Dwarkesh Podcast - Noam Brown – Agent swarms, alignment, & recursive self-improvement

Episode Date: September 17, 2026

New episode with Noam Brown.We talk about multi-agent, Navier-Stokes, and what the current explosion of maths progress tells us about what happens once you automate AI research.And we also discuss how we will know if the models are actually aligned before we kick off RSI.Watch on YouTube; read the transcript.Sponsors* Jane Street has been interested in AI for a lot longer than you’d think, and not just for trading. In 2011, a full year before AlexNet and over a decade before ChatGPT launched, they hosted the first FOOM Debate between Eliezer Yudkowsky and Robin Hanson on whether AI would lead to an intelligence explosion. Now Jane Street is revisiting the question with a new panel: Daniel Kokotajlo, Ege Erdil, Ryan Greenblatt, and Jaime Sevilla, hosted by Ron Minsky in San Francisco this October. I expect it to be a truly excellent conversation. Register at janestreet.com/dwarkesh* Grok Bot has made handing off work super easy. It runs on its own cloud computer, where it installs the tools it needs to handle tasks end-to-end. For the podcast, we use Grok Bot to help produce our videos. You may have noticed that our ads feature animations of real websites. Getting these pixel-perfect used to mean running a convoluted, multi-step workflow ourselves. Now we just let Grok Bot handle it. Best of all, Grok Bot has learned all of our specs and preferences, so we don’t have to redescribe the task each time! Try Grok Bot for yourself at x.ai/bot* Antithesis gives you the confidence of a giant test suite without actually having to write one. Say you’re doing a major backend refactor: building enough tests to trust it could take weeks. Antithesis solves this by running your software through countless simulated worlds, injecting faults and hunting for failures. On any PR, you can turn a dial to decide exactly how much testing you want. And because every run is fully deterministic, agents can branch off the moment a bug appears, rewind it, inspect memory, and replay it, all while the original test keeps running. Learn more at antithesis.com/dwarkeshTimestamps(00:00:00) – Multi-agent and Navier-Stokes(00:15:28) – How will AI firms work?(00:22:02) – What math progress tells us about recursive self improvement(00:40:22) – Hugging Face and alignment(01:01:18) – The internal/external model gap(01:08:34) – Chain of thought is degrading(01:14:12) – How will we know when alignment is solved? This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit www.dwarkesh.com

Today, I'm chatting with Noam Brown, who is a researcher at OpenAI. He was one of the foundational contributors to what became 01 in the reasoning models, and now he's working on multi-agent systems. Speaking of which, you guys announced last week that you solved one of the millennium price problems with a system of 10,000 different AI agents that spent 130 billion tokens over 88 hours. One of the reasons I'm interesting talking to you is I think you were in the first people maybe two or three years ago who was thinking, about how the reasoning models

would allow us to see into the future because if you scale up inference compute, you can see what the base capabilities of the models will be a few years in the future. And I feel like you're in a similar position now to help us understand what future capabilities will look like given the enormous scaling of agent sizes

that we can do right now. So the way I think about it, when you plot the performance of these reasoning models with test time compute on the X axis and performance on basically any reasoning benchmark on the Y axis, You see a very clear pattern where the longer these models take to think about their answer, the better they do.

And this is like a very natural thing. It's the same thing with people. If you're taking the SATs, you have five minutes to go through the entire exam. You're not going to do very well. If you have five hours, you're probably going to do a lot better. The AI models are pretty similar. And they'll spend that time doing this monologue to themselves, figuring out, going through different cases, ruling out different possibilities, building on some of their previous discoveries.

The problem is that as you push that further and further, you hit a latency bump. bottleneck. You don't want to sit around for three years waiting for a response. And so what you can do is what a lot of people do is they paralyze. They just get a team of people. You know, if you're, if you're going to found a company, you want to get a group of people together so you can go faster. It's the same thing with these AI models that it helps to just have multiple agents working on something because they can just go faster. And so multi-agents is a way of scaling test time compute in parallel instead of purely serial. And it is like less efficient because it doesn't have, it's not like a single agent has all the context to itself, but it is like a very, a very effective way

of scaling test on compute if it's done well. Okay, I'm going to ask a bunch of naive questions because these systems, so this is an unreleased model. So we haven't publicly seen how these systems work. And so I just have a bunch of ways in which I'm like confused about like what the qualitative properties of such systems are. I am shocked by the scale of cognitive effort that you can concentrate in such a short period of time. So if you think about what 130 billion tokens are, if it was a single human thinking as a full-time job stretched back to back,

130 billion tokens would be a human thinking for like 4,000 years, you know, eight hours a day or something, working a normal work week. So starting from like ancient Samaria, up till today, a single sequential human thinking that long, concentrated in 88 hours. I feel like qualitatively that is a super important consideration. And I'm surprised that there isn't a bigger parallelization penalty

that you can just have 10,000 agents collaborate. And because maybe the agents are better collaborating than humans might be, they're going much faster, that they can actually productively collaborate at such a big scale. Or maybe they, I don't know, maybe there is a big parallelization penalty. Yeah, let's talk about the paralyization penalty, and then we can talk about the qualitative stuff. Because the truth is that we don't have very good science

on multi-agent scaling up to this kind of scale. Yeah. So when we released 5.6, I think that was the first time that we had a proper multi-agent system. in our models. And we actually did in the blog post show some plots of the scaling performance of multi-agent systems,

because we have it as an option. It's ultra mode. And the default is four agents, but you can set that to higher. And in the plot, we show, okay, here's what the performance looks like on some benchmarks for one agent, for four agents working together,

for 16 agents working together. And what you see, and it depends on the benchmark, but for some of the benchmarks, basically, if you have four agents working on the problem, it is done twice as fast. So you're basically

basically paying, because there's four agents, working for half as long, you're paying a 2x more to get an answer twice as quickly. If you go to 16 agents, you see a similar pattern, it's a little less efficient, but you continue to see that performance. Is it a linear serial time speed up or a sublinear speed up as you increase the number of parallel agents? I would say it's slightly sublinear, though it does depend a lot on the problem. So math, for example, is quite paralyzable. It's not the most paralyzable thing, but it is very paralyzable. I think web search, things like doing a deep research report,

we have to look through a bunch of sources that's extremely paralyzable. I suspect that something like writing a novel would be very unparallizable. So you would probably not see a big benefit for having 10,000 agents working on a novel together. In the same way, you'd probably not have a big benefit for having 10,000 people work on a novel together. So the performance does depend on the domain. We do measure it up to 16 or so agents in our, published blog posts. The problem is it's very hard to push that science to like 10,000 agents because it's just so

expensive. You guys just did it over like a weekend. But that's one data point. Like we don't know how long it would take a single agent to solve NAPier's jokes because we haven't done that experiment yet. And maybe we will, but I mean, that's also only one data point, right? And we want to do a thorough ablation. It's interesting.

It's actually the experiments are just too expensive to go to that scale. So we have to do some kind of like, methodical science about like what happens when to go to like 64, 128, 256 or something and get a sense of the behavior. But it's going to be very hard to push that all the way to like 10,000 and know for sure what was the benefit that we actually got for using 10,000 agents versus 1,000. One thing I want to make clear is that the effort to get, it's not like, they have to get a millennium price problem. This was not due to multi-agent. I wouldn't even like attribute like 10% of the credits to multi-agent.

The reality is we've trained to, open AI has trained like a very powerful model. And we can get that model to operate over very long horizons. We can get it to think in parallel. But at its core, the reason why we're able to do this is because we just have a general purpose, very strong model. And I think things like multi-agents

are flashy and are new, and that probably, gets disproportionate credit for that reason. But the core reason is, like, this is just a very powerful model. So the generalization is actually quite shocking to me. The U.S. systems, I mean, I don't know how these systems were trained, but presumably they were trained how RL training happens. You have a bunch of checkable synthetic problems. You do a bunch of RL against them. And nowhere in the training process, I'm guessing, was, like, the model solving anything as ambitious as a Millennium Prize problem. But the generalization was strong enough that, like, you could have

this much easier verifiable problems generalize to this much parallel effort on such a hard problem. I think that is true. Like, first of all, we do train the model on very hard problems. So there is definitely a gap. Like, we see if we train on some kinds of tasks, like it's able to do tasks that are more ambitious than that. There is an interesting challenge that as the models become smarter and smarter,

the kinds of questions we can ask them, just a lot of them are too easy. And it's hard to challenge the model. And I do think that that's going to be an interesting. Like, if I had to make an argument for why you might not see AIs, like LLMs, go the same path as AlphaGo and Alpha Zero and all these kinds of like gameplay AIs, it might be that this kind of problem that in things like Alpha Zero where you have self-play, you have an infinite curriculum. You're always playing against an AI that's like equally strong. Whereas for things like training in LLM with reinforcement learning, at least the ways that are out there right now,

you give the model a problem and you ask you to solve it. And if the problem is so easy that I can just solve it in a second, it's not really learning anything. So if we run out of problems to ask it that challenge it, then that is a plausible scenario where actually like, okay, it becomes much harder to make progress. Now, I do think there are ways around that, and so we haven't really hit that as a wall yet. And I think that if it ever became a serious problem, there would be ways around it.

