Dwarkesh Podcast - AI researchers debate how close we are to recursive self-improvement

0:00Today, I'm chatting with three of my AI researcher friends from whom I learn a lot every time we talk, and who also happen to be at somewhat openish labs and companies, so you guys can actually say things on the record. I'm joined by Baron Millich, who is the CTO of Zyphera, which is developing open-source models. John Shulman, who is the chief scientist at Thinking Machines. Previously, the co-founder of OpenAI led the RLHF work that led to Chechi PT. And Charlie O'Neill, who is head of model training at Base10.

0:29The first question I have, if we're in 2036, it's been 10 years, we don't have like billions of crazy superintelligence that are running around that have like radically transformed the world. What is the most likely reason that doesn't end up being the case? Other than sort of exogenous political shocks or like if there's a war or they banned AI or something, but what is the most likely technical reason that 2036 isn't like a crazy alien superintelligence world?

0:56I mean like my reason would just be like it's got to be the sort of like there's been a classic thing almost like Morvex paradox right where like we see like you know we think of the AI be like if it can do this is going to be amazing right like if it can solve these hot math problems if it can win a chess blah blah and then it solves these things and then it's like not that impactful. Obviously it's not impactful but like not everything. It's like if somehow that continues and like there's never like the true like spark of generalization that occurs. I think that could lead to like the AI is just being like extremely good at kind of everything that people like put into a benchmark, put into an environment but like there's still some persistent like Syntorilla which is somehow blocking everything. I think this is kind of unlikely. I think we do actually see this kind of generalization even from our island practice already.

1:37But if it is just ridiculously hard to generalize meta-learning, plus we don't solve container learning, it's just super hard and impossible. This would be my default scenario in that case. Yeah, I agree with that. Humans have a lot of advantages over models now. And each time a new model comes out, it'll catch up in some of these areas.

2:02like you end up getting bottlenecked by the places where the model is weaker and where it has worse judgment or the models can't check themselves well enough. Yeah, so there's this cycle that keeps repeating where people think where a new model comes out and people are blown away and they're like, this is it, this is the, this is AGI, but then they use it a bit and then it starts to feel dumb after a month or so. So that cycle just might keep going and it's hard to predict how many times it's going to repeat.

2:31Like right now, you don't get explosive growth in capabilities because you still get bottlenecked enough when you're trying to do research in engineering that even if the model can write way more code than a person, it doesn't make you like a hundred times more productive. So maybe they're just more of these cycles than we would expect. For me, it's like a question of how far off like this global optimum of a learner you could have on a chip.

2:59is like the transformer plus like RL basically like the current recipe. So like I think people imagine that even once like once you have a an agent which is better than all humans at AI research even if it's like 0.1% better than all humans then the fact that you can run like you know hundreds of thousands if not millions of these in parallel you can run them much faster like chips kind of speed up that's going to outweigh every other like bottleneck and like you're eventually just going to like hit this like very fast takeoff with recursive self-improvement.

3:28I could imagine that if we continue along the trajectory that we're currently on with that paradigm where it's basically just like self-attention, RL, scaling up RL environments. I guess if you think about what happened with Moore's Law, we had this very nice straight line and that held for a really, really long time. But there were so many discrete discontinuities and innovations that had to happen to keep that scaling law going. And the same thing has kind of happened with LLMs.

3:55had this pre-training scaling law, and then that was kind of hitting the diminishing returns. And then we came up with RL and solve that. And then we got this new diminishing returns curve to hit that made it keep looking like a straight line going up. And so if it requires another one of those discontinuities to solve, I'm not sure that the current method of training LLs with these RL environments, even RSI targeted RL environments, would be able to discover that discontinuity. And if not, we're probably going to hit this like asymptotic like curve where like. So do you think the discontinuity will be harder than anything that's come since 2012? If we had the answer that we kind of have the ability to implement it but like maybe there's the distinguish we should distinguish between a discontinuity which adds to the current paradigm again it's like cumulative like there's some thing beyond the RL that we have to discover and maybe they're capable of like you know connecting the dots in that straight line or like.

4:49But again, how far off the global optimal are we? Do we have to go back and throw out gradient descent and neural nets in general? And I don't think if you continue to scale up the current paradigm and LLM, no matter how many LLMs you're running are capable of necessarily discovering that if it's too far away. Yeah, the only hope really is if deep learning just can't get us to an AI which is at least.

5:14can dominate human research and human development, including the human ability to come up with new paradigms and so forth. Or like, I don't know, maybe humans would also never have discovered the next learning architecture, but to the extent humans could have discovered it eventually. But it just seems like, I don't know, if you just look at the progress that's happened since 2012 till now, and you just continue that on. I mean, I know it's been powered by a huge amount of compute scaling and so forth, but it would be weird if like, It just didn't get to the point where it could like dominate humans, at least in R&D. Especially over the next few years, there's going to be Ryan Greenblatt was on the podcast recently. And he made this point that you could imagine as AI's get more and more capable and are capable of making progress on simulations which incentivize getting better at not only AI R&D but generally at science. So this is a thing that all the labs targeting many startups are targeting. Or another intuition pump is if you look at the ELO score.

6:12of chess bots since the 80s. They're just like a very linear increase in ELO over time. But there's this huge discontinuity as they cross the human range of human experts always win against AIs to like human experts never win against AIs as this linear increase in ELO happened. And you could think, I agree with your point that.

6:30So far, AI capabilities have not been that big of a deal in terms of their end economic impact in the world, but that's just because they're slowly rising in ELO relative to humans. Yeah, I agree. The only way for this to not happen is if, as you said, somehow asymptotes just before basically, because we're already pretty close, and I've been into where we'll start crossing the human ELO score. And so we'll need to asymptote before that. And that's the only way. And this scenario you posed where somehow we're sitting here in 2035, and everything is normal for this to happen, I think.

7:00there's some dramatic regulation on AI. This is kind of what I see as the most likely way for this scenario to happen actually, rather than the technical thing. Yeah, I think there's different kinds of research. There's research where it's like the auto research style where the objective is already specified very cleanly and you're optimizing that objective. And I think everyone is picturing like.

7:19if we continue along this path of like, you know, making pre-training loss go down, making our own environments back to the bottomless go up, that's going to lead to like improvement. But like, you know, maybe what Ryan is talking about is like this much more open-ended type of science, which is required for like paradigm shifts where we can't specify the objective and the AI's are definitely not able to specify that objective either. Like we have to be really, really careful about how we specify objectives for any of these things. And maybe your point is that like the nature of the breakthroughs that have happened since 2012 is that we have found like in 2012 people weren't saying

7:47I'm assuming, I don't know, you guys were there. Or at least John, you were there. I was in promiscule. I actually tried, John, I'm curious where your wisdom of the ages of, or wisdom of being in the trenches way back when. But presumably a big breakthrough was realizing that next token prediction is the, you wouldn't have thought that nano-GPT speedrun is the thing to be optimizing for in 2014.

8:13But now that we have come to this new paradigm, you wouldn't think to do a speed run on that and have AI's get really good at that. But maybe there's like a next inner loop to optimize that the AI's wouldn't anticipate. And there's an outer loop of like revenue or something that eventually should be strong, but it's a very slow outer loop. Yeah. In fact, I remember in the early open AI days having the intuition that actually just do like.

8:39Minimizing log loss wasn't going to get you to intelligence, because the important bits are accounting for such a small fraction of the loss that it was going to be overwhelmed by noise. So just training a language model on next token prediction just wasn't going to learn the interesting things you wanted to learn. And we needed to craft better objectives that would put more emphasis on the important things.

9:07You can make all sorts of arguments for this and you could say, humans probably don't learn how to model everything in our environment. Most people can't create a photorealistic reproduction of some kind of scene they've looked at. We must need a better objective, but then it turned out that it just worked anyway. As you were pointing out, the inner loop, even in current AI research of post-training benchmarks or whatever. It doesn't necessarily translate into what users like. Oh yeah, I mean, the whole field relies a lot on generalization and it's very hard to predict when you're going to get generalization or when you're going to get some kind of out of distribution generalization. So we know that if you train on the task you care about, you're going to do better. But like the most important advances are often

10:03the types of generalization that we have no right to expect. So for example, from just pre-training on this very naive next token prediction objective to various tasks of interest where some very, that require understanding of the input in some deep way or learning some skill from pre-training that's very rare and not very heavily represented.

10:32And then also generalization from these verifiable tasks to less verifiable ones. This is also a type of generalization that there's no reason priori to expect it. So this is an interesting question because one intuition pump that you could have for why you would see some sort of singularity very rapidly without even scaling up the inputs to a progress that are not just EA labor is that before every single experiment you run That's like a seven figure experiment. You spend an equivalent amount of compute on AI labor. And so you just have automated versions of you guys spending a century thinking about what is the optimal experiment to run.

11:13Do you like small scale ablations? Developing literally like a century's worth of theory. So going back even before like deep learning, before you decide what experiment to run, doing extremely optimal like setting up of the experiment, then you do a century of thinking after the experiment is over, where you're like analyzing what happened and what the next experiment to run is. Yeah. Well, I think if you think hard enough, you probably could have expected some of these things beforehand.

11:40There is probably some very clever way to do a small-scale experiment that'll let you build the theory that then will generalize to the large-scale experiment. So I would expect that we're nowhere near the ceiling of how well you can do research. I would imagine a future where AI is doing a lot of analysis and theory building.

12:05spending a comparable amount of compute to the amount that you're spending on the experiments themselves doing various kinds of analysis and building a theory around what we've seen so far. I think there's really concrete examples of this when the objective is well specified. So again, all thinking can do is update your posterior based on the bits that you've gotten since you formed your prior. You can't gain any new bits from just thinking.

