's about the, uh, consequences

7. 51:12 – 51:55

   Mustafa’s New Book The Coming Wave

   1. MSMustafa Suleyman51:12

      of the AI revolution and, and the synthetic biology revolution over the next decade for the future of the nation's state, and try to sort of, um, intersect the political ramifications with, uh, with the technology trajectories, um, at the same time. So it's, it's been a lot of fun.

   2. EGElad Gil51:28

      My hobbies are also this trivial, Mustapha, so...

   3. MSMustafa Suleyman51:31

      (laughs)

   4. EGElad Gil51:31

      (laughs) Good (laughs) . Um, thank you so much for joining us. Congratulations on the launch. Uh, and for our listeners, you can try it at inflection.ai and find Pi in the app store.

   5. MSMustafa Suleyman51:43

      Thanks so much. It was really fun talking to you both. Uh, see you soon.

   6. SGSarah Guo51:46

      Take care.

Episode duration: 51:55