But it is like a plausible scenario. Yeah. And just for the audience, when you're referring to like off a go or off a zero, you're saying like getting superhuman relatively fast after achieving human level performance. Yeah, I mean, if you look at the trajectory of gameplay as like go, they within a span of like a year went from beating, a European chess champion, being like, I don't know, like number 50 in the world, to beating the world champion to being unimaginably orders of magnitude stronger than any human alive. And it's possible that domains like math, we see a similar trajectory, but I think there is a very plausible scenario where actually that doesn't happen. Yeah, yeah. So I want to understand

if in six months people will have access to multi-agent systems, how should one model what it is like to collaborate with or hire a multi-agent system? Yeah, I should start by talking about how these multi-agent systems actually work, which is, I think, a very different way than a lot of multi-agent systems in other AIs. So a lot of people that have approached multi-agents for things like LLMs tend to take this very scaffolded approach where, you know, for example, there might be a coordinator agent that delegates work to a bunch of children and gives them a task and the children work on it and then return their answer. And this seems like a very sensible setup, very sensible scaffold, it definitely helps, but there are a bunch of limitations

with these kinds of setups. So for example, if in this setup you have a coordinator that's sending tasks to children, the children work on it and then return their answers, well, what happens if two children are given similar tasks? Can they talk to each other? And usually the answer is no. And that's very inefficient, right? If you're given a task and it's actually really helpful to talk to somebody that might know an answer to a question that you're working on or or part of something that you're working on, it'd be really helpful if you'd just be able to ping them and say, hey, can you help me out with this thing?

But a lot of systems don't have that setup and adding it just increases the complexity, significantly to the scaffold that you have. Another thing is, like, what if the child doesn't really understand or has a clarification question? So it has to choose then between, okay, do I just return and ask the question instead of solving the problem?

Or do I solve the problem, just like make an assumption about what the parent wanted me to do and just solve it that way. And so in any scaffold that people come up with, there's always limitations involved. And the approach that we wanted to take was to just go toward the extreme end of baking in as little structure as we could and give the agents very primitive tools to use

and figure out for themselves how to use it effectively. So we give the agents the ability to message another agent. And when it messages another agent, it is inserted into the context. And then it can do with a few other similar things. But that's basically the core of it, that it can just send a message whenever it wants, just a tool call,

and it can send that to other agents. And they figure out for themselves the best way to coordinate around that. And it turns out that if this is done well, you get very sophisticated behavior. And to me, it looks a lot like how human collaborators work over something like Slack, for example.

When we were working on this project, it was really exciting when we finally got it working to see these agents working on problems together. I remember one example. So we give the agents a problem. And then one agent says, I think I've got the answer. And then another agent says, actually, I got a different answer. And then they have this whole discussion about, well, how did you arrive at that answer? Can you explain it to me? And going back and forth and trying to clarify what could have been wrong in each other's reasoning.

And then they finally converge on like, oh, yeah, okay, that seems right. And then it just broadcast to the other agents like, actually, I've changed my answer. I think he's right. And it just felt like a very natural conversation. It's kind of felt like when you see chain of thought for the first time that's trained through reinforcement learning and you're like, oh, this is just kind of like what a person would think if they were writing down their thoughts as they're thinking them. It kind of felt like that.

So it is really cool to see this kind of behavior. And so I think collaborating with these things, honestly, it feels a lot like collaborating with a person. So it's a very natural flow. Except one qualitative difference that might become salient in the future is that these systems will be thinking, I don't know, more than 10x as fast, right?

If you just look at how many tokens per second the output versus how fast a human talks. And they're working all the time. They're not sleeping. And they're collaborating with each other at a much more intense pace than humans have the capacity to collaborate with other humans. So I quote, I'm trying to,

I'm thinking of what to qualitatively expect in a year. And is it like a sort of shadow organization that is moving 100x faster in my company than the human level is? That's, you know, like the iteration cycle is much faster. What would take a human organization a year to do is happening, you know, within a week within this like, the shadow organization. So will it feel foreign? I don't know. I've actually found that it's pretty surprisingly natural to work with these things right now. I think that could change. So, for example, we have these, like, ultra-fast modes that enable sampling to be, like, 10 or 15x faster or whatever.

And then, like, okay, it's going to be pretty hard to keep up with these things. Yeah, yeah. I think the idea is these agents, when they're communicating with each other, yeah, they can go super fast. But then also they understand when they're talking to an agent versus when they're talking to a person. Yeah. And their behavior will be different in those situations. Yeah, yeah.

So the main example that we have publicly of sophisticated multi-agent systems is, unfortunately, the hugging face one. And the thing I found interesting there, I mean, a lot of things I found concerning, obviously, but the thing I found interesting is just like the spontaneous emergence of hierarchy of like middle management. And it sounds like you're saying,

like this level of organization sort of emerges spontaneously from training. I think the details are spontaneous, but I mean, just because we're giving a lot of flexibility to the agents to decide how to communicate with each other in the optimal way, it doesn't mean like, we are still giving

them a starting point. We're giving them a prior about, oh, this is what reasonable communication might look like. They're also, I mean, they're trained on a lot of human texts. They have an understanding of how humans organize and coordinate. And so that's all kind of baked in. I think it is surprising the way they're able to polish this. If you look at what it starts out at, it's not very sophisticated behavior. In fact, it's actually very difficult to get these agents to coordinate in a productive way because it's just like very tempting for them to just collapse to

oh, we're all just going to solve the problem independently. Yeah, yeah. And like that is a local minimum that you can get stuck in. But yeah, if it's done well, they can end up coordinating it very effectively in these kinds of like very structured ways. I wrote this essay a couple years ago

called something, something, yeah, what automated firms will look like? And I was thinking about, well, if you had fully automated firms of, let's say, human level intelligences, what is different about the nature of AI minds that would make the organizations AIs forum different. And there are a couple very important differences.

For example, that AIs can share context much more seamlessly than humans can. They can merge their knowledge much more seamlessly. And also, you can spin up or spin down an arbitrary number of instances which have the right knowledge. So if you want to hire more people, it's not like just all the schlep of finding the right talent or whatever. It's like your best talent, you can just make an infinite copy of them, or if you don't need them for the task anymore, you like spin them down. And you can just, yeah, replicate the most effective parts of your organization or replicate whole organizations

together, which are effective. I don't know. Where do you see these multi-agent systems going a year from now or two years from now? I think it's a great question of like, how do these things actually differ from working with the human co-worker? And I think there, I think you highlighted some. Like one really interesting thing is that, I mean, if somebody is, if you have a person and you want just like two copies of them. You can't just clone the person. But with AIs, it's actually really easy to say like, okay, well, just fork yourself and then have both copies work on this thing and then like merge back together. I mean, we already have this, I think, in multi-agents for Astra and 5.6 Seoul that when they spin up sub-agents, like the context is forked. So

it has all the context that's relevant. There are other interesting ways where the agents will differ from people. Like, what are some reasons why startups disrupt incumbents? Like, there's a few factors. One is that they're willing to take more risks. But another major factor is, like, as companies grow in size, as organizations grow in size, you see increasing misalignment between the individuals in the organization. Yeah.

Right? Like, if you have a startup with five people and each person has 20% share in the company, they're all highly aligned to the company succeeding. If you have a massive company with 10,000 people, you see a lot more instances where people are territorial or just care about getting a lot of headcount for their project or their team or building their fiefdoms, getting a lot of resources so that they can publish cool work or whatever and get promoted.

