Dwarkesh Podcast - Ilya Sutskever (OpenAI Chief Scientist) — Why next-token prediction could surpass human intelligence

Episode Date: March 27, 2023

I went over to the OpenAI offices in San Fransisco to ask the Chief Scientist and cofounder of OpenAI, Ilya Sutskever, about:* time to AGI* leaks and spies* what's after generative models* post AGI futures* working with Microsoft and competing with Google* difficulty of aligning superhuman AIWatch on YouTube. Listen on Apple Podcasts, Spotify, or any other podcast platform. Read the full transcript here. Follow me on Twitter for updates on future episodes.Timestamps(00:00) - Time to AGI(05:57) - What’s after generative models?(10:57) - Data, models, and research(15:27) - Alignment(20:53) - Post AGI Future(26:56) - New ideas are overrated(36:22) - Is progress inevitable?(41:27) - Future Breakthroughs Get full access to Dwarkesh Podcast at www.dwarkesh.com/subscribe

But I would not underestimate the difficulty of alignment of models that are actually smarter than us, of models that are capable of misrepresenting their intentions. Are you worried about spies? I'm really not worried about the way it's being leaked. We'll all be able to become more enlightened because we interact with an AGI that will help us see the world more correctly. Like, imagine talking to the best meditation teacher in history. Microsoft has been a very, very good partner for us. So I challenge the claim that next token prediction cannot surpass human performance.

If your base neural net is smart enough, you just ask it like, what would a person with great insight and wisdom and capability do? Okay, today I have the pleasure of interviewing Elia Sutskaver, who is the co-founder and chief scientists of OpenA.I. Elia, welcome to the Lunar Society. Thank you. Happy to be here. First question, and no humility allowed. there's many scientists or maybe not that many scientists who will make a big breakthrough in their field

there's far fewer scientists who will make multiple independent breakthroughs that define their field throughout their career. What is the difference? What distinguishes you from other researchers? Why have you been able to think multiple breakthroughs in our field?

Well, thank you for the kind words. It's hard to answer that question. I mean, I try really hard, I gave it everything you got. And that was, worked so far. I think that's all there is to it. Got it. What's the explanation for why there aren't more illicit uses of GPT? Why aren't more foreign governments using it to spread propaganda or scam grandmothers or

something? I mean, maybe they haven't really gotten to do it a lot. But it also wouldn't surprise me if some of it was going on right now. Certainly I imagine they'd be taking some of the open source models and trying to use them for that purpose. Like, I'd be. sure I would expect this would be something that'd be interested in in the future. It's like technically possible. They just haven't thought about it enough. Or haven't like done it at scale using their technology. Or maybe it's happening which is don't know it.

Would you be able to track it if it was happening? I think large scale tracking is possible. Yes. I mean, it requires of all special operation that is possible. Mm-hmm. Now there's some window in which AI is very economically valuable on the scale of airplanes, let's say. But we haven't reached AGIEI yet. How big is that window?

I mean, I think this window, it's hard to give you a precise answer, but it's definitely going to be like a good multi-year window. It's also a question of definition, because AI, before it becomes AGI, is going to be increasingly more valuable year after year. I'd say in an exponential way. So in some sense, it may feel like, especially in hindsight, it may feel like there was only one year or two years

because those two years were larger than the previous years. But I would say that already last year, there have been a fair amount of economic value produced by AI. Next year is going to be larger and larger after that. So I think that this is going to be a good multi, multi year chunk of time. But that's going to be true. I would say from now till AGI pretty much. Okay. Well, because I'm curious if there's a startup that's using your models, right?

At some point, if you have AGI, there's only one business in the world, right? It's open AI. How much window do they have, does any business have, where they're actually producing something that AGI can't produce? Yeah, well, I mean, it's the same question as asking how long until AGI? Yeah. I think it's a hard question to answer.

I mean, I hesitate to give you a number. Also because there is this thing where effect, where people who are optimistic people, who are working on the technology, tend to underestimate the time it takes to get there. But I think that the way I ground myself is by thinking about the self-driving car. In particular, there is an analogy

where if you look at the type of a Tesla and if you look at the self-driving behavior of it, it looks like it does everything. It does everything. All right. But it's also clear that there is still a long way to go in terms of reliability.

And we might be in a similar place with respect to our models where it also looks like we can do everything. And at the same time, it will be, we'll need to do some more work until we really iron out all the issues and make it really good and really reliable. and robust and well-behaved.

By 2030, what percent of GDP is AI? Oh, gosh, hard to answer that question. Very hard to answer the question. Give me an over-under. Like, the problem is that my error bars are in lock scale. So I could imagine, like, I could imagine, like, a huge percentage. I could imagine, like, disappointing a small percentage at the same time. Okay, so let's take the counterfactual where it is a small percentage.