12:29But when the objective is well specified and there is this data sitting around, I imagine there will be this big speed up in the current paradigm we're in. And a good example of this is if you've got an AI to think about the Kaplan scaling laws, an AI at this point would have noticed that they've just taken these intermediate checkpoints and didn't account for the annealing. And so this is wrong. That would have caught that years earlier, we would have made progress, would have cut off a year or two of progress just from that observation from an AI.

12:57And like again, once the objective is well specified, which is like lower pre-training loss or whatever. Like there's many, many good examples where if you just thought about it a bit more, you would have been able to like cut down a significant on things that you've done. So like mu P and like how learning rate scales with like model size and like realizing the model width is important in that as well. Like I feel like you can really back out a lot of these things and cut off like a lot of like hanging fruits. I would imagine like a 10 times speed up if our thing is just like maximize the objective we're currently on.

13:26But I don't see how that generalizes at all to come up with the right objective in the first place. Just thinking doesn't necessarily buy you the right objective in the first place. I mean, yeah, I think this is really the key question to any kind of very rapid RSI from current AIs. It's like, how well can AIs generalize to learning their own objectives? Because to have any kind of self-propelling automated loop, you need the AI to propose objectives, optimize them, figure out how to propose a new objective, and have this not go off the rails at any point for a long, long time.

13:53coming back to Morvex products there might be like a case of Morvex products where like we think there's kind of like autonomy and sort of like being like self encapsulated so we can you know think of what we should do ourselves and then go do it and like have this loop is like super easy because we always do this and like obviously evolution needs to create creatures that can like survive on by themselves like long periods of time and like this just might be something that for some reason is like really hard for the AI in the same way that like locomotion stuff is really hard, but it's like math is super easy despite being super hard for us. I don't know. Doesn't the time horizon increasing suggest that that's... Yeah, exactly. I mean, this is another possibility, which like, but I agree, like, there's no obvious evidence for this. Like, in fact, the fact that, you know, our agent is now like super persistent and it's quite easy to do this. It's kind of evidence against this. But like, this would be, you know, potentially like one of the reasons why like we just don't get this like immediate takeoff is like if this is hard. If you look back from 2012 till now, or maybe from when you...

14:44started doing your research till now, what part of all the innovations that have happened since that time, including purely engineering ones, including purely conceptual ones, what seems like the thing that is the thing that would be the last things humans would have to do before AI's totally automate AI R&D?

15:05Probably just like iteratively asking the right questions. Like if you can get the AI to like do any experiment but like you need to decide what experiments to do and like right now I think AI's are not very good at this compared to coding the experiments at all. Like whenever we talk about research they propose like a bunch of like miscellaneous things which are like very very tiny steps. Or even going from like you know DeepMind's approach like we're going to solve intelligence by learning to play games to the superhuman level. That's going to be the approach to like one random researcher like Radford being like I'm going to try and just predict the next token.

15:35of a very wide swath of data. And then even once Radford had discovered that, it took a while before people decided to scale up because we had to come with the idea of scaling walls and the fact that you could very reliably predict these things. I would say that the last job for humans, or the role for humans at the last, the longest is defining the objective and deciding what we actually want. So in that vein, something like deciding what the assistance should behave or what it means to be helpful or what's the objective when we're doing our all-from-human feedback is one such thing. Then later, defining constitutions and model specs is another one.

16:26even if the AIs can do all the technical work, we'll have to still do a lot of that and decide what we actually want. Yeah, alignment is the final job. Yeah, alignment is sort of the answer, but it's also, alignment itself can be kind of decomposed into specification of the objective or figuring out what the right objective should be, and then actually achieving or optimizing the objective you've defined.

16:51and I think the first one is not going to go away anytime soon. If I think about a post-training team and why you need a lot of people to be on the team, it's just because there are a lot of different areas where you have to figure out how the model should behave.

17:15It would be very hard to automate the whole thing just because someone has to think about how should the model behave in this area. Jane Street started using antithesis to test their software in early 2025 and they were so impressed by the product that they decided to invest in the company. I recently caught up with Ron Manskey, who co-leads Jane Street's tech group to ask about how antithesis actually plugs in.

17:39The thing that I think is most impressive about Antenesis is we started using it in a team that was building high assurance software and being really careful and nonetheless it was able to shake out bugs that were otherwise going to be really hard to find and that's important both because it helps make those systems more reliable but also because it helps the teams that build it to just move faster.

18:00This matters more and more as code production is increasingly automated. I think in general, as we've been using agents more and more, the key problem that you run into is the verification bottleneck. Just the time it takes from people to look at code and figure out is that actually something you want to accept in your production software.

18:17Tools that make testing better are just incredibly helpful there. They just ease the verification bottleneck and make it possible for you to get more stuff done and move faster because you can have more confidence that the code generated by the agent is actually not introducing new problems. To see how Antithesis fits into your development process, go to antithesis.com.

18:42There isn't huge consolidation in model providers. There's just so many things that point to centralization here. Yeah, if we step back over the course of years, is there something that is going to prevent that? Yeah, I think distillation is the main thing that fights against the centralizing force. Because basically anything that can be learned through RL can be distilled very easily.

19:07It's a small number of bits. It's something that you can learn from a small amount of data. So if you can get trajectories from the model that show a behavior, you can easily distill it. So I think distillation is one of the things that fights centralization. There's also- I mean, there is a possibility that there'll be company specific models that it'll be possible to learn from deployment and have a company continually improving its own model. And such a system could be provided by the current oligopoly of model providers or some other currently smaller company. But I think that'll change the game a bit.

19:54Yeah, and I also want to point out that, like, continual learning doesn't stop distillation, right? Like, even if your model's improving every day, like, people could be distilling it every day. So it's like, the loops could just operate at the same pace. Right, that makes sense. Okay, so copying model behavior. I guess you need to know yourself what the right distribution to prompt is in order to get, like, the relevant model behavior.

20:16Yeah, for just distilling with supervised learning, the prompt distribution is extremely important. So it's very non-trivial to distill a model even if you have full access to it and have the chain of thought and everything. Yeah, it's non-trivial to distill all of the useful capabilities from it because you need to prompt the model with something and you need to prompt it with like realistic prompts.

20:45you need to have a really wide distribution of realistic prompts. So yeah, one thing that's been coming out recently is some of the Chinese companies are probably using these router services, which are designed to allow people in China to use the US frontier models, which would otherwise be blocked in China. But there are all these router or proxy services that allow people in China to use these models mostly for coding.

21:13and these router services are collecting and selling some of the data. So I think this is like a very useful data set for distillation because it gives you the perfect prompt distribution.

21:26I think this is one of those things where AIs help a lot here. If you actually look at the front-tier pipelines of, say, the Chinese models that they actually put in their papers, it's a lot of humans or they get seed prompts from somewhere, which is in combination with humans, this kind of data. And then they synthesize a vast coverage from their seed prompts using their existing models or the other front-tier models. And so it's like you can automate an awful lot of this prompt distribution gathering and environment creation. It's just like humans need to provide increasingly fewer amounts of bits. It's like the models get better.

21:55like by having a service which has users or users are going through. So like. Not necessarily. I mean, yeah, that's obviously very helpful. But theoretically, you can just think about what users want or like. The whole point is that the user says, make me an application like this. Oh, that didn't work. I actually want you to make this new feature. But actually, let's step back and do this other thing. And capturing that whole trace is the, or to the extent you could have done that anyways.

22:21then you just have RSI. Ultimately, if you have this fully automated loop, that is basically RSI. The AI is deciding the data, it's deciding the training, that is the loop. But yeah, it depends how much human information you need. At some point, if you're just like, I want traces that look like this, you prompt that to the model, the model will be able to come up with a pretty good approximation. But what if you want to do, make me a really good politician, and they just anticipate de novo.

22:45How would a discussion in the Senate halls go or something? I just feel like there's going to be a lot of things. Ironically, this is actually, I think, easier for the distillers than the Frontier Labs, right? Because the distillers just, like, I want a good politician. They go to the Frontier model. The Frontier model already knows how to be a good politician, so they just, like, generate those traces. Whereas, like, if you actually want to build the first model that does this, you have to, like, actually somehow get data on what politicians do every day and, like, build that.

23:08So it's actually much easier to like say like I want something like this and then like get like the idea to produce like a billion variations than to like actually create the thing like this to begin with. I think you can actually make a really concrete prediction based off like this observation that the Chinese types have this router data. So like I think the thing that that just did this originally was I was saying isn't it weird how Sonnet 5 and Opus 5 are like like almost objectively worse models than like.

23:32GLM 5.3, Kimi K3, even though they've had access to not only distillation but logic distillation from mythos. And so the counter here was that the prompt distribution really, really matters. You need to see what users are doing so that you can distill these behaviors and things in. I think the prediction from this is that the frontier labs don't necessarily have much of an advantage, if at all, in oral environments now. Because yes, user distribution matters for general behavior.

23:59and so on but like the best measure of a capability is the very very hard RL environments you've made at the frontier. And so if you have access to those RL environments as anthropic and you have access to logitizflation and you've still made a worse model then maybe like- The real world deployment matters more than the environment. That's really interesting. So but they had to incentivize those capabilities in the first place in Fable or the frontier model. And it's weird that they can't incentivize them again.

24:28Or would they get a smaller model or something? Maybe we're just in this weird uncanny valley where actually trying to copy that frontier model too much. The student teaching app, whatever it is, is just too large. And I think people made this point with Opus is it's like.

24:43The difference between Opus 4.6 and Opus 5 is that Opus 5 really feels like it's got this AI as a judge checking every possible thing it's done. That's why it uses so many tokens. It tries to think about all these things, but it doesn't necessarily have the big model smell of fable to know when to stop doing that or when's a good path to go down or whatever. The reach exceeds the grass, yeah. Yeah, it would offer a slightly different hypothesis. So I would say there are a couple of different axes for the environments you can create.