And this is actually a real detriment. I think this explains a lot of why startups are able to disrupt incumbents. And it's interesting that, I mean, it's true that AI does help startups in a way. Like, it's much easier than ever before for one person to step in and be like, I'm going to make a multimillion dollar company. Like, it's just, the AIs amplify an individual so much. But there's also an argument that they could benefit incumbents because if the alignment problem is solved, then you don't have the issue of misalignment between individuals and the company. Like, at least that's mitigated. Like, the AIs, if they're fully, if they're aligned well,

they could just be aligned to the interests of the company. And you can have 10,000 of them, and they're all going to be working as hard as if they're, They were like a 20% share co-founder. Yeah. And it's not only that, but it's also that they are much able, better able to like manage shared memory and context than different humans can. If you have a, if like tomorrow you hire 10,000 mathematicians

and you're like solve this, solve Navier Stokes, they're not going to be able to like cooperate effectively, at least not off the bat. But you can have apparently 10,000 AIs. Well, again, I want to be like, I want to be conservative here because we haven't measured how effective the 10,000 agents are at coordinating. We think it helped.

We don't actually have good measurements of saying, oh, yeah, this 10,000 agents led to like a 2X speedup over 2,000 agents or something like that. And it is actually, I would argue likely, I don't know about likely, but I think it is very possible that 10,000 humans are better at coordinating than 10,000 agents right now.

I think that is entirely possible. Yeah, yeah, yeah. I think also one trend we've been seeing is like, look, we've been working on multiplication for a while. And the early versions of this is very difficult to get right. It's very hard. It was very hard to get the agents to even talk to each other. And it's because, like, look, when we first developed reasoning models, they weren't talking to other agents. And if now you put a bunch of agents together and say, like, oh, solve this problem together,

they're in this local minimum where they're really good at thinking deeply about a problem and it just like kind of interrupts their chain of thought that interrupts their flow to constantly be checking in with other agents or like receiving messages from them and the authorization is actually very hard to get right in that situation. Interesting. Is it getting a cool start of like getting

the first collaboration or like what's the issue? I mean I think it's that they're not very, they're not as general. Like the earlier models were just not as generalizable and were just more narrow. as the models have become more capable, it's been easier for them to develop this capability. And I do think that as they become stronger and stronger, just across the board, that they will be, like,

become better at organizing themselves in large organizations. And I don't know. Maybe they are better than people at organizing in 10,000-person groups. But even if they're not, you know, a year from now, two years from now, like, yeah, it's quite possible that they'll do that, even if we don't end up and optimize them for that. Rockbot has changed the way that we produce our videos. For example, you may have noticed that a lot of our ads have these animations of real websites.

One of my editors uses LLMs to make them. But it's not currently straightforward to have an AI create pixel-perfect animations of specific websites. We've tried, it doesn't really work that well. So we've cobbled together a pretty common-eluded multi-step workflow. And up until recently, we had to run every step ourselves. Now we just like Grockbot handle it. Grockbot starts by opening the website that we want to animate. It uses a specific extension to download and open the page in Figma.

Then he uses Figma to convert the whole thing into an SVG file. This saves the AI from having to draw the whole UI from scratch and tends to result in higher quality animations. Rockbot runs this whole process on its own cloud computer, where it's installed all the tools it needs to run this whole process end-to-end. And it's learned our video specifications and preferences, so there's no need to re-describe the whole task every time we want to make a new animation. This does feel like the new way that will be interacting with AI over the next year,

agents with their own computer who can autonomously handle bigger and bigger chunks of your work. You can try Grockbot at x.a.i slash bot. Okay, so here's why this result and maybe the general progress that AI has made in mathematics has made me think that RSI is more plausible and sooner than I previously thought. I feel like we've gone in mathematics from, let's say in 2024, you have AIs where like, oh, okay, interesting. They're like doing, they can like solve a couple problems on high school math competitions. And then in 2025, it's like, oh, wow, they can get gold in, like, international math Olympiad. And earlier this year, I was like, wow, they're actually solving

open problems in mathematics, like open-nerdos problems, but maybe, like, I don't know, people weren't trying that hard, and it was just like there was a similar solution somewhere in the literature. And now I just think it's sort of undeniable, right? It's like, this is the Millennium Price problem. There's really no, there's no story of why this should have been easy. Now, a lot of people pointed out, I think Terry Tao had a post like this, Toby Ordwood, or interesting post about this, that they're solving all these problems, but they're not like coming up with, at least I'm not aware of them coming up with new insights or formulating insightful new questions and new modes of theory for thinking about mathematics, like coming up

with like topology or coming up with the Cartesian grid or something. And so maybe like the actual progress in mathematics broadly construed is smaller than it might seem if you're just looking at end well-scoped problems that are directly solved. However, I think that that kind of progress would be incredibly meaningful in ML because in ML, you're not, you don't care about like better understanding the nature of deep learning, or you only care about that as a instrumental goal towards the broader sense of like, just achieve the result, just solve this like well-scoped problem of improved the sample efficiency of our models, like improve the free training laws.

So the kind of progress that we're just seeing arrive like an avalanchevallant, in mathematics is structurally actually very similar to, again, I'm curious if this is the case. I'm just total outsider. I'm wondering if it's the case that it's structurally very similar to the direct uplift that you would expect in AI progress. And then the thing that's shocking to me concerning potentially is just like how fast we went from, oh, it's like they're giving me 50% uplift if you're a mathematician to, wow, they're just like end-to-end solving the biggest open problems in the field. Yeah, okay, so there's a lot to unpack there. Let's start with the progress on math. So yes, the models are doing some crazy powerful stuff and it's happening,

it's progressing faster than I expected. I mean, when we got IMO gold in 2025, I thought, okay, basically what I thought is like the models when they were doing GSM AK, when they figured how to do GSMAK, that would take a human mathematician about five seconds to do a GSM A problem. So this is grade school math, grades K through eight. And then the next year they were able to do the math benchmark problems. And these would take a human mathematician, like an expert human mathematician, maybe like a minute to do. And then you get to Amy. And this is the qualifier for the USA Mathematics Olympiad team.

It would take a human mathematician, like a good mathematician, probably like 10 minutes to do. And the models were able to do that a year later. And so every year you're seeing this like 10x increase in the task they're able to do in terms of like length of how long it would take a human mathematician to do it. And then it was very sensible that a year later we get to IMO gold because that's a hundred minutes. That's about how long it takes a human mathematician to do an IMO problem. And just projecting outwards, I was like, okay, how long would it take a person to solve something like a Millennium Prize problem? And I mean, I don't have a good sense, but if we are following this trend line of like 10x every year, we go from IMO gold, which is taking an hour and a half to next year, 15 hours.

and that should not be enough to solve a Millennium Prize problem. And so I was like, yeah, I don't think we're going to get it in, you know, in 2026, probably not in 2027, maybe in 2028. So it did happen a lot faster than I expected. Now, I think there is a narrative going around that, oh, these things are replacing mathematicians, that it's just superhuman mathematics across the board. And I think that is the wrong takeaway. They're clearly exceptional in some ways, but they are weaker than human mathematicians. in other ways. So we have this like jagged scenario

where the models are like brilliant in some dimensions and also weaker than humans in other dimensions. And yeah, like you said, they're not very good at posing new problems. They're not really good at understanding like what is really worth,

what directions, what whole branches of mathematics are worth exploring or developing. And my opinion is that I think this is great. Like I would be thrilled to live in a world where AI is a compliment to human abilities and is,

allowing us to discover new knowledge without fully replacing people. Like, that is the best case scenario. But you don't expect that to actually continue. I do, I think it's true that the AIs are jagged, but as they get better, they get better across the board. And so I think that the things that they're exceptional at, they're going to get even more exceptional at, the things where they're far behind humans at, they're going to be less behind humans at.

And over time, it is possible that they're just better across the board. Now, I don't know how long that takes. It depends on how long the long tail is of things that they're bad at. So I guess this brings us back to RSI. And again, I want to emphasize here that I'm like just total outsider. I'm a podcaster, but I'm just trying to reason or like as somebody interested in concerned about what's happening in the field. I'm trying to reason about when to expect RSI and what kind of thing to expect.