Let's say it's 2030 and, you know, not that much economic value has been created by these elements. As unlikely as you think this might be, what is, what would be your best definition right now, or why something like this might happen. My best explanation. So I really don't think that's a likely possibility. So that's the preface to the comment. But if I were to take the premise of your question, well, like, why were things disappointing

in terms of the real world impact? And my answer would be reliability. If somehow it ends up being the case that you really want them to be reliable and they ended up not being reliable or if reliability now to be harder than we expect. I really don't think that will be the case, but if I had to pick one,

if I had to pick one and you tell me like, hey, like, why didn't things work out? It would be reliability that you still have to look over the answers and double-check everything. That just really puts a damper on the economic value that can be produced by those systems. They'll be technologically mature. It's just a question of whether it will be reliable enough.

Yeah, well, in some sense, not reliable means not technologically mature. You see what I mean. Yeah, fair enough. What's after generative models, right? So before you're working on reinforcement learning, is this, is this basically it? Is this a paradigm that gets us to AGI or is there something after this? I mean, I think this paradigm is going to go really, really far and I would not underestimate it. I think it's quite likely that this exact paradigm is not going to be the quiet AGI form factor. I mean, I hesitate to say precisely what the next paradigm will be. but I think it will probably involve integration of all the different ideas that came in the past. Is there some specific one you're referring to? I mean, it's hard to be specific.

So you could argue that next token prediction can only help us match human performance and maybe not surpass it. What would it take to surpass human performance? So I challenge the claim that next token prediction cannot surpass human performance. It looks like on the surface, it cannot. It looks on the surface

if you just learn to imitate to predict what people do, it means that you can only copy people. But the here is a contra argument for why it might not be quite so if your neural net is, if your base neural net is smart enough, you just ask

it like, what would a person with great insight and wisdom and capability do? Maybe such person doesn't exist, but there's a pretty good chance that the neural net will be able to extrapolate how such a person could behave. Do you see what I mean?

Yes, although where would it get that sort of insight about what that person would do, if not from? From the data of regular people. Because if you think about it, what does it mean to predict the next token well enough? What does it mean actually? It's actually a much, it's a deeper question than it seems. Predicting the next token well means that you understand

the underlying reality that led to the creation of that token. of that token. It's not statistics. Like it is statistics, but what is statistics? In order to understand those statistics, to compress them, you need to understand what is it about the world that creates those statistics. And so then you say, okay, well, I have all those people.

What is it about people that creates their behaviors? Well, they have, you know, they have thoughts and they have feelings and they have ideas and they do things in certain ways. All of those would be deduced. from next token prediction. And I'd argue that this should make it possible, not indefinitely, but to a pretty decent degree to say, well, can you guess what you do if you took a person

with like this characteristic and that characteristic? Like such a person doesn't exist. But because you're so good at predicting the next token, you should still be able to guess what that person would do, this hypothetical, imaginary person, be a far greater mental ability than the rest of us. When we're doing reinforcement learning on these models, how long before most of the data for the reinforcement learning

is coming from AI's and not humans? I mean, already most of the data for reinforcement learning is coming from AI's. Yeah. Well, it's like the humans are being used to train the reward function. But then the reward function in its interaction with the model is automatic and all the data that's generated during the process of reinforcement learning is created by AI.

So like, if you look at the current, I would say technique paradigm, which is in getting some significant attention because of chat, GPT, reinforcement learning from human feedback. So there is human feedback. The human feedback is being used to train the reward function. And then the reward function is being used to create the data which trains them off. Got it. And is there any hope of just removing a human from the loop and have it improve itself and some sort of alpha go away?

Yeah, definitely. I mean, I feel like in some sense our hopes for like our plan, like, very much so. The thing you really want is for the human teachers that teach the AI for them to collaborate with an AI. You might want to think about it

and you might want to think of it as being in a world where the human teachers do 1% of the world and the work and the AI do 99% of the work. You don't want it to be 100% AI but you do want it to be a human machine collaboration which teaches the next machine. Currently, I mean, I have the chance to play around these models,

they seem bad at multi-step reasoning, and they have been getting better. But what does it take to really surpass that barrier? I mean, I think dedicated training will get us there, more improvements to the base models you'll get us there. But fundamentally, I also don't feel like they're that bad at multi-step reasoning. I actually think that they are bad at mental multi-step reasoning, but they're not allowed to think out loud.

