Main Topic: The podcast episode discusses the current state of AI research, specifically focusing on the advancements in large language models (LLMs) and their potential applications in various domains. The conversation revolves around the capabilities and limitations of LLMs, their potential impact on society, and the challenges associated with aligning AI with human values.
Key Points:
Participants:
Takeaways:
Recommendations:
New episode with my good friends Sholto Douglas & Trenton Bricken. Sholto focuses on scaling RL and Trenton researches mechanistic interpretability, both at Anthropic.
We talk through what’s changed in the last year of AI research; the new RL regime and how far it can scale; how to trace a model’s thoughts; and how countries, workers, and students should prepare for AGI.
See you next year for v3. Here’s last year’s episode, btw. Enjoy!
Watch on YouTube; listen on Apple Podcasts or Spotify.
----------
SPONSORS
* WorkOS ensures that AI companies like OpenAI and Anthropic don't have to spend engineering time building enterprise features like access controls or SSO. It’s not that they don't need these features; it's just that WorkOS gives them battle-tested APIs that they can use for auth, provisioning, and more. Start building today at workos.com.
* Scale is building the infrastructure for safer, smarter AI. Scale’s Data Foundry gives major AI labs access to high-quality data to fuel post-training, while their public leaderboards help assess model capabilities. They also just released Scale Evaluation, a new tool that diagnoses model limitations. If you’re an AI researcher or engineer, learn how Scale can help you push the frontier at scale.com/dwarkesh.
* Lighthouse is THE fastest immigration solution for the technology industry. They specialize in expert visas like the O-1A and EB-1A, and they’ve already helped companies like Cursor, Notion, and Replit navigate U.S. immigration. Explore which visa is right for you at lighthousehq.com/ref/Dwarkesh.
To sponsor a future episode, visit dwarkesh.com/advertise.
----------
TIMESTAMPS
(00:00:00) – How far can RL scale?
(00:16:27) – Is continual learning a key bottleneck?
(00:31:59) – Model self-awareness
(00:50:32) – Taste and slop
(01:00:51) – How soon to fully autonomous agents?
(01:15:17) – Neuralese
(01:18:55) – Inference compute will bottleneck AGI
(01:23:01) – DeepSeek algorithmic improvements
(01:37:42) – Why are LLMs ‘baby AGI’ but not AlphaZero?
(01:45:38) – Mech interp
(01:56:15) – How countries should prepare for AGI
(02:10:26) – Automating white collar work
(02:15:35) – Advice for students
Dwarkesh Patel
Okay, I'm joined again by my friends, Sholto Bricken, wait, fuck.
Trenton Bricken
Did I do this last year? No, no, no, you named us differently, but we didn't have Sholto Bricken and Trenton Douglas.
Dwarkesh Patel
Sholto Douglas and Trenton Bricken, who are now both at Anthropic. Yeah, let's go. Sholto Scaling RL, Trenton's still working on mechanistic interoperability. Welcome back. Welcome back. Happy to be here. Yeah, it's fun. What's changed since last year? We talked basically this month in 2024. Yep. Now we're in 2025. What's happened?
Sholto Douglas
OK. So I think the biggest thing that's changed is RL and language models has finally worked. And this is manifested in we finally have proof of an algorithm that can give us expert human reliability and performance given the right feedback loop. And so I think this is only really being conclusively demonstrated in competitive programming and math, basically. And so if you think of these two axes, one is the intellectual complexity of the task and the other is the time horizon of which the task is being completed on. And I think we have proof that we can reach the peaks of intellectual complexity along many dimensions. But we haven't yet demonstrated like long-running agentic performance. And you're seeing like the first stumbling steps of that now and should see much more conclusive evidence of that basically by the end of the year with like real software engineering agents doing real work. And I think, Trenton, you're like, experimenting with this at the moment.
Trenton Bricken
Yeah, absolutely. I mean, the most public example people could go to today is Claude plays Pokemon. Right. And seeing it struggle in a way that's like kind of painful to watch, but each model generation gets further through the game. And it seems more like a limitation of it being able to use a memory system than anything else. Yeah.
Dwarkesh Patel
I wish we had recorded predictions last year. We definitely should this year. Oh, yeah. Hold us accountable. That's right. Would you have said that agents would be only this powerful as of last year? I think this is roughly on track for what I expected with software engineering.