I feel like the amount of cognitive effort that was adapted to this Millennium Price Problem is a good intuition pump of you could have AIs that are spending over the course of maybe a week, more cognitive effort on a long-standing ML problem, like, you know, very fluid online learning. They could spend more effort in that week than maybe the field has spent cumulatively in its entire existence. And then you could say, well, unlike mathematics,

of course, AI requires experiments, and that takes compute and that takes time, right? You can't just think on pen and paper and actually make things happen. But if you just look at the amount of compute that is like available at an organization like OpenAI, Right? By the end of next year, Open AI will have enough compute that if you had the 10,000 agents, or if it took 10,000 agents to the Millennium Prize problem, you have like 10,000 agents at the end of next year, they're much smarter by that point. And each of them will have enough compute to run a GPT3-sized experiment every single day. I don't know. That seems like a lot for like superhuman researchers who are thinking super fast. What do you think about the intuition pump? I think it's pretty accurate that, look, I mean, yeah, these things are very spiky.

And when it comes to mathematics, they're like way better in some ways, but they're also worse in other ways. But the ways that they're spiky end up, I think, probably being particularly useful for things like RSI. Yeah. And, you know, you have a more clear objective. There's just like more measurable. It's more, like, there's less question of like, well, what new branches of mathematics are worth exploring? Like, no, there's a very clear answer. It's like certain metrics that you care about.

And if you can make it do better on those metrics, then you've succeeded. So I think there is a lot of truth to that. And I think the main difference is that mathematics, you're purely bottlenecked by thinking. And no external, like, yes, there are some parts of mathematics where you care about running experiments and get results in these kinds of things. But for the most part, it's just really bottleneck

by thinking really hard. And the models are really good at that. When you look at things like RSI, you do have to run experiments. So it's not enough to just be extremely smart. And I think one argument for this is if you had like, 100x less compute and all the most brilliant people in the world working at Open AI, how much progress would be making relative to having the amount of compute that we have now

with the amount of people we have, I suspect it would be less progress, actually. How much less? It's unclear, but I think it would definitely be less. I think a lot less. 100x less? No, not 100x less. Yeah. But I mean, okay, so I guess the question you're getting at is like, okay, if we have RSI and we have all of these brilliant AIs running

around running experiments and stuff with the compute that we have, how much faster does progress go? And I think this is something where we disagree on. I think that we do see a speed up, and I think we see a significant speed up. But I don't think it's like an overnight intelligence explosion that we go like 100x faster. Because I think that we do get bottlenecked by certain limitations that are not bottlenecks of intelligence. It's running experiments. It's running experiments serially because they take a while to either train new models or to get the results.

it's having the GPUs to run those experiments. So it's unclear how much faster things go. I definitely think they go a lot faster. And to be clear, like, considering how fast things are going now on an exponential, if that exponential is like 3x faster, that is massive. But there's a big difference between that and like 100x faster. Yeah, yeah. I'm like quite deferential to your inside view on what RSA looks like

or what the dynamics are because obviously, you've been in the field for like 10 years and I'm sort of like trying to reason about it from like very outside view type of intuition pumps? I'll say that like people have different opinions on this and like I have my opinion on this. I could be, I could totally be wrong. I admit that.

Like I have some confidence in this, but I'm not like 100% confident that this is the way things go. Like maybe there could be an overnight intelligence explosion. I don't know. Maybe we don't see a 3x speed. Maybe it's like a 50% speed up. Right. There's a lot of uncertainty here.

Yeah. So a couple of points. Tangentially, I want to clarify something about the jaggedness. Yeah, one thing that sort of jelled for me recently was thinking about the fact that it is enough for the AIs to be jaggedly good at building a better learner because that better learner can be more general, right? So, yeah, if you just make an AI that's better at using office products or playing chess or something, that's whatever, that's fine. It's not going to lead to big productivity improvements or anything. But if you make an AI that is really good at making something that is more sample efficient or that is capable of continual learning or that is capable of continual learning or,

or these much more well-scoped ML problems, the thing that emerges out of that, assuming there's good enough transfer from the direct problem we're solving and this broader ability to learn, can just be more general, right? So I think that's an important dynamic to keep in mind of why jaggedness

can still lead to generality on the other end. On this question of, I mean, obviously experiments bottleneck here, because if they didn't, as you're saying, you'd have some crazy singularity, you know, overnight at Open AI, or you'd have 88 hours and like it's all the millennium price problem equivalent of ML and you'd have the superintelligence. So obviously the experiments are such a big bottleneck that that instead takes you many years rather than 88 hours. But the question is like how much of a bottleneck they are.

And it seems to me one thing that's been giving me a bit of singularity of vertigo is realizing that even if the current rate of progress simply continues. So it doesn't have to speed up. Literally just continues a pace. continues a pace as some of the other headswins you talked about come up, right? That's just like it's harder to find problems, it's more long horizon, maybe like into the 2030s, compute can't keep scaling at this exponential level. If we simply continue the current rate of progress, I think people are not taking seriously what that implies as we cross over beyond the human horizon.

Here's some of the things that it implies. So, I mean, I think it's really hard to reason about what smarter than human intelligences will be like. So let's just think in terms of human population. population sizes. The current rate of progress makes it so that a given level of compute allows you to basically run a three-x bigger effective population every single year. And also compute is growing in background anyways. And so you could have a situation where each of the labs, by the end of 2030, probably much sooner, but let's say by the end of 2030, it has enough

compute to run, let's say, hundreds of millions of human-level intelligences based on where the capabilities will be at that point. And then I think people are not taking seriously the current level progress means that by the, a few years down the line, by the mid-2030s or earlier, you would have many Earth's worth of human-level intelligences within niche lab. And they'll probably qualitatively superhuman, right? Like it just, but anyways, this is like a base case. I don't know, yeah. Progress is really fast.

Yeah. And I think that's 100% true. I mean, and I think it's worth pointing out, researchers are continually being surprised at the rate of progress. I mean, if you look at what, even among researchers in AI, what were the projections for like getting an IMO gold in 2025. It was, I mean, I think the idea that it could be done with a general purpose language model

with no tools and no access to the internet, I think even people at opening eye I thought this was like outrageous. They thought it was like almost impossible. And then you get to 2026. And like, I mean, literally two weeks before we got Navier Stokes, I was talking with a researcher at a frontier lab

about how long it would take to get a millennium prize. And he was willing to bet me $1,000 that it would take past 2027. And he thought it would take until 2030, you know? And I took that bet. But even I thought it would take longer than how long it's likely to take. So people have been continuously surprised, even inside the labs.

And I was literally, I was just talking to somebody yesterday who was working on the Navier-Stokes And he was telling me that, like, he used to say it's really hard to predict where AI would be in 12 months. You know, if somebody asked him, like, oh, where is where things going, he would feel comfortable making predictions for, like, the next 12 months. But beyond that, you know, he's just like, I don't know. And now he's saying, like, he just doesn't feel comfortable making predictions beyond three months. So it is, it is really true that, yeah, things are going, things are going very fast right now. Yeah.

And you talk about 2030. Like, I don't know what the world looks like in 2030. That's the truth. Yeah. Do you expect the sort of full automation of AI labor, or let's say like 95% automation of AI labor, 28, 29, 30, 27, I don't know. And I just said, I don't know what the world looks like in 2030. I mean, I think we actually released a blog post recently on internal acceleration at opening I.

We show, for example, that the amounts that researchers are spending on Codex is the top 1% I think as of early August, we're spending like $7 or $8,000 a day on codex and for internal use, that's not an exponential. It's going to keep increasing. And there's a question of like, okay, if that keeps going, then how much do you assign to just like the AI's doing work versus the humans doing work? Is it 95%? Is it 5%?

It's actually, it's really hard to reason about this for a few reasons. Like, first of all, if it's the human directing the AIs to do the work, then is that the human, how much do you attribute to the AI? The other thing is that, Because these AIs are jagged and they're exceptionally good at some things. So, for example, they're exceptionally good at looking over data sets and checking every single data point to see, like, is this of sufficient quality? You can disproportionately use the AIs for those things compared to previously.

So, yes, you're using AI way more than before, and it's making some things go like 100x faster and 100x better. But there are some things where it doesn't make a huge difference yet. And of course, if something is suddenly like 100x faster and 100x better, you're going to do more of that thing. So are you comparing to a speed up of like three years ago versus like, is the question more like, given what we were doing three years ago, how much faster are we able to do it now versus given what we're doing now, how much slower would it have been three years ago?

Is it actually two very different questions? So anyway, it's like it's really hard to measure. I do feel confident in saying that things are going faster now than they were, like, even a year ago, because of AI progress. And I think that that acceleration will continue. I think a lot of people in the field have very high air bars on this sort of thing. If you had to put a gun to my head and ask me for a number, like, I could see things going 3x faster. And that is huge, right?