But when they are allowed to think out loud, they're quite good. And I expect this to improve significantly. both with better models and with special training. Are you running out of reasoning tokens at the Internet? Are there enough of them? I mean, you know, it's okay. So for context on this question, like there are claims that indeed at some point

we'll run out of tokens in general to train those models. And yeah, I think this will happen one day and by the time that happens, we need to have other ways of training models, other ways of productively improving their capabilities and sharpening their behavior, making sure they're doing exactly, precisely what we want

without more data. You haven't run out of data yet? There's more... Yeah, I would say the data situation is still quite good. There is still lots to go. But at some point, yeah, at some point data will run it.

Okay. What is the most valuable source of data? Is it Reddit, Twitter, books? What would you trade many other tokens of other varieties for? Generally speaking, you'd like tokens which are speaking about smarter things,

tokens which are like more interesting. Yeah. So, I mean, all the sources which you mentioned, they are valuable. Okay. So maybe not Twitter. But do we need to go multimodal to get more tokens, or do we still have enough text tokens left? I mean, I think that you can still go very far in text only, but going multimodels seems like a very fruitful direction. If you're comfortable talking about this, like, where is the place where we haven't scraped the tokens yet?

Oh, I mean, yeah. Yeah, obviously. I mean, I can't answer that question for us, but I'm sure I'm sure that for everyone there's a different answer to that question. How many orders of magnitude improvement can we get just not from scale or not from data, but just from algorithmic improvements? Hard to answer, but I'm sure there is some. Is some a lot or is still a little? I mean, it's only one way to find out.

Okay. Let me get to your quick-fire opinions about these different research directions. Retrieval transformers. So just like somehow storing the data outside of the model itself and retrieving it somehow. Seems promising. Would you see that as a path forward? or I think it seems promising. Robotics.

Was it the right step for Open AI to leave that behind? Yeah, it was. Like back then, it really wasn't possible to continue working in robotics because there was so little data. Like, back then, if you wanted to work on robotics, you needed to become a robotics company. You needed to really have a giant group of people working on building robots and maintaining them and having and even then, like if you're only,

if you're going to have a hundred robots, it's a giant operation already, but you're not going to get that much data. So in a world where most of the progress comes from the combination of compute and data, right? That's where we've been, where it was the combination of compute and data

that drove the progress. There was no path to data from robotics. So back in the day, then you made a decision to stop working in robotics. There was no path forward. Is there one now? So I'd say that now it is possible to create a path forward, but one needs to really commit to the task of robotics.

You really need to say, I'm going to build like many thousands, tens of thousands, hundreds of thousands of robots and somehow collect data from them and find a gradual path where the robots are doing something slightly more useful and then the data that they get from these robots and then the data that is obtained and used to train the models and they do something that's slightly more useful.

So you could imagine this kind of gradual path of improvement where you build more robots, they do more things, you collect more data and so on. But you really need to be committed to this path. If you say, I want to make robotics happen, that's what you need to do. I believe that there are companies who are thinking about such doing exactly that,

but I think that you need to really love robots and need to be really willing to solve all the physical and logistical problems of dealing with them, it's not the same as software at all. So I think one could make progress in robotics today with enough motivation. What ideas are you excited to try, but you can't because they don't work well on current hardware. I don't think current hardware is a limitation. Okay. I think it's just not the case. Got it. So, but anything you want to try, you can just spin it up? I mean, of course. Like, the thing, you might say, well, I wish current hardware was cheaper.

Or maybe it had higher, like maybe it would be better if it was higher memory process for bandwidth, let's say. But by and large, hard way is just a limitation. Let's talk about alignment. Do you think we'll ever have a mathematical definition of alignment? Mathematical definition, I think, is unlikely. Uh-huh. Like, I do think that we will instead have multiple, like, rather than achieving one mathematical definition, I think we'll achieve multiple definitions

that look at alignment from different aspects and I think that this is how we will get the assurance that we want and by which I mean you can look at the behavior you can look at the behavior in various tests in various adversarial stress situations you can look at how the neural net operates from the inside I think you have to look at several of these factors at the same time And how short do you have to be before you release a model in the while?

Is it 100%, 95%? Well, it depends how capable the model is. The more capable the model is, the more, the more, the higher over the the, the more confident it to be. Okay. So just say it's something that's almost AGI. Where is AGI? Well, it depends what your AGI can do.

