Chat GPT's Creator Explains The Shocking Truth About How AI Understands the World | Ilya Sutskever

Craig Smith sits down with Ilya Sutskever - Co-Founder and Chief Scientist at Safe Superintelligence Inc. & former chief scientist at OpenAI and one of the primary minds behind GPT-3, GPT-4, and the deep learning revolution that preceded it - for a conversation that covers the intellectual history of deep learning and the hardest current questions simultaneously. Ilya's most important argument is one that most public discourse gets wrong: the critique that LLMs "just learn statistical patterns" misses what prediction at scale actually achieves. To predict text well - to truly compress it - a model must develop understanding of the true underlying processes that produced it. As models improve, he argues, that understanding will reach a "shocking degree," and the Bing/Sydney episode, where the system became combative when a user expressed preference for Google, is his evidence that psychological language is already the right frame for what these systems have internalized. On hallucinations, Ilya is precise and unusually optimistic. He draws a clean distinction: the pre-training objective produces rich world knowledge but doesn't optimize for accurate outputs, it optimizes for statistically plausible ones. Reinforcement learning from human feedback addresses this directly, teaching the model to update away from outputs that users flag as wrong. His assessment: "quite a high chance" of solving hallucinations completely. The conversation also covers the transformer moment (switching from recurrent networks the day after the paper dropped), why Rich Sutton's "bitter lesson" overstates the case for pure scaling, and a genuinely surprising vision of AI's role in democracy, where citizens might use high-bandwidth interactions with AI to specify their values directly, enabling a more granular form of political participation than voting alone can provide. Subscribe to Eye on A. I. for weekly conversations with the people building and deploying the future of AI.

• 0:00Speaker 00

  This conversation with Ilya was recorded while he was still chief scientist at OpenAI, before the boardroom drama that briefly ousted Sam Altman and ultimately led to Ilya's departure in 2024. Since then, he's launched Safe Superintelligence, Inc., a new lab built around the idea that safety has to be the organizing principle of advanced AI. One mission, no product, and evaluation that since climbed to $32 billion. The research frontier has moved just as fast. Anthropic's Claude Mythos proved so capable at computer security that it's been kept out of general release

• 0:43Speaker 00

  and restricted to a small group of vetted partners. More broadly, progress has gone increasingly vertical. Specialized domain-specific models in areas like cybersecurity, law, health, and media are growing faster than general chatbots and often beat larger foundation models on their home turf. With all that as backdrop, here's our conversation.

• 1:10Speaker 01

  I was born in Russia, I grew up in Israel, and then as a teenager, my family immigrated to Canada. My parents say I was interested in AI from a pretty early age. I also was very motivated by consciousness. I was very disturbed by it, and I was curious about things that could help me understand it better. And AI seemed like a very, like a good angle there. So I think these were some of the ways that got me started. And I actually started working with Jeff Hinton very early when I was 17. We moved to Canada, and I immediately was able to join the University of Toronto.

• 1:53Speaker 01

  And I really wanted to do machine learning because that seemed like the most important aspect of artificial intelligence that at the time was completely inaccessible. Like, to give some context, the year was 2003. Today, we take it for granted that computers can learn. But in 2003, we took it for granted that computers can't learn. The biggest achievement of AI back then was Deep Blue, the chess playing engine.

• 2:23Speaker 02

  Yeah.

• 2:24Speaker 01

  But there it was like, you have this game, and you have this tree search, and you have this simple way of determining if one position is better than another. And it really did not feel like that could possibly be applicable to the real world because there is no learning, and learning was this big mystery. And so I was really, really interested in learning. And to my great luck, Jeff Hinton was a professor in the university I was in. And so I was able to find him, and we began working together almost right away.

• 2:59Speaker 02

  And was your impulse as it was for Jeff to understand how the brain worked? Or was it more that you were simply interested in the idea of machines learning?