25:12And like one of them is difficulty and the other is realism. It's sort of easy to create or it's comparatively easy to create a lot of difficult environments like that are just like involve like doing a much more complicated task or doing something that requires a lot more cleverness. And you could say this is like the benchmarking distribution because a lot of the most prominent benchmarks just involve doing some very hard puzzle-like task that's easy to verify. And then there's sort of like the realism axis where you want the model to be good in the realistic coding agent setting where there's like multiple back and forth with the human and there's like multiple objectives. And like I'd say like the people like the labs who are

25:59Crafting the model behavior for the first time needs to push in both directions. And to get good model behavior, you need to really push on the realism axis and have like rubrics or some kind of human feedback that's informing the reward function you use there. But I think when if you try to do distillation naively, you end up just sort of matching the teacher on the benchmarking distribution.

26:24But if you don't have enough of the environments that really exercise the capabilities in these tricky or realistic settings, then you're not going to get those into your student model. And I think maybe one thing that's happening is the big models generalize better from the tricky, narrow tasks to these sort of.

26:45more realistic tasks. So if you have a really good realistic prompt distribution for distillation, you can match the big model really well. But if you only have this distribution of easily verifiable tasks, then you can match the big model on all the benchmarks. But you do worse on this broader distribution. So that might even explain something about the smaller, anthropic models like Sonnet 5, though it's hard to predict exactly what they're doing to post-training those models. It could also be that they're always changing their post-training stack, and they just made, they just got a few things wrong in some of these models. So they, like, I don't know, they turned up like some.

27:36something too high and created some quirks that people really don't like. So it's like really easy to screw up post-training in some way that doesn't show up in benchmarks. I mean, just one other sort of very basic point is just like the front AI labs buy all their data from the data companies and like the Chinese can also just buy the same data from data companies. And they are. And like they are. Exactly. There's a lot of people like, you know, being annoyed about this, but like if they have exactly the same data and like they can buy that, they can also distill. It's like it's it means it's quite easy to like keep up really.

28:05Yeah, yeah, yeah. Okay, the other question I had is how the first models that are capable of automating AI R&D will actually be trained? Because there's a toy version, which is this thing that Ryan was talking about, which is you just have GPT-8 try to build GPT-3 size models that are really good at like inner loop type challenges of beating video games that require continual learning or just getting to a certain loss with like the least amount of compute, etc.

28:35But John, I think you had an interesting point that maybe that's not the way it actually will happen in practice. So I'd be curious about, yeah, by the point which you have you guys that are actually capable of automating R&D, how are they probably trained? Yeah, I think we'll probably do some combination of learning from human feedback to absorb like the researchers taste and just like creating a lot of practice environments which involve like doing multi-step research projects. So I think, yeah.

29:03people will in practice do some combination of those two things and just each iteration like patch whatever seems to be most broken in the last iteration. So researchers will be using the AIs a lot and will notice that they have some consistent weaknesses and then those things will either be patched by collecting human feedback or creating environments. Yeah, that makes sense. Maybe useful to think about this is like.

29:32how much of the lineage we roll back and then let self play from there. I think in the limit like you're picturing like you know just giving them like a GPU and maybe neural nets or something and saying like okay figure out how to train a model to like do this particular task. Like the way it currently works is like we go up to the very like edge of the lineage and say okay like here are the bugs like you know Anthropica is found in their training stack in the last few months we'll turn those into environments like.

29:57you need to try to get better on the frontier and so you obviously lock in all the previous history of the lineage but you could imagine a world in which you roll back to like, you know, before GRPO or something and then you have environments which like trying to get it to discover like the best will form to like RL models on and then maybe you roll further and further back but I think we will be still So compute bottleneck that people will just keep staying at the frontier and diffing essentially the bugs and whatever improvements they found since the last model version turning those into training environments. Which is also really good for having non-steal new data between the model generations. Again, this is basically continual learning within the AI lab of distilling the last three months of AI research progress through environments and like RLHFs type stuff back into the model itself. And it is distilling, right? And that's maybe why.

30:44some of us feel like it's asymptotic is like you're always like just trying to get the last three months of progress and that progress is being contributed to by AI as of course but it also still has humans in the loop and it feels like you know you're just constantly inching closer and closer to what the human researchers are like finding capable of doing. I mean the one thing I will say though is like obviously if you're just distilling on like trajectories you can never go above it but environments can go quite a far way above what a human can do like it's very easy to design an environment that like no human can solve but the AI can always you still try and solve it.

31:12And so that would be the path to like go ahead of just like with the human AI research. Do you have like an example of like in terms of RSI or like? With an AI speedrun, but doing it even faster than a human speedrunner. Yeah. I mean I feel like in AI research especially it's very easy to define like goals which like you know you could say like the loss needs to be like 1.3 or something and like no human can get that you know now but like that's a very extremely measurable verifiable task and if the AI gets that then great. Right on a building like a hundred million parameter model that Beast might craft.

31:41that's maybe too easy, but beats a much more complicated game or something. Isn't it crazy that a hundred million primitive models would be at Minecraft would call it not too easy, like imagine if you said that like five years ago. I would say a lot of research is not exactly like that though, where it's like hill climbing on a well-defined goal. It's sort of more like here's an intuition we have about some way models should be better and then we also have some idea for an algorithm that seems to go a little bit in this direction. So let's come up with a task that is sort of designed to show signs of life on this approach and see if we get those signs of life and then if we do, we can make successively more realistic versions of the task. Right, it's like a lot more guided by intuition. And then the inner loop is to elicit the or make

32:35test for that intuition rather than the test itself leading to the insight. Right. You're not directly optimizing for the eventual objective you care about or the practical production objective. You're relaxing your objective a little bit. You're saying, yeah, let's relax on the realism axis a little bit and find some methods that actually work and then try to get back to realism later after the method matures a little bit.

33:05And then there's research that's more oriented towards explaining things and developing a theory. Often we don't have mathematical theories in machine learning that are that predictive, but we have a lot of.

33:23more informal theories for what's going on. Yeah, I mean like presumably the models will be trained on like some combination of all of these tasks and like some will be very easily verifiable, some will be like L and misjudge or like just ask the human like does this look reasonable and then you will say the hope would be that like these would all generalize to like these much sort of hard sort of more vague fuzzy kind of tasks and like it probably will to some extent whether it generalizes enough that like we could the loop can become like self-sealing without humans being in the loop at all. It's like unclear. Yeah.

33:51Maybe taking a step back. Here's what it seems to me that the plan for AI research going forward is. And you tell me if you think it's going to work or if you agree with this characterization. So the bet is that we will scale up RLVR training across millions of diverse environments across hundreds of different kinds of domains. And what will emerge at the other end is an agent.

34:16which has like learned these basic skills or less than basic skills around being persistent, being able to triage information and context, eventually having like end-to-end optimization of working with other agents and things like that. And such an agent will be very sample efficient within the context.

34:35You've done research on how you actually scale up in context learning to make it arbitrarily long, but you just keep scaling it up. And so what comes out of the other end will be something that basically functions like a drop in remote worker over the course of a week or a month. First of all, do you agree that that is the bet the labs are making? And second, is that enough? Like basically learning how to learn within the simulacra within a data center.

34:58And then getting deployed into the real world, but not actually learning from real world deployment, only learning these meta skills from the simulated environments in the data center. Yeah, I think it's now hard to separate out how much of the lab's effort is going towards direct RSI versus making generally intelligent models that they can continue to deploy to collect revenue to fund the next big training run. I think for the latter, yes, that's probably just the bet they're making.

35:26And it's very clear the pattern of where these environments are going over the last few years. Anthropics lineage of environments is a very clear example of this. First we just focus on coding and we're going to get really, really good at that. And then the task horizon that we've got from coding, which is probably the lowest hanging fruit in terms of data available on the internet to create environments, their own internal stuff that they can turn into environments. Then we're going to generalize, we're going to go after finance next.

35:53so much Excel data and all that sort of stuff in the RR training. And then it's PowerPoints. It's like this long tail of the working economy. And that seemed to work really well. And a lot of the other labs and even the open source labs have now realized that that was the correct bet. But what is the implication from that? When I had Dario on the podcast, the thing I asked him was, if you truly expect models, which will be human in their ability to learn on the job, why would you try to beacon all these of like working with PowerPoint or something. Wouldn't you just expect the model to be able to pick that up on while it's deployed? And so yeah, there's multiple different explanations. One is just that this is, we expect models to get there soon, but they're not there yet. So why not amortize these skills into the model training? Another is that we're not concentrated on making it really good at widely deployed work. We just want it really good at RSI. And this is just like a way for us to like get revenue so that we can forward back into.

36:47a model that is actually really good at doing our site development, and then once the singularity happens, the thing that comes out the other end will be really good at all the things which seem like modeling next to the current generation of models. John, I don't know if you have a taste on what you want to construe, why there is so much task-specific knowledge in these models, if the path is this kind of generalization. Yeah. I mean, if the models were good enough at learning in context, then in theory, you wouldn't need to train them on finance, they would just be able to figure out, read all the books on the fly and figure out how to do everything in the appropriate jurisdiction. Yeah, and you could argue that you need to do a lot of this domain specific training just to make them more efficient. So even if they were smart enough to figure this out on the fly, you still might want to do a bunch of RL and bake the

37:43they call these intuitions into the weights. So the model would be more efficient at runtime. Yeah, I'd say in practice, it does seem like model providers are going domain by domain and trying to strengthen the models in the highest value domain. And I'd say that that's one of the answers to why the models have gotten so much better. It's just because the model providers have covered a lot of the high value domains and the most common types of skills.