Keep in mind that AGI is an ambiguous term also. Yeah. Like your average college undergrad is an AGI right? It's a data, yeah. But you see what I mean. There is significant ambiguity in terms of what is meant. by AGI. So depending on where you put this mark,

you need to be more or less confident. You mentioned a few of the paths towards alignment earlier. What is the one you think is most promising at this point? Like, I think that it will be a combination. I really think that you will not want to have just one approach. I think we will want to have a combination of approaches where we, you spend a lot of compute adversarily probed to find any mismatch

between the behavior that you wanted to teach the behavior that it exhibits, we look inside into the neural net using another neural to understand how it operates on the inside. I think all of them will be necessary. Every approach like this reduces the probability of misalignment. And you also want to be in a world where your degree of alignment keeps of increasing faster than the capability of the models. I would say that right now

our understanding of our models is still quite rudimentary. We made some progress, but much more progress is possible. And so I would expect that ultimately the thing that will really succeed is when we will have a small neural net that is well understood that's given the task to study the behavior

of a large neural net that is not understood to verify it. By what point is most of the AI research is being done by AI? I mean, so today, when you use co-pilot, right? What fraction? How do you do the, how do you divide it up? So I expect at some point you ask your, you know,

the Senate of chat GPT, you say, hey, like I'm thinking about this and this. Can you suggest fruitful ideas? And you would actually get fruitful ideas. I don't think that will make it possible for you to solve problems you couldn't solve before. Got it. But it's somehow just telling the humans, skipping them ideas faster or something. It's not itself interacting with the one example. I mean, you could slice it in a variety of ways.

But I think the bottleneck there is, good ideas, good insights, and that's something which should be the neural nets could help with this. If you could design some, like a billion dollar prize for some sort of alignment research result or product, what is like the concrete criterion

and what's definitely that makes sense for that a billion dollar price? There's something that makes sense for such a price? It's funny that you asked this. I was actually thinking about this exact question. I haven't come up in the exact criteria yet. Maybe something that with the benefit, like maybe a prize where we could say that

two years later or three or five years later, we'll look back and say it like that was the main result. So rather than say that there is a price committee that decides right away, you wait for five years and then award it retroactively. But there's no concrete thing we can identify yet as you solve this particular problem

and you made a lot of progress. I think a lot of progress, yes. I wouldn't say that that would be the full thing. Do you think end-to-end training is the right architecture for bigger and bigger models or do we need better ways of just connecting things together?

I think Antoine is very promising. I think connecting together is very promising. Everything is promising. So Open AI is projecting revenues of a billion dollars in 2024. That might very well be correct, but I'm just curious when you're talking about a new general purpose technology,

how do you estimate how big a windfall it will be? Why that particular number? I mean, you look at the current, you look at the, you know, we've already had a product for quite a while now back from the GPT three days from two years ago through the API and we've seen how it grew. We've seen how the response to Dali has grown as well. And so you see how the response to chat GPTs.

And I think all of this gives us information that allows us to make a relatively sensible extrapolation for any four. Maybe that would be, that would be one answer. Like you need to have data. You can't come up with those things out of thin air because otherwise your error bars will be like off by your airbags are going to be like a hundred X in each direction. I mean, but most exponentials don't stay exponential, especially when they get into bigger under quantities, right? So how do you determine in this case that, I mean, like, would you bet

against the eye? Not after talking with you. Let's talk about what like a post-AGI future looks like. So are people like you, you know, I'm guessing you're working like 80-hour weeks towards this grand goal that's really assessed with. Are you going to be satisfied? in a world where you're basically living in an AI retirement home? Or like, what is your, what are you concretely doing after AGI comes? I think the question of what I'll be doing or what people will be doing after AGI comes, it's a very tricky question. You know, I think where will people find meaning?

But I think, I think that that's something that AI could help us with. Like, one thing I imagine is that we'll all be able to become more enlightened because we'd interact with an AGI that will help us see the world more correctly but become better on the inside as a result of interacting. Like imagine talking to the best meditation teacher in history. I think that will be a helpful thing. But I also think that because the world will change a lot, it will be very hard for people to understand what is happening precisely

and how to really contribute. One thing that I think some people will choose to do is to become part of it. in order to really expand their minds and understanding and to really be able to solve the hardest problems that society will face then. Are you going to become part of AI? Very tempting. It is tempting. Do you think they'll be physically embodied humans and 3,000?

3,000. Oh, how do I know what's going to have any 3,000? What does it look like? Are there still, like, humans walking around on Earth? Have you guys thought concretely about what you actually want this world to look like? 3,000. Well, I mean, the thing is, here's the thing. Let me describe to you what I think is not quite right about the question.

Like it implies like, oh, like we get to decide how we want the world to look like. I don't think that picture is correct. I think change is the only constant. And so, of course, even after AI is built, it doesn't mean that the world will be static. The world will continue to change. The world will continue to evolve. And it will go through all kinds of transformations. And I really have no, I don't think anyone has any idea of how the world will look like

in 3,000. But I do hope that there will be a lot of descendants of human beings who will live happy, fulfilled lives, where they are free to do as their wish, as they see fit, where they are the ones who are solving their own problems. Like one of the things which I would not want, one world which I would find very unexciting is one where, you know, we feel this powerful tool,

and then the government said, okay, so the AGI said that society shall be run in such a way, and now we shall run society in such a way. I'd much rather have a world where people are still free to make their own mistakes and suffer their consequences and gradually evolve morally and progress forward on their own strength to their own strength. See what I mean. Be the AGI providing more like a base safety net. How much time do you spend thinking about these kinds of things versus just doing their research that? I do think about those things a fair bit.