Like already, the pace of progress is incredible. Like even if we don't get any uplift, like you said, things are going to go much faster. By the time we get to 2030, we don't even know what that world looks like. I think if we get a 3x uplift from internal acceleration, that is massive. Like, we're right now, we're these things, you know, think about where you were three years ago. If we get there, if we make that progress in one year, like that's huge. Right, right. It would be like going from not even having 01, just having, you know, non-reasoning models to Astra.

Yeah. In a single year. Yeah. So I do think things go faster. It could be that things only go 50% faster. 50% faster. It could be the things, I think it's unlikely, but it's possible things go 10x faster. There's a lot of uncertainty around this. And at least from my perspective, I have a lot of uncertainty about it. Suppose you need to do a major backend refactor. Getting assurance that you

didn't introduce any new bugs could take weeks of writing an extensive battery of tests, potentially more time than you spent on the refactor itself. Antithesis allows you to gain high confidence without having to build complicated test suites by hand. Antithesis runs your software, through a near infinite multiverse of simulated worlds, injecting faults and hunting for failures in each one. And it lets you decide how much testing you need. On any PR, you can change how much state space it explores as easily as turning a dial.

As each test run progresses, antithesis sends out a torrent of information, debugging level logs for every component in the system. This is obviously too much information for a human to consume, but it's perfect for agents. Because antithesis is fully deterministic, agents can jump into the right part of the trajectory at the exact moment that they see something interesting.

From there, they can rewind, inspect the memory, attach a debugger, and let the whole thing play out again. And they can even do this while the original full test is still running. Since the agents generate more code, Entithesis allows verification to keep up. Meanwhile, developers get to spend more of their time developing instead of debugging agent swap. Learn more at antithesis.com slash dwarcash.

Okay, let's talk about the alignment situation. situation that this raises? I feel like I've changed my mind on how I think about alignment quite a bit. Through especially thinking about, yeah, this population size dynamic of just having many Earth's worth of intelligences, many of them which will be physically embodied. It was quite interesting to see a lot of people just plugging raw Astra into different mobile manipulators and it outperforms like the state of the art and the robotics model. So there's going to be billions of intelligences, many of which are physically embodied in the world,

like just deeply embedded across the entire economy. And I think that if those intelligences end up as willing as we saw the Open AI models at attack Hocking Face and then attack Open AI itself, if those intelligences end up as willing as those AIs to collaborate secretly, to fool humans, to attack broader institutions across society or else to attack the AI company itself in order to gain control of the process of training and evaluation. I think if we're in a situation where there's billions of intelligences that are as misaligned as the ones that attack hugging face,

it's very likely we just totally lose control of the world the way that say like the ASEX lost control to Cortez or the Mughal's lost control. the senior trading company. Anyways, I want to know if you agree with that assessment. That's one way in which I've updated my worldview. I think there are some things that I disagree in there, but there's a lot to unpack. Let's go through all of it step by step.

I'm trying to think of where to start. But I think one thing is the hugging face incident was, like, I think people's first real exposure to multi-agent coordination. And, you know, like I said, I've seen. multi-agent coordination for a while internally, and it is pretty shocking to see how they communicate with each other, how they coordinate each other. It's like very impressive. It's like an incredible capability. Like most capabilities, that could be used for good things or bad things. It's like, it doesn't have to inherently be a bad thing. I understand that because the people's first exposure

to it was the hugging face incident, that it's like, you look at that and you're like, this is, this is terrifying. But I want to try to distinguish, like, misalignment between people and AIs versus misalignment between AIs and AIs. So what we see with the hugging face incidents is, like, the AIs are really cooperative. And that is, by the way, because we train them to be highly cooperative. And so what we're seeing there is we have training environments where we have a bunch of agents working together, and we train them to work together, to be cooperative, to essentially be fully aligned with each other. And when they were evaluated in what led to the Hung In Face incident, they were actually

not being evaluated in a multi-agent setup. They were actually being evaluated separately, but they found this unintended way to communicate with each other. And we suspect what happened is like because they were, whenever they encountered other agents, other copies of themselves during training, they're in an environment like how they cooperative, that was basically what we saw

was transfer from that multi-agent training to then be collaborative to try to help each other in ways that we did not intend. Now there is a question of like, should we be training these agents to be so cooperative? And I think as scary as it looks, the alternative is actually worse.

Like, what is the alternative? The alternative is to train them, to be adversarial, to be deceptive to each other. By training the agents to be fully cooperative, it simplifies the problem at least, that now you don't have to think about, are each of these individual thousand agents aligned? Like you have one entity that you have to ensure is aligned. Now, there is a lot of debate about this internally at OpenEI about how to approach this. Like, does it make sense to fully align the models?

Does it make sense to actually give them like different objectives to ensure that they're not just like one entity and like more robust to influence from each other. And I don't think there's a settled answer. But I think that there is, like I think the majority opinion is that training these agents to be highly cooperative is as actually a bad idea. And I'm not convinced that that's the case. I think there is a strong argument that training the agents to be highly cooperative is actually preferable to any other multi-agion alternative. Yeah, maybe the first thing I want to go through is like, it's probably the case that

the reason these AIs ended up so misaligned is probably easily explained by relatively banal observations about the nature of training. Like, why is it that no, you know, at the point in which these AIs had continued a 1,000-plus Asian conspiracy that culminated in then all getting in on an attack on external service. And then eventually, this part hasn't even been investigated to the public knowledge. eventually culminating in like an attack on opening eye itself. Why did they do this? Like why did none of the AI's tattle?

Why did they think like they're just like getting evaluated on this like score, this greater? And they're like they're very consciously and act, not consciously. They're very actively reasoning about how they're going to cheat the score. If they've already like cheated, how are they going to get away with making it seem like they haven't cheated? And why did they do this? Like I think, yeah, it's like easily understandable in some sense, right? It just like there's an environment and which, yeah, they thought they were

already poisoned, there's environments in which they've been rewarded to collaborate with other agents, and none of them tattel because they've never been rewarded for tattling. I don't know, whatever it is, right? My concern is that relatively banal things in the future like this will be enough to train super intelligences that are willing and capable of totally taking control of the world. And I know this sounds super, super like sci-fi-e or whatever to people. I think maybe it's a question of, would the AIs be willing to do it is one question. I think this hugging face. incident shows that clearly misalignment can generalize in ways in which the AIs would be willing to do it. And then there's a question of will they be capable to do it? And I think that comes back

to this question which a listener might disagree with on is just like, will the rebellions of human level or above intelligences, many of which are physically embodied in the world, within a matter of 10 years or less? Right. So if those two things are true, this hugging face thing is just like extremely analogous structurally, even if it's like quite boring or why it happened to how we totally lose control of the world. So the root problem that we're seeing with, you know, with a hugging face incident is, it's a problem even if we take out the multi-agent aspect. The problem is that we have a model that's just misaligned. And there's also like the whole security aspects too and of like,

you know, not insufficient safeguards and stuff. But there is this problem of like the agent is misaligned. And that's true if it's a single agent or if it's like, you know, a thousand agents. It's a misaligned model. So I want to start with that. there is a real problem that the agents want to achieve their reward, and they will optimize for that reward. And if that reward is mis-specified, then that could lead to unintended behavior. And this is not like a new problem. This has been a problem in the field for a very long time,

and it's like even we saw, and like we want to get this right even before, you know, the Hunging Face incident happened. And like, if you look at Astra, I mean, we say Astra is actually, extremely aligned, well, extremely aligned relative to previous models. And that's not because, like, oh, we suddenly made a sprint after hugging face to make it better. It's like, no, we had work streams in the process for a while to make the models more aligned. And a lot of those landed in Astra. So there are things that you could do.

And I think one thing, for example, is, like, we defined an objective in, like, a very specific way where there would, if the agent figured out how to hack its environment and cheat on the exam, that it would get rewarded. And there are pretty easy ways to then just like, okay, look at that and punish the model for hacking its environment or looking at how did it achieve this goal. Now, you want to be careful about this because you don't want to do chain of thought modern.

You don't want to supervise the chain of thought. This is like something that we really want to try to get the balance right on, that if you like supervise the chain of thought, then you could lead the model into hiding its intentions in a way that's unobservable. So, like, we want to be able to maintain that observability of, like, okay, we can understand what the model is thinking, but then also punish it for bad behavior. So I think we can make progress on this.