38:10I mean I think another thing is just like it's not that expensive to do both at the same time right because like the models are massive they can easily afford in terms of their parameters to like learn everything. And like there is likely some transfer in sort of even just even if like finance is not specifically like the information is important for like RSI just the general like meta learning of like how to figure what's important how to have taste how to like do long-horizon work is potentially generalizable and like there's not that much RSI like data in the world as well. Like it's kind of hard to generate and like that requires a lot of effort to like if you can amortize in this other data gets

38:40some transfer format, you already have masses of compute and masses of parameters based on why not do that as well as obviously the direct commercial intent of selling a model. That makes sense. Yeah, I'll add that. I mean, there's one question about whether this current paradigm of doing sim to real will be the dominant one forever. So basically, you look at what the real world tasks are like, and then you try to create a bunch of environments that can be simulated in the data center.

39:09you can do RL on them. And I think obviously this has been very successful, but it has a lot of weaknesses because a lot of things are just kind of hard to simulate, especially if they involve like interacting with a bunch of humans in real time. Yeah, so there's some question about like whether SIM2Real will be the dominant.

39:30framework forever. I think some tool has to be the dominant framework where like sample efficiency is kind of low because like right now you need like you know thousands of thousands of interactions with the humans and no human is going to sit there and like deal with this basically be in the loop of oil training. Yeah. And so like we kind of have to simulate that now to like get the samples you need but like obviously if sample efficiency improves a lot you'd expect learning from deployment to like become like a much bigger part of it.

39:51though there are also other things you could do like you can learn off policy so you can take all the traces and even without re-simulating everything you can potentially learn something from them. Jane Street just launched a new competition and it's their most ambitious one yet. Design a protocol emulator ASIC. Basically if you have a chip that you want to test you can connect it to this ASIC and then this ASIC will simulate realistic traffic.

40:15That way, you can see how the chip responds without having to plug it into a live system. Jane Street is looking for flexible, general-purpose designs, not single protocol emulators. When I was chatting with them, they suggested that I start off by trying to implement what are apparently three very common protocols, UART, SBI, and I2C. Jane Street also mentioned that they hoped that more ambitious designs will also tackle low-speed USB and Ethernet, and any other protocols that flex your chip's specific architecture. Importantly, your design should be reprogrammable rather than smashing a bunch of specific protocols onto a chip. If a new protocol comes out after your ASIC is taped out, your chip still needs to be able to handle it. How exactly it does it is up to you, but there is one hard constraint. Your design must target an open source 130 nanometer process node. That's because Jean Street will pay to tape out the most novel submissions and send the physical copies to the winners. The competition is open till January 18th, 2027.

41:13and working in teams is highly encouraged. Go to jainstreet.com.org to download the template code and get started. I want to ask more about this because it's sort of weird that you have 50% of compute that's spent on inference that is not directly helping the model become better. Like one of the key advantages you'd expect eventually digital minds to have is unlike a human who gets to have 50 years of like real world experience, A model will get to, through all its instances, will get to experience, I don't know, millions of years of deployment across all kinds of economically relevant work in the economy. And right now, that data is just not in a meaningful sense helping the model get better. It just seems so obvious that eventually model should be able to learn from this data. And once they do, you would have something that almost feels like a widely deployed intelligence explosion because the model is assimilating so much information.

42:08across all these deployed instances. But when do you expect this kind of hive mind kind of crazy shit to start happening? I think broadly like at a very basic level this is already happening, right? Like just in the next generation of models. So like right now you can always take your deployment data and put this in the pre-trained or the mid-trained of like future models. Especially if you do like some kind of filtering or some kind of like judgment or annotation or like recent you know synthesization of that. How much do you think that explains the generation over generation improvement? I think it explains like quite a bit. I mean especially like I mean this is you know

42:35I don't know whether the labs do this because theoretically they claim not to train on people's data but like the Chinese 100% do and like they definitely get this advantage both like obviously deploying this is basically what distillation is like they take other models they get some of their like deployment data they get some fraction of that by like pinging the model and then they train their next generation of models on it and they can suddenly do it on their own models as well like there's no reason not to whatsoever. I can clearly agree with this. I think if you zoom out far enough this is like definitely happening like you're picturing this like and we're all picturing this is like what continual learning like the holy grail is is like

43:05this very, very organic live loop of an individual model getting an experience and live updating on the spot and learning from that. And a lot of things break when you zoom into that level of granularity, but the big labs are doing this, the closed models are doing this. There's also early signs of life of people using open source models doing this at a much faster cadence. So a good example is probably Composer. You have some sort of model and you are able to, or Harvey's doing the same thing with legal.

43:34legal agents. It is getting very specific environments from the data that you have for that particular task and things that users are complaining about. And all the feedback that you're somehow extracting from your specific deployments. And a lot of these companies have the advantage over the big labs in that they can use this data really, really well. And then they will create environments. They will do a big post train of Kimmy K3. They will go deploy it. They might do some online learning as well. Composer did online, basically reinforced for a long time.

44:03So yeah, there's still a human in the loop. There's still a human saying, okay, these are the signals we care about. Here's how we're going to create environments from the data that we have. And there's still a longer cadence than maybe the one that you're thinking of, but it really is happening and eventually that loop will become faster and faster. I mean, the composer thing is interesting because this is where the model, in cursor, people press tab or they don't press tab on the next completion that the model suggests.

44:26And based on that, every single day composer gets better at predicting the next year. So that was the old tab model. They actually did the same thing for the actual, not just the tab model, but the actual generative model. Oh, that's interesting. And it's hard because when you do online reinforcement learning, you don't have groups, right? You just have one user saying one thing and then you get one rollout. And so you have a big variance reduction problem. And Curse's kind of fuzzy answer to this was like, oh, we have very good heuristics which we're able to estimate.

44:52how much better than average this response was or how much worse than average this response was. And then they would do this big reinforce update. And then their solution to whether it got worse or not was if it improved on cursor bench, they would deploy the new model every five hours. If it didn't, they would throw that version out. Interesting. Yeah, I think your biggest problem is actually just not knowing what the reward function should be from natural data. And if you use some kind of superficial signal, did they accept the the code, the edit, that might get reward hacked in some way. But this seems like a bigger issue with the Sim2Real thing, where the longer and longer horizon tasks get, the harder they are to simulate within a data center, right? It seems to me already, potentially, even in coding, we're getting to a point where there's not some your long coding task that doesn't eventually require you to

45:46talk to a client or interact with the company or interact with users. If you think about the gamut of things we would want ESP capable at, you want eventually super intelligent people to run a business or start a new business and make it profitable or have a profitable day trading in the markets or win a court case. These are all things which are very hard to simulate in a data center. Inherent part of the learning there is interacting with the real world. Maybe they had to learn how to get better at these things from the transfer between sim to real.

46:14But alternatively, maybe you do need weight updates from these kinds of interactions in order to get better at them. And then if that is the case, if transfer isn't strong enough and you do need weight updates, then the fact that the models are quite sample inefficient is like maybe a deeper problem. And the reason I'm curious about this, I feel like by default, I don't see how you don't get some kind of crazy recursive stuff from Prevent within the next 10 years. But the one reason why that might not happen is in terms of like weight updates, The sample efficiency of weight updates, they just seem way far behind humans, right? Like plausibly a million fold behind humans in terms of how much data a human sees from birth to adulthood versus how much a model sees from cold start to finishing training. And so, yeah, this is all to say, first of all, is there going to be good transfer between simulations and extremely long horizon, really complicated real shit that we want the EIS to do in the real world? And if not, does that really mean that the actual sample efficiency in these models comes to bite us?

47:14I think maybe the way I'd break down the two types of tasks in which models get good and models will still continue to struggle is whether the task is cumulative or you have this non-stationary distribution you have to keep learning and re-litigating a bunch of stuff. So maybe an example of a cumulative task might be RSI.

47:32It's theoretically possible to maybe have a less than a million token Python file, which from scratch trains a model that is capable of recursive self-improvement. And every discovery that you make is kind of a line in the sand that you hold. If it's true that for RSI, we don't need to discover a new attention barrier and whatever. Once you've discovered attention and then once you've discovered mixture of experts, once you discover JRPO, you just add that to the training stack and that's there. And a good example of this is 5.6 sole training.

48:015.6 terrible, whichever one opening I told you to train. It didn't have to go back and discover attention. It basically probably would have called a bunch of scripts, which is pre-training.sh and post-training.sh and just did that. So that's an example of a cumulative task. I think the real world and the reason people are thinking so much about continual learning is it's not really a cumulative task. Imagine in a law firm, you have an agent acting as a legal associate.

48:25That's a very non-stationary distribution. You have to be able to fit in your context all the relationships between all the important people at that company, which are also changing all the time. You have all these implicit ways about how things are done, where to find information, et cetera. That's not as clean of an example of accumulated parcels like RSI is.

48:44I think that there will be this breakdown between tiles. But if the labs realize that and they do believe that RSI is cumulative in the sense that we don't need to go back and discover some brand new architecture or whatever, then maybe more and more effort and compute gets focused on that. It's so unfortunate that RSI happened to be easier than they apparently are. Yeah, I don't know if you guys have thoughts on this. Yeah, I would say there's like models. Today's models are weaker than humans in a lot of different ways.

49:13some of them might have to do with sample efficiency in a certain regime where, I mean in some regimes models are very sample efficient like learning in context. But then there might be some like medium length regime where they're less sample efficient because humans can do some kind of weight update more efficiently than models. So I think like being less sample efficient in certain regimes might be.

49:40one of the sources of weakness, but then I think there are other sources of weaknesses that are completely different than that. For example, having lower diversity of thought than humans or being bad at certain kinds of long horizon judgments. I mean, I think a lot of what people call taste is something about behavior that works in the long run and that people have realized works in the long run.

50:09not everything, but some aspect of taste, especially for something like software engineering, I think a lot of taste is what are the systems that are going to be maintainable and work well in the long run of this project. So I think the weaknesses of humans, which limit RSI along with other things, there's a variety of them and some of them are related to sample efficiency and some of them aren't.