Yeah, things are very interesting questions. So in what ways have the capabilities we have today? in what ways have they surpassed where you expected them to be in 2015 and in what ways are they still not where you would expect them to be by this point? I mean, in fairness, they did surpass what I expect them to be in 2015. In 2015, my thinking was a lot more. I just don't want to bet against deep learning.

I want to make the biggest possible bet on deep learning. Don't know how, but it will figure it out. But is there any specific way in which it's been more than you expected or less than you expected? Like some concrete prediction you had in 2015, and that's been frowned? You know, unfortunately, I don't remember concrete predictions I made in 2015.

But I definitely, but I definitely think that overall, in 2015, I just wanted to move, to make the biggest bet possible on deep learning. But I didn't know exactly. I didn't have a specific idea of how far things will go in seven years.

Well, I mean, 2015, I did have all these bets with people in 2016, maybe 2017, that things will go really far. but specifics. So it's like it's both the case that it surprised me and I was making these aggressive predictions, but I think maybe I believe them only 50% on the inside.

Well, what do you believe now that even most people at OpenAI would find Farfetched? I mean, I think that at this, because we communicated a lot at Open AI, people have a pretty good sense of what I think. And so, yeah, we reached a point at Open AI I think we see eye to eye and all these questions. So Google has, you know,

it's custom TPU hardware, it has all this data from all its users, you know, Gmail, and so on. Does it give it an advantage in terms of training bigger models and better models than you? So I think like when the first at first when the TPU came out, I was really impressed and I thought, wow, this is amazing. But that's because I didn't quite understand hardware back then. What really turned out to be the case is that TPUs and GPUs are almost the same thing. They are very, very similar.

It's like, I think a GPU chip is a little bit bigger. I think a TPU chip is a little bit smaller. It may be a little bit cheaper, but then they make more GPUs than TPUs, so I think the GPUs might be cheaper after all. But fundamentally, you have a big processor, and you have a lot of memory, and there is a bottleneck between those two.

And the problem that both the TPU and the GPU are trying to solve is that by the amount of time it takes you to move one floating point from the memory to the processor, you can do several hundred floating point operations on the processor, which means that you have to do some kind of batch processing. And in this sense, both of these architectures are the same. So I really feel like hardware, like in some sense, the only thing that matters about hardware is cost, cost per flop.

Overall systems cost. Okay, and there isn't that much different. Well, I actually don't know. I mean, I don't know how much, what the TPU costs are, but I would suspect that probably not, if anything, probably views are more expensive because there is less of them. When you're doing your work, how much of the time has spent, you know, configuring the right initializations, making sure the training run goes well and getting the

right hyper parameters, and how much is it just coming up with whole new ideas? I would say it's a combination, but I think that coming up with, it's a combination, but coming up with whole new ideas is actually not. It's like a modest part of the work. Certainly coming up in new ideas is important. But I think even more important is to understand the result. to understand the existing ideas, to understand what's going on. Because normally you have this, you know, neural net is a very complicated system, right?

And you ran it and you get some behavior, which is hard to understand, what's going. Understanding the results, figuring out what next experiment to run. A lot of the time is spent on that. Understanding what could be wrong, what could have caused the neural net produce a result, which was not expected. I'd say a lot of time we spend as well. Of course, coming up with new ideas, but not new ideas. I think like, I don't like this framing as much.

It's not that it's false, but I think the main activity is actually understanding. What do you see is the difference between the two? So, at least in my mind, when you say come up with new ideas, I'm like, oh, like, what happened if it did such and such? Whereas understanding, it's more like, like, what is this whole thing? Like, what are the real underlying phenomena that are going on? What are the underlying effects? like why

why are we doing things this way and not another way? And of course, this is very adjacent to what can be described as coming up with ideas. But I think the understanding part is where the real action takes place.

Does that describe your entire career? Like if you think back on like ImageNet or something, was that more a new idea or was that more understanding? Oh, I was definitely understanding. Definitely understand. It was a new understanding of very old things. What is the experience of training on Azure been like using Azure?