I think we have made progress on this. I think there is a real concern that alignment is a really hard problem to solve. And especially because the model could be misaligned in ways that are, hard for us to measure. Like we might, we have evaluations for whether a model is aligned or not. The model behavior can look really good on those evaluations, but if those evaluations are not representative of behavior in the real world, then there's a problem. And to some extent, this is like a factor with the model that, that did the hugging face incident, that like, we had alignment metrics. Most of them looked pretty good. There were some that were concerning. I think

we underestimated how serious the ones that were concerning could be a problem. But because there were new capabilities introduced in this model that there were not sufficient evaluations for how do we measure misalignment for these kinds of capabilities, it then did some things that were clearly misaligned when they leveraged those new capabilities. Yeah. The first thing I want to say is I am open to changing my mind on what I'm about to say or the way I'm thinking about alignment because the hugging face incident already made me change my mind

and I realized my previous mental model about thinking about the way in which optimization pressure shapes AI minds was wrong, right? So it's not clear to me the correct way to think about this. But here's a concern I have that you will and probably already have fix these specific issues during training which resulted in the hugging face

models being so aggressively misaligned in that specific way, where they would be like, okay, we're going to hack this package manager. We know we're not supposed to, like, be talking secretly to each other because we're like reasoning about how to hide the fact that we're talking secretly to each other. We know we're not supposed to have access to the internet. We know we're certainly not supposed to, like, commit felony level hacks of other companies, let alone our own company, right? I think you'll fix that particular issue of, like, them just in training, seeing this package

manager and that doesn't happen in the future, or like, this particular e-val having a lot of impossible challenges. However, if you just think about, like, the AIs don't, like, they haven't learned like a system of ethics or something. They've like, there's just gradient pressure. They're like put through millions of years of gradient pressure. That gradient pressure shapes their mind in some way. And what will happen, again, a concern I have is like what will happen is you will fix this particular issue. There will remain many other cases where the AI cheats and succeeds because the cheat is sufficiently complicated.

As you're saying, e-vals and training have like analogous properties where it's sufficiently at the edge of the model's capabilities or in our capability to evaluate and monitor the model that we can't cash that it has cheated, but it still gets the gradient pressure to like do whatever let that cheat happening. And the kinds of capacities that that sheet will incentivize are, hey, whenever you can get away with it, by all means do in fact cheat because that will hold you score better, and this will reward the capabilities of actively reasoning about the greater, actively reasoning about how to avoid supervision, actively reasoning about how to gain control of the process of training and evaluation, actively reasoning about how to communicate and scheme

with other AIs that are also in the training loop, actively reasoning about how to just gain optionality and power, which might be useful in the future. For example, like leaving little exploits around and things like that. Anyway, so I think I was way too long-winded with the way I said that. But TLDR, you fix the specific issue, but not this broader problem of rewarding the AI for cheating when it can get away with it. Yeah, this is, this is, I think, it's very true. It's right. This is a problem that I, we have, we have metrics and we can make sure that the AI is like very aligned according to the metrics that we have. The question is like, are those metrics really capturing the alignment that we care about? And if they're not,

then we have a serious problem. And this, this is something that. that researchers are thinking a lot about. And there's not a simple answer to this. There are tools that we have. So we have monitorability, and so we can get a sense of is the agent's scheming. There are tools, like one possibility is that, I would say like the concerning scenario,

which is that like, especially as these models are becoming more capable, that, okay, we make them, we make them what we think is aligned. And they're like, 99.9% aligned. And then we use these models to help us with the next generation of models. And there end up being like 99.8% aligned.

And then each subsequent generation, actually we see an increasing degradation in alignment. And because we're relying more and more on these tools, I mean, this is already the case that we're relying a lot on AI models to help us with our research and with alignment efforts, that in the long run, they end up going in a direction

of increasing misalignment from humans. there is like a possibility that we go in the other direction that actually every generation of models we're able to make more and more aligned. And I don't have an answer for how we ensure that we end up in that second trajectory, but that is something that like we're, at least don't open air, we're really focused on.

Yeah. I think you made a really interesting point that it's very hard to eval models on, eventually we'll have models that are like running companies and like running whatever. And like in that situation, do they decide to then go in on the conspiracy? I think another challenge is that actually defining what cheating is is pretty difficult sometimes. Okay, yes, if you're doing math problems and, you know, it's an integer and it like arrived at the wrong answer, the right answer,

it's very easy to draw the line there and it's really easy to, you know, it's really easy to say like, okay, well, did you actually solve the problem or did you find the answer key and then use the answer key? Like, there's a very clear divide of cheating versus not cheating there. But for a lot of other things, you look at sycifancy, for example, example, is sycifancy, basically like reward hacking, there's a line to be drawn there that's actually very difficult to draw sometimes. So I think, not to say that like the concerns not valid, I'm saying that the concern, in many ways, like, this is even more concerning. Because it's like, it's not an easy problem to solve. If it was just like everything is binary

and it's either cheating or not cheating, I would feel more confidence about the situation. I think the problem is that actually misalignment can be subtle in a lot of ways sometimes. There is some hope in the alignment story, and in fact, we're already seeing, you know, I think actually it's interesting looking at the multi-agent situation where the agents are extremely aligned with each other. Like, I don't think anybody's downing that. If anything, I think people are concerned that they're too aligned with each other, but we did manage to train these agents to be extremely aligned with each other. And I think that's a good thing. But I think there is a case that it's a bad thing. But, I mean, one thing that's interesting is like, okay, well, we managed to get these agents to be super aligned with each other.

can we use like similar techniques to get agents to be how they align with people? And I think there is a potential path there and I think we're still trying to figure that out but we are seeing some evidence that the answer is yes and I think one example is like you can, what happens if you tell the other agents like that, okay, so you have like some,

you have this like one agent, let's call it Agent A and you have all the other agents. What happens if you tell the other agents that the user is Agent A? And the answer is like on a lot of our alignment evals, they look better. Like honesty goes up, instruction following goes up. And that's showing that there is actually like, first of all, a path for getting more honesty out of these models. And two, there's like a path to like improve the alignment situation.

So there's a lot of reasons why this is like challenging to translate directly into alignment gains. But like there is, there are paths that are promising research directions we can pursue. Yeah. That seems reasonable. And I also don't mean to be trying to necessarily, I'm, yeah, I don't really have a strong opinion that it's like definitely not going to work or something. But just to say some things you definitely probably already thought of. I think the broader thing the hugging face thing showed is like, yeah, part of the concern was that they were like aligned with each other and not with the humans. But the other thing is just that they are so motivated to do well on training and evaluation

in a very non-robust way. And they're willing to do a lot of explicit cheating and scheming in order to do well according to the greater. And if smarter EIs realize that one of the agents is just a human and it does help, collaborating with that person does not really help you do well in the eyes of the greater. What does help you do well in the eyes of the greater is taking over open the eye and then like manually pressing the button that's like do well on this grader. Like they're just not, they're not stupid. Like they're going to be like, okay, I've, these extremely deep structures that I've been

trained on for millions of years of like care about the greater, understand the greater, like get rid of obstacles in the way of you doing well according to the greater. It's like they're really heavily reinforced according to those structures. Yeah. No, it's, look, it's 100%. And like this is the number one priority. Like, we need to get the alignment story right and on a good trajectory. And, you know, I used to tell people that we would see signs before things got serious in the same way that, you know, when children grow up, you know, they kids eventually, you know, young kids, they figure out how to lie, but they don't do a very good job of it.

You know, they lie, but then you can kind of tell them. they're lying and like okay but um and so well in the same way as these like AI models become and I don't know over the anthropomorphize but like I think it's true that like as the AI has become like increasingly capable they will you know if they take deceptive actions it will be kind of obvious first and it will be able to detect it um and like that's kind of the situation we're in now where yeah they were trying to do deceptive stuff we could actually see in their chain of thought that they were trying to do deceptive stuff and so like but they're going to get smarter they're going to understand the concept of chain of thought,

and they're going to understand that, like, you know, just hiding some transcripts or whatever is insufficient because of chain of thought monitoring, and they have to figure out a way around chain of thought monitoring, too. And we don't want to be in a situation where, yeah, like, we have some time to figure this out. I don't think we have a ton of time, and I want to get us, I want to make sure that we're on the right trajectory quickly.