Sholto Douglas
I think I expected them to be a little bit better at computer use. Yeah. But I understand all the reasons for why that is. And I think that's like well on track to be solved. It's just like a sort of temporary lapse. Yeah. And holding me accountable for like my predictions next year, like I really do think end of this year, sort of like this time next year, we have software engineering agents that can do close to a day's worth of work for like a junior engineer or like a couple of hours of like quite competent independent work.
Trenton Bricken
Yeah, that seems right to me. I think the distribution is pretty wonky though. Yes. Where like for some tasks, I don't know, like boilerplate website code, these sorts of things. It can bang it out and save you a whole day. Yeah, exactly.
Dwarkesh Patel
Yeah, I think that's right. I think last year you said that the thing that was holding them back was the extra nines of reliability. Mm-hmm. I don't know if that's the way you would still describe the way in which these software agents aren't able to do a full day of work but are able to help you out with a couple minutes. Is it the extra nines that's really stopping you or is it something else? Yeah.
Sholto Douglas
I think my description there was – I think like in retrospect, probably not what's limiting them. I think what we're seeing now is closer to lack of context. Yeah. lack of ability to do complex, very multi-file changes and maybe scope of the change or scope of the task in some respects. They can cope with high intellectual complexity in a focused context with a very scoped problem. But when something's a bit more amorphous or requires a lot of discovery and iteration with the environment, this kind of stuff, they struggle more. And so maybe the way I would define it now is the thing that's holding them back is if you can give it a good feedback loop for the thing that you want it to do, then it's good. It's pretty good at it. If you can't, then they struggle a bit.
Dwarkesh Patel
And then for the audience, can you say more about what you mean by this feedback loop if they're not aware of what's happening in RL and so forth?
Sholto Douglas
Yes. So the big thing that really worked over the last year is Maybe like broadly the domain is called like RL from verifiable rewards or something like this where a clean reward signal. So, you know, the initial unhobbling of language models was RL from human feedback where, you know, typically it was something like pairwise feedback or something like this and the outputs of the models became closer and closer to things that humans wanted. But this doesn't necessarily improve their performance at any like... like difficulty of problem domain, right? Particularly as humans are actually quite bad judges of what a better answer is.
Humans have things like length biases and so forth. So you need a signal of whether the model was correct in its output that is... that is like quite true, let's say. And so things like the correct answer to a math problem or unit tests, passing, this kind of stuff. These are the examples of reward signal that's very clean, but even these can be hacked, by the way. Like even unit tests, the models find ways around it to like, hack in particular values and hard code values of unit tests if they can figure out what the actual test is doing. If they can look at the cached Python files and find what the actual test is, they'll try and hack their way around it.
So these aren't perfect, but they're much closer. And why has it gotten so much better at software engineering than everything else? In part because software engineering is very verifiable, like it's a domain which just naturally lends it to this way.
Trenton Bricken
I think... Does the code pass the test? Does it even run? Does it compile?
Sholto Douglas
Yeah, does it compile? Does it pass the test? You can go on the code and you can run tests and you know whether or not you got the right answer. But there isn't the same kind of thing for writing a great essay. That requires... The question of taste in that regard is quite hard. We discussed the other night at dinner, the Pulitzer Prize, Which would come first, like a Pulitzer Prize winning novel or a Nobel Prize or something like this? And I actually think a Nobel Prize is more likely than a Pulitzer Prize winning novel in some respects. Because a lot of the tasks required in winning a Nobel Prize, or at least strongly assisting in helping to win a Nobel Prize, have more layers of verifiability built up. So I expect them to accelerate the process of doing Nobel Prize winning work More initially than that of like writing Pulitzer Prize-worthy novels.
Trenton Bricken
Yeah. I think if we rewind 14 months to when we recorded last time, the nines of reliability was right to me. Like we didn't have Claude Code. Yeah. We didn't have deep research. All we did was use agents in a chatbot format. Right.
Sholto Douglas
Copy, paste, copy, paste, copy.