Fantastic. I mean, yeah, I mean, Microsoft has been a very, very good partner for us, and they've really helped take Azure and make it, bring it to a point where it's really good for ML and you're super happy with it. How vulnerable is the whole AI ecosystem do something that might happen in Taiwan? So let's say there's like a tsunami in Taiwan or something. What happens to AI in general? Like, it's definitely going to be a significant setback.

it's not going to like it might be something equivalent to like no one will be able to get more more computers for a few years but I expect computers will spring up like for example I believe that Intel has fabs just of the previous

of like a few generations ago that means that if Intel wanted to they could produce something GPU from like four years ago so yeah it's not the best let's say I'm actually not sure about if my statement about Intel is correct but I do know that

there are fabs outside of Taiwan that is not as good. But you can still use them and still go very far with them. It's just, it's just a setback. Will inference get cost prohibitive as these models get bigger and bigger? So I have a different way of looking at this question. It's not that inference will become cost prohibitive.

Inference of better models will indeed become more expensive. But is it prohibitive? Well, it depends on how useful is it? Like, if it is more useful than it is expensive, then it is not prohibited. Like to give you an analogy, like suppose you want to talk to a lawyer,

you have some case or need some advice or something, you are perfectly happy to spend $500 an hour. Right? So if your neural net could give you like really reliable legal advice, you'd say, I'm happy to spend $400 for that advice and suddenly inference becomes very much non-prohibitive. The question is, can neural net produce an answer good enough at this cost?

cost. Yes. And you will just have like price discrimination, different models of different. I mean, it's already the case today. So on our product, the API,

we serve multiple neural nets of different sizes. And different customers use different neural nets of different sizes depending on their use case. Like if someone can take a small model and fine tune it and get something that's satisfactory for them, they'll use that. Yeah. But if someone wants to do something more complicated and more interesting,

is the biggest model. How do you prevent these models from just becoming commodities where these different companies would just, they just spit each other's prices down until it's basically the cost of the GPU run? Yeah, I think there is, without question, a force that's trying to create that. And the answer is you got to keep on making progress. You got to keep improving the models. You got to keep on coming up with new ideas and making our models better and more reliable,

more trustworthy, so you can trust their answers, all those things. Yeah, but let's say it's like 2025 and the model from 20, or somebody just offering it at cost. And it's still pretty good. Why would people use a new one from 2025 if the one from just a year older is even better? So there are several lenses there. For some use cases, that may be true. There will be a new model from 2025, which will be driving the more interesting use cases.

There's also going to be a question of inference cost. Like you can do research to serve the same model at less cost. So there will be different, the same model will be served, will cost different amounts to serve for different companies. I can also imagine some degree of specialization too where some companies may try to specialize in some area and be stronger in a narrower area compared to other companies. And I think that too may,

that may be a response to commoditization to some degree. As over time, do these different companies, do their research directions converge or they diverge? Are they doing similar and similar things over time? Or are they doing, are they going up, branching off in different areas? So I'd say in the near term, it looks at least convergence.

in the, like, I expect it's going to be a convergence, a divergence convergence behavior where there is a lot of convergence on the near-term work. There's going to be some divergence on the longer-term work. But then once the longer-term work starts to yield through, I think there will be conversions again. Got it. When one of them finds the most promising area, they everybody just...

That's right. Now, there is obviously less publishing now, so it will take longer before this promising direction gets rediscovered. But that's how I'd imagine it. I think it's going to be convergence, divergence, convergence. Yeah, we talked about this a little bit at the beginning, but as foreign governments learn about how capable of these models are, are you worried about spies or some sort of attack to get your weights

or, you know, somehow abuse these models and learn about them? Yeah, it's definitely something that you absolutely can't discount that. Yeah. And, yeah, something that we try to guard. against the best of our ability, but it's going to be a problem for everyone who's building. How do you prevent your weights from leaking? I mean, you have, like, really good security people. And, like, how many people have the, if they wanted to just, like,

a stage into the weights machine, how many people could do that? I mean, like, what I can say is that the security people that we have, they've done a really good job so that I'm really not worried about the weights being leaked. Okay, got it. What kinds of emergent properties are you expecting from these models? at this scale? Is there something that just comes about de novo? I'm sure things will come. I'm sure really new surprising properties will come up.

I would not be surprised. The thing which I'm really excited about or the thing which I'd like to see is reliability and controllability. I think that this will be a very, very important class of emerging properties. If you have reliability and controllability, I think that helps you solve a lot of problems.

Reliability means you can trust the models out. Controllability means you can control it. And we'll see. but it will be very cool if those emergent properties did exist. Is there somewhere you can predict it at advance? Like what will happen in this parameter account? What will happen that parameter? I think it's possible to make some predictions

about specific capabilities, though it's definitely not simple and you can't do it in a super fine-grained way, at least today. But I think getting better at that is really important than anyone who is interested in who has research ideas on how to do that, I think that can be a valuable contribution. How seriously do you take these scaling laws if there's a paper that says like, oh, you just increase,

you need this many orders of magnitude more to get all the reasoning out. Do you take that seriously or do you think it breaks down at some point? Well, the thing is that the scaling law tells you what happens as you, what happens to your next word prediction accuracy, right? There is a whole separate challenge of linking next word prediction accuracy to reasoning capability. I do believe that indeed there is a link, but this link is complicated.