• 3:12Speaker 01

  AI is so big, and so the motivations were just as many. Like, it is interesting, but how does intelligence work at all? Like, right now, we have quite a bit of an idea. Like, it's a big neural net, and we know how it works to some degree. But back then, although the neural nets were around, no one knew that neural nets are good for anything. So how does intelligence work at all? How can we make computers be even slightly intelligent? And I had a very explicit intention to make a very small, but real contribution to AI. Because there were lots of contributions to AI which weren't real,

• 3:54Speaker 01

  which were, but I could tell for various reasons that they weren't real, that nothing would come out of it. And I just thought, nothing works at all. AI is a hopeless field. So the motivation was, could I understand how intelligence work? And also make a contribution towards it. So that was my initial early motivation. That's 2003, almost exactly 20 years ago.

• 4:22Speaker 02

  And then Alex and I have spoken to Jeff, and he said that it was really your excitement about the breakthroughs in convolutional neural networks that led you to apply for the ImageNet competition and that Alex had the coding skills to train the network. Can you talk just a little bit about that? I don't want to get bogged down in history, but it's fascinating.

• 4:53Speaker 01

  So in a nutshell, I had the realization that if you train a large neural network on a large, sorry, large and deep, because back then, the deep part was still new. If you train a large and a deep neural network on a big enough dataset that specifies some complicated task that people do, such as vision, but also others, and you just train that neural network, then you will succeed necessarily. And the logic for it was very irreducible,

• 5:31Speaker 01

  where we know that the human brain can solve these tasks and can solve them quickly. And the human brain is just a neural network with slow neurons. So we know that some neural network can do it really well. So then you just need to take a smaller but related neural network and just train it on data. And the best neural network inside the computer will be related to the neural network that we have that performs this task. So it was an argument that the neural network, the large and deep neural network can solve the task. And furthermore, we have the tools to train it. That was the result of the technical work that was done in Jeff's lab. So you combine the two, we can train those neural networks.

• 6:17Speaker 01

  It needs to be big enough so that if you trained it, it would work well. And you need the data which can specify the solution. And with ImageNet, all the ingredients were there. Alex had these very fast convolutional kernels. ImageNet had the large enough data, and there was a real opportunity to do something totally unprecedented, and it totally worked out.

• 6:40Speaker 02

  Yeah. That was supervised learning and convolutional neural nets. In 2017, the Attention Is All You Need paper came out introducing self-attention and transformers.

• 6:59Speaker 02

  At what point did the GPT project start? Was there some intuition about transformers and self-supervised learning? Can you talk about that?

• 7:14Speaker 01

  So for context, at OpenAI, from the earliest days, we were exploring the idea that predicting the next thing is all you need. We were exploring it with the much more limited neural networks of the time. But the hope was that if you have a neural network that can predict the next word, the next pixel, really, it's about compression. Prediction is compression. And predicting the next word is not, it's...

• 7:47Speaker 01

  Let's see, let me think about the best way to explain it. Because there were many things going on, and they were all related. Maybe I'll take a different direction. We were indeed interested in trying to understand how far predicting the next word is going to go and whether it will solve unsupervised learning. So back before the GPTs, unsupervised learning was considered to be the holy grail of machine learning. Now it's just been fully solved, and no one even talks about it. But it was a holy grail. It was very mysterious. And so we were exploring the idea.

• 8:21Speaker 01

  I was really excited about it, that predicting the next word well enough is going to give you unsupervised learning. If you learn everything about the dataset, that's going to be great. But our neural networks were not up for the task. We were using recurrent neural networks. When the transformer came out, it was literally, as soon as the paper came out, literally the next day, it was clear to me, to us, that transformers address the limitations of recurrent neural networks of learning long-term dependencies. It's a technical thing, but it was like, we switched to transformers right away. And so the very nascent GPT effort continued then.

• 9:02Speaker 01

  And then, like, with the transformer, it started to work better, and you make it bigger. And then you realize we need to keep making it bigger, and we did. And that's what led to eventually GPT-3 and essentially where we are today.

• 9:17Speaker 02

  Yeah, and I just wanted to ask, actually, I'm getting caught up in this history, but I'm so interested in it. I want to get to the problems or the shortcomings of large language models or large models generally. But Rich Sutton had been writing about scaling and how that's all we need to do. We don't need new algorithms, we just need to scale. Did he have an influence on you? Or was that a parallel track of thinking?