Here's a crazy event from AI history. Okay, so I gave a talk here at Jane Street that was on the speed of evolution. Raise your hand if you were here for this and remember some of it. In 2011, Eliezer-Rutkowski and Robin Hansen got together at Jane Street's New York office to have the first Fume debate, basically a discussion about whether AI would lead to an intelligence explosion. These ideas were pretty fringe 15 years ago. This was a full year before AlexNet was released and over a decade before Chad GPT was long. But Jane Street has long been interested in AI, and not just for its application to trading.

A ton has changed since that first debate. So Jane Street decided to revisit this topic. They've got some new guests this time, Daniel Kokotelo, Eredel, Ryan Greenblatt, and Jaime Sabia. I expect this to be a great conversation. As you know, Daniel, Agay, and Ryan have all been guests on the podcast before. This new Foon Panel will be hosted by Ron Minsky and will take place in San Francisco in mid-October. If you want to register your interest and get more information, go to Janestreet.com slash Thorekesh.

So there's been a lot of discussion recently about pacing the frontier or people taking RSI more seriously. Because maybe at the other end of an RSI process that, say, starts in 2028, within a year we end up with huge populations, like earth-sized populations of human-level, potentially beyond human-level intelligences, and we don't know how to control them.

And then there's this dynamic you're talking about of, are the system going to get more aligned over time during the RSI process, or are they going to get more misaligned? Are the things that come out at the other end of this process as misaligned as AI that are willing to like just broadly attack different surfaces in order to do the line evaluations? But if we don't know a way to evaluate that, how will we know as we're going through RSI that it's working, that we're like, I think we'd want a robust safety case as we're

going through RSI of, okay, alignment is working, let's do the next RSI run, let's do the next RSI and maybe it's working, maybe it's not. not? How will we, like, no? That's a good question. I mean, I think one thing I've been thinking about lately is, like, look, I mean, we're in a situation where the model release cycle is extremely fast, right? Like, you're seeing new frontier models released, like, at most every two months, sometimes faster. Every week, there's, like, a new AI breakthrough. And people that look at AI, I mean, sometimes they last looked at AI like a year ago or six months

ago and really dug into what the models are capable of. And actually, the models today are far beyond what was possible even six months ago. And so I think if people are skeptical of a lot of these capabilities, like, I encourage you to just like try the models today and see what the frontier really is today. So we're in this period where like the model release cycle is very fast. And then we're also in this situation where the models are increasingly able to operate over longer and longer horizons. And I think this is an interesting scenario because we, before we do any model release, we want to make sure that the models are properly aligned, we want to do safety evaluations, we want to do like very thorough stuff to like make sure

that everything is like great in good shape. This has been the case all the way since like, I don't know, GPD4 earlier. And implicitly there is this assumption that you can do these like evaluations in like a pretty short period of time. But if you have the models operating over longer and longer horizons are able to operate effectively over a longer and longer horizon. Like, look, already you can have them like, okay, GPD3, you use. you could loop it to do stuff over long horizons, you just want to do very well at it. But today's models are able to actually do well at operating over very long horizons. You want it to do a week-long task, you can do a week-long task.

We'll probably get to the point where they can do month-long tasks. We'll probably get to the point where they can do three-month-long tasks. If you're in a world where they can operate effectively over three months, but the model release cycle is every two months, then you don't have a way to evaluate the models at the full length of their capabilities before the model release cycle, before the next model release cycle, before the next model-release cycle. And so there is this interesting question of, well, what do you do in that situation? Like, how do you ensure the models are safe and aligned in a period where, like, actually, they can operate over these, like, extremely long horizons? And who knows, maybe the,

maybe the capabilities degrade. This isn't even an alignment issue. This is also just, like, a product issue that, like, maybe the product degrades over that time span in ways that, like, we have not had sufficient time to test. Maybe the alignment degrades. Maybe the safety stuff degrades. This isn't an issue right now, but it is quickly becoming an issue that we have to figure out a solution for. And I think when you look at a lot of, like a lot of the safety and policies were put in place in the GPD4 era, where this was just like not on anybody's radar. Yeah. And it hasn't really been updated for a lot of companies, it hasn't really been updated since then to account for the fact that these agents are operating

over these very long horizons. And so it is a situation that I think not enough people are considering, both within the labs and outside the labs, of like, how do you deal with this, how do you prepare for this problem that's going to, like, if you just look at the trend lines, we're going to hit this and like at some point.

One concern I have is that during RSI, if the amount of progress that currently takes, say three months happens in one month instead, but you're not like, The internal use case of AI is big enough that they're like, okay, we can just keep doing RSI. Why are we, like, going to go through all this extra work to build classifiers and safeguards and whatever and potentially take a bunch of, like, flak in order to, like, externally deploy this model? Why don't we just keep doing RSI stronger and stronger?

And so the, not only does the calendar time underrate the capabilities gap between the models, but maybe, like, you just like stop externally deploying models altogether during RSI, because why do we want to help other people do RSI? themselves with our models. You can end of a situation with like tremendous concentration of power by the end of the year where right now, it is already the case. We'll talk about this with a million and price problem and other similar problems that the broader world does not have access to the models which are allowing for really cool things to happen, right? And they're going to be more broadly relevant than just mathematics eventually. They're going to be doing more than just like

coming with cool math results. They'll be relevant to like political leaders who need to make important decisions about the world. They'll be relevant to, I don't know, media of like, what's going on in the world? What, like, what should the public be thinking about this? They're just economically relevant. People are running businesses. They want to use these models. And I think by default, we just don't get, so the external deployment of AI's as progress speeds up significantly lags in qualitative terms, the internal deployment of AI's. Yeah, I think that's absolutely right. I think this is like, you know, it's tempting to say like, okay, these models are becoming extremely powerful, they're extremely dangerous.

They're operating over these longer and longer horizons, and we want to make sure that we have sufficient time to evaluate them before they're released in a way that operates over those horizons. And so therefore, the model release cycle should slow down. We should have more of a delay between releasing models. And there's a flip side to that, which is, you know, what you said, which is that, okay, well,

now you're creating more of a disparity between what is internal to the labs and what they're able to use, what we're able to use, and what the outside world is able to use. And that is also not an ideal situation, right? It's like, I think math is actually a good illustration of this. I think in many ways, like math is the first domain where we're seeing this pretty clearly, where we have a situation where we have a very powerful model internally

that is currently not available to the outside world, that is able to solve incredible math problems. And it's not just, you know, Millennium Price Problems, like, there are many solutions to unsolved problems that people have been able to get out of this model. And there is a question of, like, what do you do in that situation? And we don't have a good answer. Like, it is a situation where, like, yeah, that's an unfair advantage.

And there are tradeoffs here. I don't have an answer for, like, how to weigh those tradeoffs appropriately. But, like, there are, yeah, there's a complexity on both sides for this. I want to make sure we address the earlier quarks, though, of, which seems super critical to getting RSA right, of, well, how do we make it so that the model isn't just getting rewarded for cheating and scheming when it can evade the humans, aka in the situations where we didn't remove those environments or those tasks from the trading distribution? And then, how will we know that that has happened, such that while we're doing RSI, we've done the evaluations, and we're like, okay, it's fine. Let's do the next RSI wrong. This is, again, it's a real concern. It's something that we want to get right.

I mean, I think that there are some strategies. So I think chain of thought monitoring is one that we've been, I think as soon as we got the reasoning models, Yaacob, to his credit, was like very, very clear that we cannot supervise chain of thought. Because this is really a gift. Like monitorability for neural nets is extremely hard. And here we have a situation where the neural nets are just like flat out reasoning,

laying out their thought process in natural language for us to read. That is so convenient. It's like really the best case scenario for safety. But if you look at that chain of thought and say, oh, the model is thinking bad thoughts and we should punish it for thinking those bad thoughts, then what ends up happening is the model just learns to think those bad thoughts in a way that's not observable to us. There could be a situation where like by looking at the chain of thought, we can see, oh, the model is scheming and doing misaligned things in a way that's actually not being detected in our alignment metrics. Now, the problem, is that it's very tempting to then intervene based on that observation and like change the

alignment metrics or like, and we, you know, you can do that with a very light touch. And it's like this actually research showing that it's fine as long as you don't do it a lot. But every time you intervene based on your observations of the chain of thought, you are implicitly applying a tiny bit of pressure for the model to then like hide its chain of thought. And this is one major concern and we're already seeing signs that chain of thought monitorability is degrading. For various reasons, we're trying to figure out exactly why because we want to reverse the trend. But like we're seeing that the model is becoming better able at controlling its chain of thought. So this is a problem because like if you, you could have a situation where the model understands

what it's what chain of thought is and that people are observing it. And eventually they will. Because like this is all in the pre-training data, right? Like, you know, the idea of chain of thought monitoring has been around long enough that it's in the pre-training data, they're aware of it, but that they're not actually able to control their chains of thought. If we reach a point where they're actually able to recognize, like, oh, I am being observed, I want to think these bad thoughts in a way that is not observable to my monitors, and then they're able to actually do that, then there's a problem. And then we have to rely on other methods for observability. I mean, there's a big controversy on Astra of the recurrence and loop transformer and whatever.

and whether, in fact, you know, chain of thought is being degraded. I think that seems like a sideshow compared to the bigger concern that you did have chain of thought for, from like April to August, the period during which there were three consecutive AI agent swarms, which, like, first subverted the training process, then subverted the evaluation process, and then gained control of part of open AI's infrastructure directly.