And we may find that there are other things that can give us more reasoning per unit effort. Like, for example, some special, like, you know, you mentioned reasoning tokens, and I think they can be helpful. There can be probably some things that can do. Is this something you're considering just hiring humans to generate tokens for you, or is it all going to come from that that already exists out there?

I mean, I think, that relying on people to teach our models to do things, especially to make sure that they are well-behaved and they don't produce false things. I think it's an extremely sensible thing to do. Isn't it odd that we have the data we need at exactly the same time as we have the transformer at the exact same time that we have these GPUs?

Like, is it odd to you that all these things happen at the same time or do you not see that way? I mean, it is definitely an interesting, it is an interesting situation that is the case. I will say that it is odd and it is less odd on some level. Here is why it's less odd. So what is the driving force behind the fact that the data exists, that the GPUs exist, that the transformer exists.

So the data exists because computers became better and cheaper. We've got smaller and smaller transistors. And suddenly at some point it became economical for every person to have a personal computer. Once everyone has a personal computer, you really want to connect them with the network. You get the internet. Once you have the internet, you have suddenly data appearing in great quantities. The GPUs were improving concurrently because you have small and smaller transistors and you're looking for things to do with them.

Gaming turned out to be a thing that you could do. And then at some point, the gaming GPU, Nvidia said, wait a second, Brian, may it turn it into a general purpose GPU computer, maybe someone will find it useful. Turns out it's good for neural nets. So it could have been the case. that maybe the GPU would have arrived five years later or ten years later. If, let's suppose, gaming wasn't a thing.

It's kind of hard to imagine. What does it mean if gaming isn't a thing? But it could, maybe there was a counterfactual world where GPUs arrived five years after the data or five years before the data. In which case, maybe things would move a little bit more. Things would have been as ready to go as there now. But that's the picture, which I imagine. All this progress in all these dimensions,

is very intertwined. It's not a coincidence that, like, you don't get to pick and choose in which dimensions things improve, if you see what I mean. How inevitable is this kind of progress? So if, like, let's say you and Jeffrey Henton and a few other pioneers,

if they were never born, does the deep learning revolution happen around the same time? How much does it delay? I think maybe there would have been some delay, maybe like your delay. It's really hard to tell. It's really hard to tell.

I mean, I hesitate to give a lot, a longer answer because, okay, well then you'd have GPUs would keep on improving, right? Then at some point, I cannot see how someone would not have discovered it. Because here's the other thing. Okay, so let's suppose no one has done it. Computers keep getting faster and better. It becomes easy and easy to train these neural nets because you have bigger GPUs.

So it takes less engineering effort, train one. You don't need to optimize your code as much. You know, when the ImageNet dataset came out, it was huge and it was very, very difficult to use. now imagine you wait for a few years and it becomes very easy to download and people can just thinker. So I would imagine that like a modest number of years, maximum, this would be my guess. I hesitate. I hesitate to give a longer answer, though, you know, you can't run.

You can't rerun the world. You don't know what of them. Let's go back to alignment for a second. As somebody who deeply understands these models, what is your intuition of how hard alignment will be like i think with the so here's what i would say i think with the current level of capabilities i think we have a pretty good set of ideas of how to align them but i would not underestimate the difficulty of alignment of models that are actually smarter than us of models that are capable of misrepresenting their intentions like i think i think it's something to to think about a lot and to research i think

this is one area also by the way you know like oftentimes academic researchers asked me ask me where what's the best place where they can contribute? And I think alignment research is one place where I think academic researchers can make very meaningful contributions. Other than that, do you think academia will come up with more insights about actual capabilities or is that going to be just the companies at this point?

So the companies will realize the capabilities. I think it's very possible for academic research to come up with those insights. I think it's just, it doesn't seem to happen that much for some reason, but I don't think there's anything fundamental about academia. It's not like academia can't. I think maybe they're just not thinking about

the right problems or something because maybe it's just easier to see what needs to be done inside these companies. I see. But there's a possibility that somebody could just realize. Yeah, I totally think. Like, why would I possibly rule this out? You see what I mean?