50:37maybe an interesting thought experiment is like if you were able to give a model like a context window of I don't know a trillion tokens or whatever you would have needed to fit in like your experience prior to like let's say RLHF and like it's got all that experience in the context window and it has the same sample efficiency and in context learning ability as it does at a million tokens like do you think taste is then solved like would it be able to like make the same judgments that you did or is there like something fundamentally missing apart from just a longer context window with the same sample efficiency.

51:07Yeah, I mean, it would have to be trained to learn from that context. So I'm not sure. Yeah, either it would have to be trained to learn the right update to make from that context or you just like dump it all in like your whole like life, like research experience. I mean, like you still need the data to train it long context, right? Like even if you could theoretically get like a trillion context, you would need a trillion lengths of data to train it like right now if you have like. I'm just asking if you had that.

51:37Very, I think yes. I mean this really just comes down to the question of like how meta-learnable is taste from like short to horizon episodes. And I feel like there's no obvious reason it's super long, because humans somehow develop taste with not having many long episodes. We don't live to be like 10,000. We have like, we develop pretty quickly, right? And so if you think about, even in a PhD, the difference between a first year PhD student and a final postdoc or something, that's like five years maybe. And they've only done maybe like 10, 30 research projects in total, but somehow they develop taste quite quickly from a relatively short succession of small things. And so theoretically, it's

52:12possible to develop it like that. The AI obviously will have vastly more experience in which to develop tastes like meta-learned and there's like how well does that generalize to like really long horizon things is I think the question which I think is really unsolved at this point like we don't know. Going back to this question, eventually there should be a regime where AI's are learning a ton from each individual instance of deployment that they have. Well currently you could say there's a meta fuzzy process by which models do improve for deployment but I feel like it's a very weak.

52:41very weak feedback loop. Do you see this around the horizon where there's this like hive mind kind of learning that's very rapid? And if so, how exactly does it happen? Actually, I would say that around, will we get a hive mind that learns from all of its deployment experience? I mean, a big part of that is actually about incentives rather than being a technical question.

53:06companies aren't going to want to have the model provider learn from all of their deployment because that might just reduce the advantage of their business. I think that maybe the economics of this will pressure not necessarily weight updates to one big common shared model, but kind of like modules that get subbed in. So a very obvious example, this is a Laura, but it might be something else like there's been a lot of work to.

53:32try and fit an arbitrary context length into a fixed size. This is all the linear attention stuff and all that sort of stuff and cartridges which are essentially KVKs trained to be very, very compressed KVKs to fit in a lot of information. That's another example of something that companies may be willing to sign up for if that's getting subbed into the model and it's not actually changing the base underlying model itself.

53:53There's many different versions of learning from your data in real time and the latter ones are not really helping the big labs because they are just these modules. But I think the economic pressure will force the labs to go down that path first before they.

54:07can embark on this like, you know. Which economic pressure this I feel like even if you have like a bunch of cartridges or laws or whatnot, you can still just like take all these traces and just like distill this dub this to the pre training of like your next generation. Yes. So it may be a more indirect form of learning that the labs are getting. And that's obviously still really valuable to them. But I can't imagine a world in which we start off with like, you know, we're going to just directly train this one big model on all the exact data. No, I think it will definitely go through stages because I mean this is assuming there's one discontinuous event where suddenly we fix weight updates continuously and in practice I think it's much more likely to be like.

54:41The cartridges and stuff allow you to specialize in deployment. Then you generate traces, you put that in your model. Like three months later you come out with a model which is better at this stuff. You specialize it again, you like consolidate it again. And then eventually we'll just like make this loop faster and faster. So instead of like every three months we will release a model. Now it's like every week and then every like day and then every hour and which point we basically have always resolved it. And I think this is a good point as well because you asked like kind of how far off the current paradigm we are from being able to do this. We've done a bit of research of this and people have done a lot of research.

55:09at a really large scale like when you wash out enough noise and you have large enough patches like this outer loop process of like putting data into mid-training creating our own environments like it does work in like some sort of continuing regime but the problem is like when you're zooming close enough at like a micro level it's like I've got one model and I'm trying to update it again for like a law firm or something and I'm trying to do that very continuously like with a relatively small amount of data like all the methods kind of break down a bit so like if I SFT the model on just like you know successful traces

55:39off policy, on policy, eventually in the very iterative regime, when you're doing hundreds of these micro updates, you see both catastrophic forgetting, you see forgetting of previous information I've learned on top of the base model that was much earlier on. And I see degradation of general use, general capabilities. On policy distillation seems to push this horizon out a little bit, but it still eventually succumbs to the same thing.

56:04And RL is not very good at like, it is good at like getting capabilities in, but it's not as good as getting like knowledge in. And like just this very explicit knowledge of like, okay, like this person does this law firm and like this is a very specific person we find. And you have to pour in a lot of compute to create the right environments to get the knowledge of this RL. Do you think that the fundamental issue here, why you get worse at these other skills or there's forgetting and stuff, do you think it's fundamentally an issue of capacity?

56:29Or it's an issue of techniques? A little bit of both. I think like SFT and even like on policy distillation can be like way too destructive. Like the reason RL is so nice is because like yeah it changes a very very small amount about the model and there's like a lot of evidence for why this is the case. And so like it kind of just like tweaks it in this very very very small like loss value to like get it into the right point. But that also then limits what you can do with RL like how much you can actually change the model.

56:57You're saying the reason this is the winner take-all potentially is that it's just very hard to distill that much information into the base model. Without ruining something in an editor. It's easy to distill it into a different base model. This is why I think it's mostly technique. It's not like...

57:11it's definitely not like just like there isn't capacity. Like if you had some modeling with all this data and you take like literally the same size model and pre-trained from scratch with like all of this stuff in mid-training, it will be better and I think that's a lot of what's happening today. And so it's very much like there's a bottleneck that stops us from just keeping training the same model forever versus just like getting all the data from the old model and like training a new model from scratch. And this is exactly.

57:31I was saying some combination like Plasticity and Catastrophe forgetting. If you just naively train on non-stationary data, because you're adding new data as you go, basically this is messing with the data distribution. So the old stuff is just forgotten. And we don't really have good methods to stop that from happening. And maybe at the limit, you're just bottlenecked by retraining the model from scratch with all this new information. Yes, which of course is very expensive. Training model from scratch is expensive. But you're going to do that anyways. Not necessarily. Maybe eventually, if you have continued learning, you never train any model.

58:00have a model and it keeps learning and it keeps expanding right? But there might be some deep technical reason why that's very difficult because of these like- I mean that's the question. That's the question. I think we have pushed back like how much from scratch we need to do. Like it is definitely possible now to take like the pre-trained base and like do very good mid-training on top of that like kind of continuously plus some RL from like different checkpoints that are later on in the training and like- That's looking more like a tuning line, but certainly not the case of like, you know, take the most recent model, apply a couple of very small updates and like iteratively like never lose it. But isn't this like, I'm a bit confused because isn't this literally what happens during training or during post training or something. You just have like, you have a model that's already gone through so much training.

58:39And then you distill some fork that's been further RL or something. Isn't that literally what happens? But it's still at a large enough scale, I think, that you're washing out a lot of the noise. And you're not just focused on one distribution, which, as Baron said, is like, that is now a very, if you're just focusing on one task, that's not a distribution. In the eventual regime, you'd be doing, I don't know, there's billions of deployed instances. You're learning from all of them at once. And so hopefully there's some.

59:07Washing out of noise and stuff from that right? Maybe that's go. Yeah Yeah, I mean I think you like definitely is a child saying like you can do continual mid-training for like a long time And you can like well back to a checkpoint given you mid-training data Yeah, but at the same time like you can't do this like indefinitely like if you just keep continuing me training the same base forever It just like get it does it sort of asymptote at some point like you can't just learn new stuff in that base And this is why people end up training new bases like otherwise you just keep me training the same base forever Whenever I finish recording an interview I immediately brain dump all my thoughts into slack

59:36Things like, what was the most interesting? And what should get cut? This ensures that my editors have all the context they need to start editing the episode. But it's not like these brain dumps have any clear timestamps, and my unedited recordings are many hours long. It can take a ton of editor time to even find the exact moments that I was referencing. So we decided to try adding a GrockBot producer to our chat. And now, whenever one of my editors posts a rough cut of the episode, GrockBot opens a transcript on its own computer and starts working.

1:00:04usually before I've even seen the message. It takes the notes that I dropped in a slack and it highlights the relevant snippets in the transcript. It also uses a big case file that I've compiled with all my preferences, so it can suggest potential edits. And when it's done, it sends me its top clip candidates so that I can review everything from my phone. This has worked really well. Being able to send informal messages like I'm texting my editor and then having the transcript immediately reflect my preferences has just been so helpful. Try GrockBot yourself at x.ai slash bot.

1:00:33Okay, let's talk a bit about data now. So I'm generally interested in this question of how much of AI progress is just explained by data progress. It doesn't mean it will be necessarily hard to automate whether this is ever a question. So is there some data distribution which if you trained current architectures on would result in a super intelligence that totally dominates human experts across every single field? Are we talking about like pre-training plus post-training data like environments as well?

1:01:03I think the existence of this is obvious. It's just like whether we can create the right environment. In the trivial case, we could just train it to output the Python file which like trains the actual superintelligence, like just have a memorize in the weights. Yes, there's probably like a ladder of RL environments that is possible to construct such that you would get a research which is at least as good as a human researcher, but the effort to climb each successive wrong grows like.