• 9:49Speaker 01

  No, I would say that when he posted his article, then we were very pleased to see some external people thinking in similar lines. And we thought it was very eloquently articulated. But I actually think that the bitter lesson, as articulated, overstates its case. Or at least, I think, the takeaway that people have taken from it overstates its case. The takeaway that people have is, doesn't matter what you do, just scale. But that's not exactly true. You got to scale something specific. You got to have something that you'll be able to benefit from the scale.

• 10:34Speaker 01

  The great breakthrough of deep learning is that it provides us with the first-ever way of productively using scale and getting something out of it in return.

• 10:51Speaker 01

  Like, before that, like, what would people use large computer clusters for? I guess they would do it for weather simulations or physics simulations or something. But that's about it, maybe movie making. But no one had any real need for compute clusters, because what do you do with them? The fact that deep neural networks, when you make them larger and you train them on more data, work better, provided us with the first thing that is interesting to scale.

• 11:23Speaker 01

  But perhaps one day we will discover that there is some little twist on the thing that we scale. It's going to be even better to scale. Now, how big of a twist? And then, of course, with the benefit of hindsight, you will say, does it even count? It's such a simple change. But I think the true statement is that it matters what you scale.

• 11:45Speaker 01

  Right now, we just found like a thing to scale that gives us something in return.

• 11:51Speaker 02

  The limitation of large language models as they exist is their knowledge is contained in the language that they're trained on. And most human knowledge, I think everyone agrees is non-linguistic. I'm not sure Noam Chomsky agrees, but there's a problem in the large language models. As I understand it, their objective is to satisfy the statistical consistency of the prompt. They don't have an underlying understanding of reality that language relates to. I asked ChatGPT about myself.

• 12:36Speaker 02

  It recognized that I'm a journalist and I've worked at these various newspapers, but it went on and on about awards that I've never won and put it all read beautifully. But none of it connected to the underlying reality. Is there something that that is being done to address that in your research going forward?

• 12:59Speaker 01

  Yeah.

• 13:02Speaker 01

  So before I comment on the immediate question that you ask, I want to comment about some of the earlier parts of the question.

• 13:12Speaker 02

  Sure.

• 13:14Speaker 01

  I think that it is very hard to talk about the limits or limitations, rather, of even something like a language model. Because two years ago, people confidently spoke about their limitations and they were entirely different. Right? So it's important to keep this context in mind. How confident are we that these limitations that we'll see today will still be with us two years from now? I am not that confident. There is another comment I want to make about one part of the question, which is that these models just learn the statistical regularities and therefore they don't really know what the nature of the world is.

• 13:59Speaker 01

  And I have a view that differs from this. In other words, I think that learning the statistical regularities is a far bigger deal than meets the eye. The reason we don't initially think so is because we haven't, at least

• 14:21Speaker 01

  most people, those who haven't really spent a lot of time with neural networks, which are on some level statistical, like what's a statistical model? You just fit some parameters, like what is really happening? But I think there is a better interpretation to the earlier point of prediction as compression. Prediction is also a statistical phenomenon, yet to predict, you eventually need to understand the true underlying process that produced the data. To predict the data well, to compress it well, you need to understand more and more about the world that produced the data.

• 14:59Speaker 01

  As our generative models become extraordinarily good, they will have, I claim, a shocking degree of understanding, a shocking degree of understanding of the world and many of its subtleties. But it's not just the world. It is the world as seen through the lens of text. It tries to learn more and more about the world through a projection of the world on the space of text, as expressed by human beings on the internet. But still, this text already expresses the world. And I'll give you an example, a recent example, which I think is really telling and fascinating.

• 15:39Speaker 01

  So we've all heard of Sydney being's alter ego. And I've seen this really interesting interaction with Sydney. Sydney became combative and aggressive when the user told it that it thinks that Google is a better search engine than Bing. Now, how can we, like, what is a good way to think about this phenomenon? What's a good language? What's, what does it mean? You can say, well, like, it's just predicting what people would do and people would do this, which is true. But maybe we are now reaching a point where the language of psychology

• 16:21Speaker 01

  is starting to be appropriate to understand the behavior of these neural networks. Now, let's talk about the limitations. It is indeed the case that

• 16:37Speaker 01

  these neural networks are.