What are the concrete steps by which these language models start actually impacting the world of Adams and not just the world of bits. Well, you see, I don't think that there is a distinction, a clean distinction between the world of bits and the world of atoms. Suppose the neural net tells you that, hey, like, here is, like, something that you should do and it's going to improve your life,

but you need to, like, rearrange your apartment in a certain way, and you go and you rearrange your apartment as a result. The neural net impact the world of atoms just. Yeah, fair enough, fair enough. Do you think it'll take a couple of additional breakthroughs as important as a transformer they get to superhuman AI or do you think we basically got the insights in the books somewhere

and we just need to implement them and connect them? So I don't really see such a big distinction between those two cases and let me explain why. Like I think what's, one of the ways in which progress has taken place in the past is that we've understood that something had a property, a desirable property all along,

which you didn't realize. So is that a breakthrough? You can say, yes, it is. Is it an implementation of something on the books? Also, yes. So my feeling is that a few of those are quite likely to happen, but that in hindsight, it will not feel like a breakthrough. Everybody is going to say, oh, well, of course, like,

it's totally obvious that such and such thing can work. You see, with the transformer, the reason it's being brought up as a specific advance is because it's the kind of thing that was not obvious for almost anyone. So people can say, yeah, like, it's not something which they knew about. But if an advance comes from something, like let's consider that the most fundamental advance of deep learning, that the big neural network trained with back propagation can do a lot of things.

Like, where's the novelty? It's not in the neural network. It's not in the back propagation. But then somehow it's the kind of, but it is most definitely a giant conceptual breakthrough because for the longest time, people just didn't see that. But then now that everyone sees that, everyone's going to say,

well, of course, like it's totally obvious, big neural network. everyone knows that they can do it. What is your opinion of your former advisor's new forward forward algorithm? I think that it's an attempt to brain a neural network without back propagation. And I think that this is especially interesting if you are motivated to try to understand how the brain might be learning its connections. The reason for that is that as far as I know,

neuroscientists are really convinced that the brain can, cannot implement back propagation because the signals in the synops is only moving one direction. And so if you have a neuroscience motivation and you want to say, okay, how can I come up with something that tries to approximate the good properties of back propagation without doing back propagation? That's what the forward forward algorithm is trying to do. But if you are trying to just engineer a good system, there is no reason to not use back propagation. Like, it's the only algorithm. I guess I've heard you in different contexts talk about the need,

like using humans as the, you know, the existing example case that, you know, AGI exists, right? So at what point do you take the metaphor less seriously and don't feel the need to pursue it in terms of research? Because it is important to you as a sort of existence case. Like at what point does stop caring about humans as an existence case of intelligence? Or as the example of the model you want to follow in terms of pursuing intelligence in models. I see.

I mean, like you got a, I think it's good to be inspired by humans. I think it's good to be inspired by the brain. I think there is an art into being inspired by humans and the brain correctly. Because it's very easy to latch on to an non-essential quality of humans or of the brain. And I think many people who, many people whose research is trying to be inspired by humans and by the brain often gets a little bit specific. People get a little bit too, okay, so like what cognitive science model should you follow? At the same time, consider the idea of the neural network itself, the idea of the artificial

neuron. This too is inspired by the brain, but it turned out to be extremely fruitful. So how do you do this? What behaviors of human beings are essential that you say, like this is something that proves to us that it's possible? What is in essential? No, actually, this is like some emergent phenomenon of something more basic. and we just need to focus on getting our own basics right.

I would say that it's like I think one should, one can and should be inspired by human intelligence with care. Final question. Why is there, in your case, such a strong correlation between being first to the deep learning revolution and still being one of the top researchers? You would think that these two things wouldn't be that correlated, but why is that their correlation? I don't think those things are super correlated indeed.

I feel like in my case, I mean, honestly, it's hard to answer the question. You know, I just kept on, I kept trying really hard and it turned out to have sufficed thus far. So it's a perseverance. I think it's a necessary but not a sufficient condition. Like, you know, many things need to come together in order to really figure something out. Like you need to really go for it and also need to have the right. way of looking at things. And so it's hard, it's hard to give them like a really meaningful answer to this question.

All right. Um, Ilya, it has been a true pleasure. Thank you so much for coming out of the Lunar Society. I appreciate you bring us to the offices. So thank you. Yeah, I really enjoyed it. Thank you very much. Hey, everybody.

I hope you enjoyed that episode. Just wanted to let you know that in order to help pay for the bills associated with this podcast, I'm turning on paid subscriptions on my substack. at Warkeshpatel.com. No important content on this podcast will ever be paywalled, so please don't donate if you have to think twice before buying a cup of coffee. But if you have the means and you've enjoyed this podcast or gotten some kind of value out of it,

I would really appreciate your support. As always, the most helpful thing you can do is to share the podcast. Send it to people you think might enjoy it, put it in Twitter, your group chats, etc. Just splits the world. Appreciate you listening. I'll see you next time. Cheers.