1:01:31kind of exponentially. And that's going to be the two things that you have to trade off against as to whether like, like how fast we're going to hit like that final run where it's where it's better. I think that's fairly clear. I think there's like, you know, we're still relatively early in like our own environment creation. Like there's a lot of asymmetries that we exploit in order to create good environments. So one of those asymmetries which we've talked about before is like, there's environments where it's easier to go backwards and forwards. And like what I mean by that is like, It's very easy to define this complex data generating process and this is this kind of latent variable you keep hidden from the model. You can generate arbitrarily complex environments and the model has to do a lot of irreducible token spend and irreducible work to figure out what that data generating process was. There's asymmetries in terms of you can inject information from the world. Anthropic finds a bug through tens of thousands of human and LMLs combined and turn that into a very, very neat environment.

1:02:26a single LLM because they're radically fine within a few million tokens. So there's all these asymmetries which we're cherry picking and we're counting on this task, horizon generalization. But I think, yeah, again, there's just going to hit diminishing returns at some point. At some point, it's diminishing returns and how hard it is to create those environments in the first place, coming up with them because you can't necessarily just have these processes where it's easier to go backwards than forwards. You actually have to sit down and construct something that looks With humans, a long enough time horizon, it's going to be a really complex task to create. And then there's also going to be the compute and time bubble next for the agent to actually do those tasks. So I think you're just going to start seeing this curve to flat now. I saw something about how someone fine-tuned the Taki model, which is only trained on data up to 1930 on this modern coding agent data. And it did better than Claude III opus on sweep bench.

1:03:23So this model that has no knowledge of code whatsoever can be fine tuned on a moderate amount of data and behave better as a coding agent than this much larger pre-trained model is pretty crazy and it kind of shows you that once you have an example of the right expert behavior, it's actually surprisingly easy to copy that into a relatively weak model. But a counter example to that.

1:03:51kind of is that there was a paper recently where they trained it up to like fifth grade maths. And like also like primary school like English and stuff so it was like a decent language model and they tried to RL it to do like you know late high school and college maths and the gap was just too large like they couldn't get it to climb at all like but if you did like successive wrongs of like you know you seven maths and then your eight maths and so on like you could obviously climb to year 12 so like again it's just like what is the distance between the wrongs on those letters and how hard is it to create.

1:04:18Yeah, and this just comes back to like the RL signal problem. Like RL is not very good at like exploring right now. And so if the model can't like get in like, you know, 128 relapse, it's very unlikely to get signal to like progress. And this is why like in RL, we need like curricula, whereas like in pre-training we don't, because like it's that's not a problem for pre-training at all. Yeah. And again, pre-training data is different to post-training data. And I imagine as we continue on like, yeah, humans will be involved less and less, but that doesn't change the fact that you're bottlenecked on like.

1:04:46how much signal you can extract from the real world. So there's a lot of signal in the world and that's true. There's people doing spreadsheet tasks, there's people doing legal tasks and all this sort of stuff. But the capability frontier of where the models are at now, how many bits in the world are actually really relevant to improving the model's capabilities? How many new maths problems are being solved? They're just beyond the reach or grasp of the current models. How many new coding problems are being created or solved that will be on the reach of the current models?

1:05:16I think that's why the diminishing returns kicks in because even the world as a whole is not giving you the bits that are useful for tipping you into the next like basin of capability. Yeah, I totally agree with this. It's like really a question like where the signal is coming from. And so like the signal doesn't, you know, in pre-training the signal is like already in common call, right? Like for the tasks that you care about in pre-training.

1:05:36The problem is there's not just like getting signal at all. It's like filtering out all the noise that exists. And that's quite an automatable process. But as the models get better as we end and to mid-training and post-training, the signal just doesn't exist anywhere in the original data we have. No amount of filtering will get this. There's no hidden proof of the Millennium Prize problem sitting in Common Core. We can just filter until we see it. And so at that point, you have to get bits some other way, either from humans directly asking them to write out their reasoning or by creating environments where humans decide what environment should be created.

1:06:05what the objectives of these environments are, or some kind of training on the human data that exists in deployment. You have to get the bits from somewhere. Yeah, yeah. There's a question of how much of the progress in pre-training is being driven by data. I did this investigation with Jerry Hahn, who's a student at Princeton, where we basically trained all the recipes from 2019 till now, pairwise with all the data sets from 2019 to now. So you change like GPT2 on the newest data set like Ultra Fine Web and you train Delphi which is the newest training recipe or the open source training recipe on like the pile or some old data set and you do like the whole grid and you see the getting to some level of capabilities how much less compute does it take across this grid. And you see that the data seems to explain like 9x of a compute efficiency gain.

1:06:59But the architecture improvements explains the 3x compute efficiency gain at a very small scale. And so to the extent that that is true at large scale, that most of the pre-training compute efficiency gains are coming from better data, how much can that continue? Can you keep just filtering data more and more and building more and more synthetic data until, yeah, do you have a sense of how much this kind of pre-training progress can continue? I think my prior is that like, Again the low hanging fruit is like somewhat exhausted with like we got the internet as this big block and Like there's it's not like the internet is necessarily like growing at the same like ratio all the useful stuff in the Internet is growing at the same rate So like we've probably got like a bunch of like point one cent lost drops to go But like not definitely not as many as have currently occurred. Yeah, but like that's also really interesting They're like, you know, you find this like what cumulative like 27 times improvement across both I think like it was epoch or someone who estimated like

1:07:54three times a year since 2019, which would imply something like three to the seven, over 2,000 times improvement. So where's that missing 100 times or whatever coming from? That probably gives you a good signal of how much this is post-training. I think the explanation has to be that a lot of the compute efficiency gains are scale dependent. And we were studying at extremely small scale. And that raises the question of, do the data compute efficiency gains?

1:08:21or the algorithmic computer efficiency gains have more scale dependence. I don't know if you guys are prior on that. We just didn't have enough computer to investigate that question. I mean like just naively right like the theoretically the scale dependence of the architecture is like fairly well known. Yeah. And like you can fit a straight line to it whereas like I would have no idea how to do that for like combining pre-training plus post-training data and mid-training data. I feel like data is actually more important with scale. I feel like architecture is kind of like a one-time, you know, an architect. They're like combining like...

1:08:53saying just like an X percent efficiency gain is kind of misleading because like what an architecture does is like letting you reach like a qualitatively new regime which you couldn't reach with the old architecture and then within that regime obviously the data is like the primary thing determining it. But like you know if we say it didn't have like even like GQA we're doing like full attentional day we wouldn't be able to do like a million it would be like ridiculous expenses to a million context and then like because of that we couldn't we can never use the data which is like actually at a million context and so we couldn't get these capabilities even though like if you just do a naive like how much does this do at like 2k context where the

1:09:23architecture isn't unlocking anything then like the data you know that will look much more important than in some sense it is right it's unclear to me that these things are like really just like multiplicative gains in this way. I see sorry but then what does it take away from the scale dependence of data? So I mean on scale dependence I think like A lot of the like mid-training and post-training data we have now is like actually but gets better with scale. Because like a lot of it like the very long context horizon of stuff really requires like big models to be able to like make use of them. And like this is not, you know, if you try and train like your 100 million parameter model on like three bench traces, it's not going to get anywhere. Like it's not going to show you the same kind of improvement that you would get if you train like an actual sensible size model on it. Yeah. And like it's hard as well now because so many of the architecture changes like you look at like Kimi for instance like or DeepSeq, they're

1:10:08doing these architectural modifications with not just like dropping the pre-training loss in mind, but like for instance how the models are going to be used in the real world. So like, yeah, the inference efficiency, like having some form of compressed attention in the deep-seq models is not necessarily geared around, you know, this is fundamentally like a pre-training improvement. It's just like, okay, we're considering how the models are going to be used. Right, right, right. One question I'm curious about to understand the future is how parameter scaling will go as we're getting into more RL heavy regime.

1:10:34I don't know how fast historically. Yeah, you can look at sort of open source architectures and see how fast parameters have been scaling. And maybe it's roughly 2x every year for frontier open source models. And to the extent that even frontier close source models have 100B or 200B active parameters. Do you think that keeps 2xing year over year or another RL regime where you also want to conserve compute on rollouts? And also, maybe there is a threshold effect where you have enough capacity.

1:11:03And at that point, increasing parameters arbitrarily doesn't matter as much. Do you guys have a sense of, in 2030, how many active parameters will a frontier model have? Yeah, I think for the next few years, we're going to be like, because we're so focused on doing longer and longer horizon rollouts for RL, where like inference efficiency matters a lot. It feels like the moles aren't necessarily saturated on their ability to do that, where the bottleneck is still the environments. And so we might see like a little bit of plateau. Like I have a feeling that, you know, like Mythos and the GPT models are much smaller than the 10 trillion parameter range that the people are talking about. Even just naively comparing the open source models, you can probably back out that conclusion. So yeah, probably for the next few years, I wouldn't imagine a huge growth in the number of parameters. But again, there's so many different things to trade off here. You decide the size of your model based on how much pre-training data you have and then the difficulty of the RL environments that you've got to train on. And you ideally want to get to the optimal point where

1:12:02you can get a decent pass at one or something on the hardest environments you have, and it wouldn't make sense to make a bigger model pass there because then you're just paying much more inference swaps than you need to. So there's a lot of inputs to this. It depends on how quickly Macore and then in-house these guys can scale up the complexity of the RL environments they're training on. I would expect the models to keep getting bigger just because people are scaling up compute and the GPUs are getting bigger.

1:12:27I would say exactly how much they get bigger depends a bit on the scaling laws in non-obvious ways. So one thing is that I think like data efficiency is going to be a bigger driver than compute efficiency of like the exact architectures people use. Now that we're getting to the regime where we're sort of running low on like high quality pre-training data. So that might affect how sparse you want to make the model. And then I also think we don't understand sparsity that well. And it's like parameters are a different resource than active parameters. And like sparsity has definitely increased a bit, but it's not clear that it's going to keep increasing without bound. There might be some kind of sweet spot. There's an argument that sparsity should make data efficiency worse because you might have to learn the same thing on multiple experts. So that's debatable.