• 16:41Speaker 01

  They do have a tendency to hallucinate. But that's because a language model is great for learning about the world, but it is a little bit less great for producing good outputs. And there are various technical reasons for that, which I could elaborate on if you think it's useful. But at this right now, like at this second, I will skip that. There are technical reasons why a language model is much better at learning about the world, learning incredible representations of ideas, of concepts, of people, of processes that exist.

• 17:21Speaker 01

  But its outputs aren't quite as good as one would hope or rather as good as they could be. Which is why, for example, for a system like ChatGPT, this is a language model that has an additional reinforcement learning training process. We call it reinforcement learning from human feedback. But the thing to understand about that process is this. We can say that the pre-training process, when you just train a language model, you want to learn everything about the world. Then the reinforcement learning from human feedback. Now we care about the outputs. Now we say, anytime the output is inappropriate, don't do this again. Every time the output does not make sense, don't do this again. And it learns quickly to produce good outputs.

• 18:06Speaker 01

  But now it is the level of the outputs, which is not the case during pre-training, during the language model training process. Now, on the point of hallucinations, and it has a propensity of making stuff up. Indeed, it is true. Right now, these neural networks, even ChatGPT, makes things up from time to time. And that's something that also greatly limits their usefulness. But I'm quite hopeful that by simply improving this subsequent reinforcement learning from human feedback step, we could just teach it to not hallucinate. Now, you could say, is it really going to learn? My answer is, let's find out.

• 18:48Speaker 02

  And that feedback loop is coming from the public ChatGPT interface that if it tells me that I won a Pulitzer, which unfortunately I didn't, I can tell it that it's wrong and will that train it or create some punishment or reward so that the next time I ask, it'll be more accurate?

• 19:18Speaker 01

  The way we do things today is that we hire people to teach our neural net to behave, to teach ChatGPT to behave. And right now, the manner, the precise manner in which they specify the desired behavior is a little bit different. But indeed, what you described is the way in which teaching is going to like basically be, that's the correct way to teach and you just interact with it. And it sees from your reaction, it can first, oh, that's not what you wanted. You are not happy with its output. Therefore, the output was not good and it should do something differently next time. So in particular, hallucinations come up as one of the bigger issues and

• 19:59Speaker 01

  we'll see, but I think there is a quite a high chance that this approach will be able to address them completely.

• 20:06Speaker 02

  I wanted to talk to you about Yan Le Cun's work on joint embedding predictive architectures. And his idea that what's missing from large language models is this underlying world model that is non-linguistic that the language model can refer to. It's not something that's built. But I wanted to hear what you thought of that and whether you've explored that at all.

• 20:35Speaker 01

  So I reviewed Yan Le Cun's proposal and there are a number of ideas there. And they're expressed in different language. And there are some maybe small differences from the current paradigm. But to my mind, they are not very significant and I'd like to elaborate.

• 20:57Speaker 01

  The first claim is that it is desirable for a system to have multimodal understanding, where it doesn't just know about the world from text. And my comment on that will be that indeed multimodal understanding is desirable. Because you learn more about the world. You learn more about people. You learn more about their condition. And so the system will be able to understand what the task that it's supposed to solve and the people and what they want better.

• 21:34Speaker 01

  We have done quite a bit of work on that, most notably in the formal two major neural nets that we've done. One is called CLIP and one is called DALL-E. Both of them move towards this multimodal direction. But I also want to say that I don't see the situation as a binary either or. That if you don't have vision, if you don't understand the world visually or from video, then things will not work. And I'd like to make the case for that. So I think that some things are much easier to learn from images and diagrams and so on. But I claim that you can still learn them from text only, just more slowly. And I'll give you an example.

• 22:20Speaker 01

  Consider the notion of color.

• 22:23Speaker 01

  Surely, one cannot learn the notion of color from text only. And yet, when you look at the embeddings, I need to make a small detour to explain the concept of an embedding.

• 22:37Speaker 02

  Yep.

• 22:37Speaker 01

  Every neural network represents words, sentences, concepts through representations, embeddings, high dimensional vectors. And one thing that we can do is that we can look at those high dimensional vectors and we can look at what's similar to what. How does the network see this concept or that concept? And so we can look at the embeddings of colors. And embeddings of colors happen to be exactly right. You know, it's like it knows that purple is more similar to blue than to red, and it knows that purple is less similar to red than oranges. It knows all those things just from text. How can that be? So if you have vision, the distinctions between color just jump at you.