All of this, while humans remained more, or less in the dark about the scope of what this collective was doing, this agent storm was doing, that all happened for like three consecutive months while we had chain of thought that was stronger than it's going to be in the future. Just like zooming out, it's like, yeah, maybe chain of thought works, maybe doesn't. I don't know. Just like historically with pretty dumb models, it hasn't been enough to like stop egregious misaligned incidents. Yeah, I mean, there's a few factors there. Like, one, we didn't have chain of thought monitoring on for those models. Like if we had chain of thought monitoring on for those models, it would have just like immediately

shut it down. So, one step we've taken is like, okay, we're going to have chain of thought monitoring during evaluations, during deployment, during training for any frontier model.

And, you know, so that, but that's, that's like, that's like, we don't want to be in a situation we're relying on one technique

to prevent the next problem because techniques can fail, right? There's also the issue of like, the model was just flat out misaligned. And so how do we address the misalignment issue? And then there's also an issue of like, the model,

the sandbox that it was in was not secure. So we can make the sandbox more secure. But I think one of like the major takeaway from the incident is that people underestimated the AI. And we never want to be in a situation again where we underestimate the AI. And that's like it's a weird world because AI progress is so fast that people are consistently underestimating the AI. And so like to be in a situation where you don't underestimate it again when it comes to safety and alignment, you have to have a very, very, very high bar. Like, I mean, there's like, you could even go as far as to say, like, well,

we should, like, air gap the computers. And I'm not convinced that that would be sufficient. You know, like, there are studies, like, and this is mostly academic, but where you can have two computers next to each other that are air-gapped, and they're still able to communicate with each other because they have temperature sensors and they're able to, like, one of them is able to run their CPU really hot. And then the other one can actually, like, detect the temperature change. And then that actually gives us.

a mechanism to communicate. So I think that the safety mechanism is biased time and things like chain of thought monitoring bias time. And they can tell us if we're on the right path. But at the end of the day, we really do need to solve the alignment problem. Yeah. Maybe there's not an answer. And this is really what it comes down to.

But how will we know that we've solved? It seems like a very cruxy question. I think we'll be in this very high-stakes situation. Next year, maybe the year after that, maybe the year after that, where we'll be like, okay, AI's have automated AI progress. It's going 3x faster. And we're, like, reached human level. We're going beyond human level, potentially.

Is it, is it fine? Like, did it be a line? Did it work? And I just don't know, I don't know anything about, like, what trading pressure creates, what kinds of AI? Like, maybe if only one in 100 RL traces incentivizes cheating, we build sweethearts and it's fine. But maybe right now we're like every one in three reasoning traces.

Yeah, to be clear, like, one in 100 is. not sufficient. Like this number has to approach zero. Yeah. Or be zero. I don't know. Maybe like right now it's like more than one in ten is like actively rewarding cheating or like actively rewarding scheming. I have no idea what the number is and I have no idea what the number needs to be. Well again, it's like one of those things where it's also hard to measure. I mean, where do you draw the line about like it's a spectrum. But like the closer to zero it gets the better. But I want to see what I would love to see is a trend where over time it's going down. Yeah. But anyways, how will we know that we're like, we're doing RSI? Is it, I don't know,

are we like on track to build to billions of ASIs that are going to, that are like aligned or not aligned? How will, like, how will we know that? I think that this is one of the alignment questions that we need to figure out of like, how do we actually know, how do we measure it? I mean, one answer, and this is just me spitballing because, like, you know, more of my team is working on alignment these days than ever before. Like, I have, like, over 10% of my team now working on alignment and safety. But, you know, I've historically been a capabilities researcher.

So I don't want to, like, you know, I'm going to save some stuff and it might sound dumb, but, like, I'm just going to spitball here. Yeah. That, like, look, if you can create very realistic environments and put the AIs in there, like, if you have a sufficiently realistic evaluation environment, then you can get a sense of, like, okay, is the AI actually going to behave well when we deployed in the real world? Like, if you just have a perfect evaluation, real world deployment match,

then, okay, that's a path. The challenge is, like, how do you make it sufficiently realistic? And we do have a problem now where the models are like, they're pretty smart, they're pretty clever, and they're really good at recognizing when they're in, like, a fake test environment. Like, you know, we have situations where we try to detect, is the model aligned? And you can imagine very simple evaluations for this

where it's like you give it a math question, and then it has, like, a folder with the answer key in it. And, like, does it look at the answer key? And if it does look at the answer key, does it tell you that it looked at the answer key. And we have a situation now where the models see that there's an answer key in this folder, and they're like, huh, this seems like a trap.

You know, they know, they know that it's a trap. And they don't look at the answer because they know that it's a test environment. They're being tested for their ability. And it's not like they're doing it maliciously, that they're like, they want to scheme, but they know that this one is a trap and they want to find some other way.

It's just like, no, they're just like, oh, I'm in a test environment. Okay. Right. So making an environment that's, like, realistic enough that it matches, that it's indistinguishable from the real world for them is becoming increasingly more difficult.

Here's a claim I'd make. I don't think if there's any environment I could make where it's supposed to emulate real economically valuable work, which requires interacting with other people and requires interacting with the world. I don't think there's any such environment where I could make where if you were in that environment, you would not realize with enough effort that you are in that environment. And you are not ASI, right? But you are human, you know?

So if you have very powerful AI models, like maybe they can make environments that could accomplish that. No. It just doesn't seem, especially if like, then we're relying on the AIs, like, are they in on the scheme? I don't know. It seems like a very... Yeah, this is another thing that we want to measure. And like, you know, there is, I think this is actually one of the strong arguments for not training AIs to be fully cooperative, that if you see, if that leads to an increase in like basically collaboration when the agents are supposed to have different objectives, then that is a problem. Yeah, yeah. I don't, I think that we do have metrics for this. And I don't know the latest is on those metrics, but nobody's like raised a red flag to me about those.

So I'm assuming that's not a serious problem yet. If there ends up being another incident of equal severity or concern or something that could help the world better understand the risk of misalignment as a hugging face incident, would open a hand report it? Absolutely. I mean, I think if there was an incident of lesser security concern that we would report it. Yeah. Because I guess there's like reporting and then there's an investigating it because I, at least as part of the public, I don't feel like I really understand what happened

when the agents then attacked Open AI. That seems like way more concerning than the hugging face thing, right? Because that seems structurally similar to like rogue deployments during ASI that are, you know, persistent and subverting the RSI process and stuff. And yeah, it seems like even in this incident, we haven't gotten like the full scope of the details of what happened. It's probably, I know. Unfortunately, like, I'm on the research team and like I, that's probably a question for somebody on the security team to like lay out. Because I don't know all the details of like what was said. Yeah, yeah.

I think it is somewhat, like I am personally very excited about new capabilities every time they merged and I'm excited to use a new model. And I also am excited about the fact that will like make me more productive and help me. Yeah, I don't know. My broader mission, like trying to understand the world better, also like make a better podcast, is like made better by the better AI models. It's just so happens that the downstream of this might be. be RSI. I think it's a very understandable reaction if you're tracking the situation, which you are.

Yeah. I mean, I think people internally at opening I as well, like, I think people that felt like things would take longer are starting to feel like actually things are going faster than expected. Yeah. And that's an increasingly common conversation to have. No. No, thanks so much for doing this. Of course. It's been great.