1:13:24So I don't think we have a good enough theory of scaling laws that we really understand why sparsity is helping and how much it'll help and if that'll plateau at some point at a certain level of sparsity. So can you spell out exactly what the implication of data efficiency would be on? So it sounds like you'd say, well, there should be less sparsity, but what are the other implications on pre-amor scaling?

1:13:49I guess just that the scaling law, you're not necessarily looking for the most, you're not trying to optimize compute efficiency. So you have all your choices you can make on the architecture and each of these gives you a different scaling law. And then traditionally you would look at some kind of envelope based on compute. So you would look at performance versus compute and take the envelope of the best models.

1:14:16Like if we're making that decision based on data, so it's like, yeah, we're sort of assuming we can spend a lot of compute and like we're sort of data is on our x-axis instead of compute. Then we just get a different set of optima or a different set of models that are on that frontier. Yeah, and I also don't think that we've necessarily like, you know, doubled like the size of the models.

1:14:45every year for the last few years, the people have been training one trillion parameter models for at least a few years. There was even an open source one called Falcon, but Liam from Periodic Labs, I think, posted on Twitter about how an early experiment at OpenAI was training a one trillion parameter model that was very, very sparse. Yeah, that was what they did before OpenAI. That was like Google to switch transform. So it was very, very good at like.

1:15:09knowledge, but terrible at reasoning because it was so sparse. And so it feels like we've been playing in this $100 billion to up to $2 trillion parameter range for at least a little bit, and it certainly hasn't been. This is a nice linear increase. I mean, I feel like there's two things. So as Charlie was saying, like inference efficiency is super important for RL rollouts. And so this will really push down active parameters quite a lot. And then I think the total parameters really depends a lot on the hardware as well.

1:15:35So you really need to get very high memory bandwidth and VRAM size to actually be able to serve multi-trillion parameter models. And so right now, people are still using a lot of H100s and stuff. And so as everyone moves to GBs and then their rubens will get the ability to scale and actually serve and do larger RL inputs at larger scales.

1:15:55The data question I think is interesting because naively larger models are much more sample efficient in the actual data points. And so even if you're not saturating the model, it's still better to go bigger because the larger models generalize better and get to a better loss for the same amount of data. And so right now I think we're kind of.

1:16:11have a lot of data, and that's not the constraint rather than computing. So we're having small models, which are very infant sufficient. But if computers are no longer the bottleneck, it might come back to larger models which are sort of under saturated, but they have this generalization ability because they're much larger. If you just look at the basic exchange of less scaling law, and you just maximize out parameters, it actually decreases the amount of data you need to get to the same loss very little.

1:16:36Yes. If you go to infinity on parameters, the amount of data you need I think goes on less than 10x just because of the nature of the power line. But we're now on the way too much data side of the Chinchilla laws, right? So right now we over-train more of the Chinchilla and so we could easily get back to a point to which as we're running our data we move back to the Chinchilla optimal points. We even like a bit on the over-training and the training model side. But surely like even with these new chips that come online and stuff like.

1:17:01We're just going to be so compute bottlenecked for the next few years that that won't necessarily be the case. This depends on the ratio you have training and inference to compute really. It's like if you're super bottlenecked on data, not on compute, you should go bigger. If you're super bottlenecked on compute, you should always go smaller. But you can also use computer-generated synthetic data. So it's one of these very hard things to predict. Yeah, I think part of the reason it took people so long to figure out the scaling laws in the first place was that if you don't get all these things right, then

1:17:27you don't get such a clean relationship. And the beautiful straight lines on graphs hide a lot of complexity on how you have to make sure to scale every hyperparameter the right way or parameterize your optimizer in a way that scales and where you don't have to change your hyperparameters as you change the model size. And bugs have their own clean scaling rules as well, right? Like with Kaplan for getting the.

1:17:52cosine and alien thing or like even just like not considering embedding parameters I think and so that messed up the estimated smaller models because embedding parameters are a decent size of the model. A bit on RL. So I feel like a year ago a lot of people were making this argument that RL will not be super successful at scaling for models. I think John you wrote a research paper where you were pointing out that models learn one bit per episode.

1:18:21when you RL basically learned, did I get the answer right or did I get it wrong? And I wrote some blog posts earlier this year, I was like, it's even worse than that because when the pass rate is low and the model is very unlikely to get the answer right, it learns almost nothing at all from an RL episode. But I look at the models today and they seem pretty smart and it seems to be the result of scaling up RL. Baron, you had a post I think a few weeks ago where you're trying to explain what's going on. But why has RL been more successful than one would have naively thought?

1:18:50I mean, so I think the success of RL comes down to a bunch of different things. So first, I think what is slightly underestimated is actually the mid-training. So an awful lot of what we see as successes of RL actually comes from very, very good mid-training data, which is basically where we're essentially doing pre-training, but on synthetic reasoning data and the kind of environments that get the model warm started for RL. So this actually takes the model almost 80 percent of the way to the final RL checkpoint often.

1:19:15And then what RL does on top of that is it does a lot of essentially tweaking to the policy. And so this is one of the reasons why it doesn't need as many bits as you would naively think. It doesn't have to learn all of these behaviors from scratch. It needs just a few bits from these episodes, which you do get. And then the other thing that I really point out in my blog is that these bits are actually extremely high signal compared to regular pre-training, which is why you need RL at all versus just SFTing on successful reasoning traces.

1:19:42bits about how to get the answer right. Well, there's two things. So yes, one, it's exactly the bits about how to get the answer right. But like, this is not exactly how you think of it because In SFT, you have a trace, right? You have like a bunch of math reasoning and then the answer at the end. The bit is still there, like you still SFT on the answer token. So that bit is still there. What's important is that the objective ignores all the other bits. So in SFT, you like have like, you know, to try and match like the exact reasoning tokens of the model produces. So you're essentially getting like too many bits about like the exact way this other model you're training on reasons. For RL, you only get the one bit. And that means that like this, that signal is not drowned out in the noise of like all the other bits the model has. And so that's what really like

1:20:20It's really super dramatic, increasing to the signal to noise ratio during training, which is why RL is so dramatically efficient in terms of steps. I don't know if you guys have thoughts on that. Yeah, there's been so much debate about what RL does to the model versus mid-training or SOT or whatever. Everyone talks about how Passive 1 will go up, but Passive 256 will go down. Very rare, correct reasoning traces will be down-weighted and outweighed by a gradient signal from easier.

1:20:49kind of reasoning traces and I think the simple way to view RL now is that if you have a large enough like a large enough amount of compute to sample a large enough group size such that your probability of getting a bunch of correct answers is like pass some like not insignificant probability, then like it will be up weighted and like to Barron's point like.

1:21:08basically mid-training and you know more pre-training like the the pass-up one the starting point for RL like scales in the log number of pre-training tokens. Can I ask a very basic questions? I guess I that answer makes sense and maybe there's empirical researchers shows that this is what's happening but then I just look at the models themselves and I don't know what's happened.

1:21:33Yeah, maybe you can give me a sense of what is the basis of the AI progress over the last year. But yeah, maybe it's just up waiting the policies which we're going to do the correct thinking anyways. But it just seems like qualitatively, the models have gotten so much more capable. And anyways, maybe there's nothing to, there's no inherent contradiction there. But how do we square like the relatively.

1:22:00small impact this take would imply that RL would have from the actual qualitative capabilities. the models seem to be gaining. One thing I want to point out here is that it doesn't necessarily imply that RL has a small effect. Even if you have a few bits and you only change the parameter to a small amount, the actual impact on function space, the model and the input to upper mapping can still be super dramatic. Even one bit can change your function space a lot and it can work out half the hypothesis space, which is huge. I don't think it's necessary in the case of small amount of bits, small amount of RL. Once you're starting from a really good point, means that you don't have dramatic impacts in behavior.

1:22:33At least not necessarily. I think it comes down to two things. I think the first thing is that everyone was hoping that RL would generalize this reasoning across all these different domains. And I don't think we necessarily got this horizontal generalization. Just training on math doesn't necessarily make you the greatest coder. You do have to do RL on code environments. I think what we did get though is horizon generalization. The models just learned how to use more tokens for longer and still make progress on some sort of task.

1:23:02And so you can train on environments where they get longer, longer, longer, and then put them into a completely new environment. And yes, they may not have generalized the reasoning patterns which allowed them to do well in that environment. But they've at least generalized the ability to continue on that task for longer, which is correlated with success. I think there's a paper called Edgebench, which showed that the rate at which models can work for longer is doubling every three months. And so that's a clear evidence of generalization. And I think the final way to think about it is in pre-training, there's this idea of

1:23:31So you have this very smooth like pre-training loss curve and when you actually look at what's happening in the model like the model is learning all these like very discreet like tasks and there's like always like emergent points. There's like kind of a phase transition like it didn't have induction heads now it has induction heads and there's like tens of thousands millions probably like hundreds of millions of these things and you average them all together and you get this very smooth loss curve.

1:23:53I think to an extent, a similar thing is happening for RL. There is this very slow outer loop, as Beren mentioned, of we will train a model and then RL and then the next model iteration of training, we will dump a bunch of these synthetic reasoning traces into mid-training data. We're hitting all these quanta for all these different tasks. On an individual task level, it may look like a phase transition and you're suddenly going from like a 0.5% pass rate to a 90% pass rate on that particular like finance task or Excel task or whatever. But you average all these things together and plus the horizon generalization, you kind of would go, wow, we've got like qualitatively.