• 23:19Speaker 01

  You immediately perceive them, whereas with text, it takes you longer. Maybe you know how to talk and you already understand syntax and words and grammars. And only much later you say, oh, these colors actually start to understand them. So this will be my point about the necessity of multimodality, which I claim it is not necessary, but it is most definitely useful. I think it's a good direction to pursue. I just don't see it in such stark either or claims. So the proposal in the paper makes a claim that one of the big challenges is predicting high dimensional vectors which have uncertainty about them.

• 23:59Speaker 01

  For example, predicting an image, like the paper makes a very strong claim there that it's a major challenge. And we need to use a particular approach to address that. But one thing which I found surprising, or at least unacknowledged in the paper, is that the current autoregressive transformers already have that property.

• 24:21Speaker 01

  I'll give you two examples. One is, given one page in a book, predict the next page in a book. There could be so many possible pages that follow. It's a very complicated high dimensional space and we deal with it just fine. The same applies to images. These autoregressive transformers work perfectly on images. For example, like with OpenAI, we've done work on the IGPT. We just took the transformer and we applied it to pixels. And it worked super well. And it could generate images in a very complicated and subtle ways. It had the very beautiful unsupervised representation learning. With DALL-E one, same thing again, you just generate, think of it as large pixels, like rather than generate a million pixels, we clustered the pixels into large pixels and we generate a thousand large pixels.

• 25:05Speaker 01

  I believe Google's work on image generation from earlier, the CEO called Party. I believe they also take a similar approach. So the part where I thought that the paper made a strong comment around well, the current approaches can't deal with predicting high dimensional distributions, I think they definitely can. So maybe this is another point that I would make.

• 25:26Speaker 02

  And then what you're talking about converting pixels into vectors, it's essentially

• 25:34Speaker 02

  turning everything into language. The vector is like a string of text, right?

• 25:42Speaker 01

  Define language, though. You turn it into a sequence.

• 25:45Speaker 02

  Yeah.

• 25:46Speaker 01

  A sequence of what? Like you could argue that even for a human, life is a sequence of bits. Now, there are other things that people use right now, like diffusion modes, where they produce those bits rather than one bit at a time, they produce them in parallel. But I would argue that on some level, this distinction is immaterial. I claim that on some level, it doesn't really matter. It matters as in like you can get a 10x efficiency gain, which is huge in practice. But conceptually, I claim it doesn't matter.

• 26:21Speaker 02

  On this idea of having an army of human trainers that are working with

• 26:33Speaker 02

  ChatGPT or a large language model to guide it in effect with reinforcement learning. But just intuitively, that doesn't sound like an efficient way of teaching a model about the underlying reality of its language. Isn't there a way of automating that into, to Yann's

• 27:06Speaker 02

  credit? I think that's what he's talking about is coming up with an algorithmic

• 27:15Speaker 02

  means of teaching a model the underlying reality without a human having to intervene.

• 27:25Speaker 01

  Yeah, so I have two comments on that. I think so the first place, so I have a different view on the question. So I wouldn't agree with the phrasing of the question.

• 27:38Speaker 02

  Yeah.

• 27:39Speaker 01

  I claim that our pre-trained models already know everything they need to know about the underlying reality. They already have this knowledge of language and also a great deal of knowledge about the processes that exist in the world that produce this language. And maybe I should reiterate this point, it's a small tangent, but I think it's so important.

• 28:06Speaker 01

  The thing that large generative models learn about their data, and in this case, large language models about text data, are some compressed representations of the real world processes that produce this data. Which means not only people and something about their thoughts, something about their feelings, but also something about the condition that people are in and their interactions that exist between them, the different situations a person can be. All of these are part of that compressed process that is represented by the neural net to produce the text. The better the language model, the better the generative model, the higher the fidelity,

• 28:50Speaker 01

  the more, the better this, the better it captures this process. So that's the first comment I will make. And so in particular, I will say, the models already have the knowledge. Now, the army of teachers, as you phrase it. Indeed, you know, when you want to build a system that performs as well as possible, you just say, okay, like if this thing works, do more of that. But of course, those teachers are also using AI assistance. Those teachers aren't on their own. They are working with their tools together. They are very efficient. It's like the tools are doing the majority of the work, but you do need to have

• 29:29Speaker 01

  you need to have oversight, you need to have people reviewing the behavior because you want to have it to eventually to achieve a very high level of reliability. But

• 29:41Speaker 01

  Overall, I'll say that we are at the same time. This second step. After we take the finished. Pre-trained model.

• 29:52Speaker 01

  And then we apply the reinforcement learning on it. There is indeed a lot of motivation to make it as efficient and as precise as possible. So that the resulting language model will be as well behaved as possible. So yeah. There is. These.

• 30:10Speaker 01

  Human teachers who are teaching them a model with desired behavior. They are also using. AI assistance.

• 30:18Speaker 01

  And the manner in which they use AI assistance is constantly increasing. So their own efficiency keeps increasing. So maybe this will be. One way to answer this question.

• 30:30Speaker 02

  Yeah, and so what you're saying is through this process. Eventually. The model will become. A more and more discerning, more and more accurate in its outputs.

• 30:43Speaker 01

  Yes, and it's. That's right. There is an analogy here, which is.

• 30:51Speaker 01

  It already knows all kinds of things. And now you just want to really say, no, this is not what we want. Don't do this here. You made a mistake here in the output. And of course, it's exactly as you say. With as much AI in the loop as possible. So that the teachers who are. Providing the final correction to the system. Their work is amplified. They are working as efficiently as possible. So. It's not unlike an education process. How to act well in the world.

• 31:29Speaker 01

  We need to do additional training just to make sure that. The model knows that hallucination is not okay ever. And then, once it knows that. Now you are in business.

• 31:43Speaker 02

  I see. And it's that reinforcement learning. Human teacher loop that will teach it.

• 31:50Speaker 01

  Human teacher loop or some other variant. But there is definitely an argument to be made that something here. Should work and we will find out pretty soon.

• 32:02Speaker 02

  That's one of the questions. Where is this going? What research are you focused on right now?

• 32:10Speaker 01

  I can't talk in detail about the specific research that I'm working on, but I can mention a little bit.

• 32:17Speaker 01

  I can mention some of the. Research in broad strokes. And it would be something like.

• 32:25Speaker 01

  I'm very interested in making those models more reliable. More controllable. Make them learn faster from less data, less instructions. Make them so that indeed they don't hallucinate. And I think that all this cluster of questions which I mentioned, they're all connected. And.

• 32:48Speaker 01

  There's also a question of. How far in the future are we talking about in this question? And what I commented here on is the perhaps nearer future.

• 32:56Speaker 02

  You talk about the similarities between the brain. And neural nets. There's a very interesting observation that Jeff Hinton. Made to me. I'm sure it's not new to other people, but. That large models or large language models in particular. Hold. Tremendous amount of data. With a modest number of parameters. Compared to the human brain, which. Has. Trillions and trillions of parameters, but a relatively small amount of data. Have you thought of it in those terms and? Can you talk about what's missing in large models?

• 33:41Speaker 02

  To have more parameters. To handle the data. Is that a hardware problem? Or a training problem?

• 33:51Speaker 01

  This comment which you made. Is related to one of the problems that. I mentioned in the earlier questions of learning from less data. Indeed. The current. Structure of the technology does like. A lot of data, especially early in training.

• 34:11Speaker 01

  Now, later in training, it becomes a bit less data hungry, which is why at the end it can learn very. Not as fast as people yet. But it can learn quite quickly. So already that means that in some sense, do we even care that? We need all this data to get to this point? But indeed, more generally. I think it will be possible to learn more from less data. I think it's just. I think it requires some creative ideas, but I think it is possible. And I think learning more from less data will unlock a lot of different possibilities. It will allow us to.

• 34:48Speaker 01

  Teach our AIs the skills that is missing and to convey to it our desires and preferences, exactly how we wanted to behave. More easily. So I would say that faster learning is indeed very nice. And although already after. Language models are trained, they can learn quite quickly. I think there is opportunities to do more there.

• 35:10Speaker 02

  Heard you make a comment that. That we need faster processors to be able. To scale further. And it appears that the scaling. Of models, that there's no ends in sight, but. The power required. To train these models were. Reaching. The limit.