1:24:29better models. Yeah. I mean, I think a lot of this as well is just like, I think RL does generalize a bit like, suddenly you get like some transfer between like math and code or like puzzles and math and this kind of stuff. Also just like the show amount of environments I think the people are targeting is just like vastly greater. So like, you know, before when you tried to do, you know, some tasks, which like you do in your daily life, like two years ago, like the labs wouldn't really care about this. They wouldn't like train the model for it. And now like it's just so much broader. They have a lot of environments targeting this specific thing. Early in the conversation, we're talking about RL in the context of causing this

1:24:58entropy collapse or just concentrating probability on solutions the base model were already done and causing relatively sparse updates in the policy. But when I think about, I think there's also another story about RL, which is going back to the Atari games and then AlphaGo coming up with Move 37, the super creative move that because it was never initialized on human data, it can think in ways that humans are not even thinking and come up with extremely creative solutions.

1:25:29Yeah, do you have a sense on when we should expect or if we should expect RL on LLMs to result in things like move 37, just extreme creativity, even beyond human creativity because there's just de novo, de novo initialization of intelligence? I mean, so a couple of things here, like first off, I think that the AlphaGo is using MCTS, which obviously does like more exploration and like stuff than regular policy gradients. But I kind of also think that like RL doesn't necessarily reduce the creativity. I think this is obviously qualitative, but if we look at the open air hugging phase incident, these models were coming up with multiple zero days at a time to break out of the sandbox. This is clearly some level of move 37 creativity, I think already, which we just get from the general generalization properties of the LMs. I don't think it's definitely not the case that RL is totally destroying the entropy, especially on long horizons.

1:26:23Yeah, I mean one thing that people call creativity is just solving hard search problems. So that's like move 37 is obviously an example of that or like writing some kind of poem that satisfies a ton of different constraints. So that's something AI is obviously going to be extremely good at if trained for it. Then there's another way in which the models like The diversity of their outputs is a lot lower after RL, and they sort of develop these ticks. And even though the models seem like they're good at writing, when you do some kind of distributional analysis, you find that they're reusing certain themes.

1:27:08like all the time and they're using the same character names all the time. So there's actually, it's not like you're getting the same kind of diversity that you get when you, like from human authors, you're sort of getting one really good like style. So I think that, like that kind of diversity has definitely been like cut down by RL a lot. And in fact, yeah, since we were talking about.

1:27:32distillation earlier, that's sort of something, yeah, one thing that's happening is that so many people are distilling mostly from Claude that like all the open-weight models write the same way as Claude and use the same, like have the same ticks. So this seems kind of concerning to me that we're having this like this monoculture emerge.

1:27:54Again, I don't think this is like fundamental to RL is like a method though and same with distillation like even with distillation Like you're just training on the data. It's like just because your data is not like superboard That doesn't mean like the training method itself is somehow wrong. It's like a problem with the data and I think a lot of for instance like The RL entropy collapses basically due to exploitation of fairly simple verifiers when you don't have a huge diversity of environments. Because for instance, the writing is presumably graded by some judge and the judge has some specific ticks and the model is learning to award hack the judge and that's why it collapses. But this is really a problem with the judge. It's not a problem with RL in general. Super rapid fire predictions about the future. So I want timelines on the following couple questions.

1:28:39By when do we have models which you can, here's what the, it feels like to a user. You basically hire them as a drop-in remote worker for all kinds of white collar work, not just coding, but I don't know, video editing, law, paralegal, et cetera. Like it's like literally an actual remote worker with like full computer use with like literally a month of seamless learning.

1:29:06and operation and executing on complex projects and requiring interacting with other people, et cetera, et cetera. So everything a human worker could do over a month. If you like mandated to use like a browser or whatever rather than like these, again, the firm setting up the information to be like programmatically accessible like maybe a couple of years, but if it's not like browser based, like you can send Slack messages, they can do all this stuff, but I'd still probably say around a year. Yeah, I mean, I would say maybe like for the like full generality, maybe like three years, but I think

1:29:36To try this point, we will end up with a lot of people making their organizations easier for the AIS to use. And so you get like 80, 90% of the way there before that. Sorry, but the thing that's, the diff between one year and three years, there was just literally like. I think there's going to be like a long tail of miscellaneous stuff which some human can do, which will take the models quite a while to do. Are you thinking of sort of computer stuff or basic cognitive capabilities?

1:29:59I mean I think this really comes down to a question of like how quickly can we solve this kind of like online learning and like whether we can like get like 80-90% of the way there with like compaction and like writing files to yourself and stuff and like that's my big uncertainty. I really don't know. And another like maybe an example something that wouldn't be good at is like you know if I have to like yell at someone to get something at work or like really push someone to get something done like the model isn't just going to do that it's just going to be too nice.

1:30:24I'd say there's a wide variation in quality of human remote workers. So if you try to hire someone off of up work to do a software engineering project, there's going to be a huge variation. It's often quite hard to get them to do a good job or pay attention to all the feedback you're getting. And I would guess that in some cases it'll be worse. The pre-AI version of this was worse than what you can get now from existing AI.

1:30:54So I think it might end up being a little complicated because maybe to some extent we already have this like for some like not so high quality of work but then like then it's obviously like we're not matching human level in certain like higher quality like forms of work. So but I basically agree with Charlie and Baron that maybe Yeah, we'll have some version of this in a year, so it's like okay, and it will be able to do, maybe we'll have that form factor, and it'll be able to do some things really well, some things not so well, and things will be improving from there. Like we ship the goalpost based on the very long tail all the time. Like, I think, I feel like you've used an example before of like doing your taxes or something. Like this year I literally just like told Codex to like go get everything I needed to do and send it to the accountant.

1:31:45And there was this massive list of stuff it had to use computers to click through and download some stuff. And it didn't. It was fine. It was perfect. So I don't know. A lot of this stuff, it can already do. Yeah. OK. Give you 10x total productivity uplift. Basically, if it takes you a year to make a breakthrough now, you make a breakthrough every month. I think I would just refuse to give you a scaler on this.

1:32:10Like, we might already be past that in some, like, types of work. Like, let's say you're just trying to prove, you're trying to do, like, certain types of math. Oh, sorry, sorry. But for you as AI researchers, trying to make, like, advance, you know, the state of AI research? Because how much are, like, AI researchers sped up? Yeah. We're uplifted. Somewhere between five and ten years? Oh, really? Okay, that's far away. Well, do you think it's longer than, like, four-general room at Huaca? Yeah. Interesting.

1:32:39I think you're right. I think I'm realizing you probably have very different definitions of fully general remote worker. I could have specified that. Yeah, this is true because I mean like yeah because obviously like an AI researcher can be a remote worker and so like. Yeah, I'm picturing like you know normal white collar work over the period of a month. Yeah, I think it starts to diverge a little bit past a month. A very competent white collar worker, but not necessarily like a super creative researcher. Yeah, I would say like two years. Two years? Yeah. 10x? Okay. How are you burned?

1:33:06I can kind of see that actually because like it really is just like right now it's already like definitely more 10x of like coding stuff and so it's like if it can do it even like one or two loops of like experimental feedback that would actually be massive already. So 10x uplift of AI researchers within two years if you just plug it into like a very naive model of like AI progress and how much is coming from AI researchers and they're like there's like a 10x increase in their productivity.

1:33:31Yeah, you have like radically accelerated space of AI progress starting two years from now. Yeah, I mean I think like this will mean that AI progress doesn't get bottlenecked on like AI research ability to run like small experiments that gets bottlenecked on other things. Of course, of course, but it just like happens 10x faster. For sure, yeah. It's a huge deal. And that also like helps the next thing which makes you 100x speed up, happen sooner, etc. Yeah, I'm happy to stick with longer on that one. And what's like the crux?

1:33:58my capacity to absorb information and make the Bayesian optimal decision on the next experiment. Makes sense. Yeah, I mean, I'm assuming that you can delegate some of this to the AI. So the AI is becoming decent at deciding, it's run this experiment, it's got this result, it runs the next experiment. And then if it can run two or three experiments in a row without crashing, then that is actually a big update in uplift. And okay, final question. An AI which dominates top human experts across every single field.

1:34:27of could work that can be done over a computer. So not only a research, but all cognitive work. And not just like short horizon work, but like literally if it takes like three years or something, they also will do better than humans. This is basically just like ASI, okay. I would say like three or four years. The fuck? I mean, that doesn't seem wrong. I mean, I would say like, like AI is obviously being more, getting more attention. So it's like one of the harder things.

1:34:58but it's like a lot of energy is being put into it and it's also not one of the hardest things for AI because it involves a lot of code and math which models are really good at. Maybe for things that involve 3D and spatial stuff and physical stuff, I think that will take a little longer. Especially if it's like mechanical engineering or something and it's not getting like the most attention right now that might take a little longer. But it also does include fields where there is relatively little data because of the nature of the field and it has to like learn that data on the fly. So for example, it has to become superhuman at like being an engineer at TSMC or something. Oh yeah, so you would have to assume that like the onboarding, yeah you can give the AI the same onboarding material and

1:35:50Oh yeah, then there's some like something has to be solved about like sort of longer horizon learning or yeah. I'd say five to ten. It means that yeah, you think automating AI research is like ASI complete or something. Yeah, I think so. Yeah, I think there's so many things in the world which like even if you have some sort of memory system external to the model and even if like context length grows a little bit like.

1:36:17There are just fundamentally things like even if you could research the information or write notes to yourself, like you'd need more than a matter of the context field today. Yeah. Yeah. I mean, I kind of agree in like the five-year range, at least for like the stuff that like labs are focusing on. But I think like there's going to be a long tail of stuff, which like, yeah, I could theoretically go out and learn about, but like no one has bothered to do it. And like the computer hasn't been allocated to that. So that might take longer for like literally every single human expert. And sorry, but by this I also included like the ability to learn as fast as a human, a new domain.

1:36:46I mean, I think that's not necessarily necessary, actually, because the AI will have vastly greater experience than any human. Right. Thanks so much for doing this, guys. I feel like this is a great format for getting different experts to disagree and debate and discuss things together. It was very productive. Thank you. Absolutely. Thanks for having us.