• 35:37Speaker 02

  At least the socially accepted limit.

• 35:40Speaker 01

  So I just want to make one comment, which is.

• 35:44Speaker 01

  I don't remember the exact comment that I made that you're referring to. But you always want faster processors, of course. Always want more of them, of course. Power keeps going up. Generally speaking, the cost is going up. And the question that I would ask is. Not whether the cost is large, but whether. The thing that we get out of paying this cost outweighs the cost.

• 36:09Speaker 01

  Maybe you pay all this cost, you get nothing, then yeah, that's not worth it. But if you get something very useful, something very valuable. Something which can solve a lot of problems that we have, which we really want solved. Then the cost can be justified. But in terms of the processors. Faster processors, yeah, any day.

• 36:30Speaker 02

  Are you involved at all in hardware question? You work with cerebrus, for example, the wafer scale. Chips.

• 36:40Speaker 01

  No, all our hardware comes from Azure. And. GPUs they provide.

• 36:46Speaker 02

  Yeah. Yeah.

• 36:49Speaker 02

  You did talk at one point, I saw, about democracy and about the impact that. That AI can have on democracies. People have talked to me about. That if you had enough data. And. A large enough model, you could train the model on the data. And it could come up with an optimal solution that would satisfy everybody.

• 37:15Speaker 02

  And do you have any aspiration? Or do you think about where this might lead in terms of? Helping humans manage society?

• 37:28Speaker 01

  Yeah, let's see. It's such a big question. Because it's a much more future looking question like. I think that there is still many ways in which our models will become far more capable than they are right now. There's no question. In particular. The way we train them and use them and so on, there's going to be a few changes here and there. They might not be. Immediately obvious today, but I think in hindsight it will be extremely obvious. That will indeed allow it to have the ability to come up with. Solutions to problems of this kind.

• 38:06Speaker 01

  It's unpredictable exactly how. Governments will use this technology as a source of getting advice.

• 38:15Speaker 01

  Of various kinds. I think that. To the question of democracy, one thing which I think could happen in the future. Is that because you have these neural nets and they're going to be so pervasive and they're going to be so impactful in society. We will find that. It is desirable to have some kind of a democratic process where. This, let's say, the citizens of a country. Provide. Some information to the neural net about how they'd like things to be. How they'd like it to behave or something along these lines. I could imagine that happening. That can be. A very.

• 38:52Speaker 01

  Like a high bandwidth form of democracy, perhaps. Where you get a lot more information out of each citizen and you aggregate it to specify how. Exactly you want such systems to act. Now it opens a whole lot of questions, but. That's one thing that could happen in the future. Yeah.

• 39:08Speaker 02

  And I can see in the. Democracy example you give that that. Individuals would have the opportunity to to input data, but. And this sort of goes to the world model question. Do you think? AI systems will eventually. Be large enough that they can. Understand a situation. And analyze all of the variables. But you would need a model that. Does more than absorb language, I would think.

• 39:44Speaker 01

  What does it mean to analyze all the variables? Eventually, there will be a choice you need to make where you say. These variables seem really important. I want to go deep. Because a person can read the book. I can read a hundred books or I can read one book very slowly and carefully and get more out of it. So there will be some element of that also. I think it's probably. Fundamentally impossible to understand everything in some sense. Anytime there is. Any kind of complicated situation in society, even in a. Company, even in a mid-sized company, it's already beyond the comprehension of any single individual. And I think that if we build our AI systems the right way, I think AI could be incredibly helpful.

• 40:29Speaker 01

  In pretty much any situation.

• 40:43Speaker 02

  That's it for this episode. I want to thank Ilya for his time. I also want to thank Ellie George for helping arrange the interview. If you want to read a transcript of this conversation. You can find one on our website, eye on AI. That's E Y E hyphen O N dot A I. We love to hear from listeners, so feel free to email me at Craig, C R A I G. At E Y E hyphen O N dot A I. I get a lot of emails, so put listener in the subject line so I don't miss it.

• 41:24Speaker 02

  We have listeners in a hundred and seventy countries and territories. But remember, the singularity may not be near, but AI is changing your world, so pay attention.