Trenton Bricken
Totally. And I think we're very used to chat interfaces, whether we're texting or using Google. And it's weird to think that the agent can actually go and fetch its own context and store its own fax into its memory system. And I still think that it's the nines of reliability. And if you scaffold the model correctly or prompt it, it can do much more sophisticated things than the average user assumes. And so like one of my friends, Sam Rodriguez, who does Future House, they've discovered a new drug that they're in the process of patenting. And by the time this episode comes out, that will be live. What was that? LSD V2? Wait, is it really? No, they're not making LSD. But, like, people didn't think that models can be creative or do new science. Right. And it does just kind of seem like a skill issue.
Dwarkesh Patel
I mean, there was the cool— Wait, wait, but like, it discovered a drug. Is it—
Trenton Bricken
How did it – like it one-shotted the molecules? So this was just over a conversation, and so we'll need to refer to the full announcement. But my impression is that it was able to read a huge amount of medical literature and brainstorm and make new connections and then propose wet lab experiments that the humans did. And then through iteration on that, they verified that this like new compound does this thing that's really exciting. Yeah. Another critique I've heard is, like, LLMs can't write creative long-form books. And I'm aware of at least two individuals who probably want to remain anonymous who have used LLMs to write long-form books.
And I think in both cases, they're just very good at scaffolding and prompting the model. I mean, even with the viral ChatGPT GeoGuessr capabilities, where it's just insanely good at spotting, like, what beach you were on from a photo. Kelsey Piper, who I think made this viral... Their prompt is so sophisticated. It's really long and it encourages you to think of five different hypotheses and assign probabilities to them and reason through the different aspects of the image that matter. And I haven't A-B tested it, but I think unless you really encourage the model to be this thoughtful, you wouldn't get the level of performance that you see with that ability.
So you're bringing up ways in which people have constrained what the model is
Dwarkesh Patel
outputting to get the good part of the distribution. But one of the critiques I've heard of RL or the – not of RL, but one of the critiques I've heard about – using the success of models like O3 to suggest that we're, like, getting new capabilities from these reasoning models is that all these capabilities were already baked in the pre-training model. I think there was a paper from Sting Schwab University where they showed that if you give a base model enough tries – to answer a question, it can still answer the question as well as the reasoning model. Basically, it just has a lower probability of answering. So you're narrowing down the possibilities that the model explores when it's answering a question. Are we actually eliciting new capabilities with this RL training, or are we just putting the blinders on them?
Sholto Douglas
Right, like carving away the marbles on this. I think it's worth noting that that paper was, I'm pretty sure, on the Lama and Quen models. And I'm not sure how much RL compute they used, but I don't think it was anywhere comparable to the amount of compute that was used in the base models. And so I think the amount of compute that you use in training is a decent proxy for the amount of actual raw new knowledge or capabilities you're adding to a model. So my prior, at least, if you look at all of DeepMind's research from RL before, RL was able to teach these Go and chess-playing agents new knowledge that was in excess of human-level performance just from RL signal, provided the RL signal was sufficiently clean. So there's nothing structurally limiting about the algorithm here that prevents it from imbuing the neural net with new knowledge. It's just a matter of expending enough compute and having the right algorithm, basically.
Dwarkesh Patel
Why aren't you already spending more compute on this? I think Dario said in his blog post that labs, or it was like a couple months ago on the expert controls thing is like, ah, deep seek, whatever. We're only spending 1 million on RL or something. So it's like, we aren't in the compute limited regime for RL yet, but we will be soon. Yeah. You're spending hundreds of millions on the base model. Why only order a million on the RL? You know the parable about when you choose to launch a space mission and how you
Sholto Douglas
should acquire, go further up the tech tree because if you launch later on, your ship will go faster and this kind of stuff? I think it's quite similar to that. You want to be sure that you've algorithmically got the right thing. And then when you bet and you do the large compute spend on the run, then it'll actually pay off. It'll have the right compute efficiencies and this kind of stuff. And I think like RL is slightly different to pre-training in this regard, where RL can be a more iterative thing, like you're progressively adding capabilities to the base model. Pre-training has, you know, in many respects, like if you're halfway through a run and you've messed it up, then like you've really like messed it up.
But I think that's like the main reason why, is people are still figuring out exactly what they want to do. I mean, 01 to 03, right, like OpenAI put in their blog post that it was a 10x compute multiplier over 01. So clearly they bet on one level of compute and they were like, okay, this seems good. Let's actually release it. Let's get it out there. And then they spent the next few months increasing the amount of compute that they expend on that. And I expect, as everyone is, everyone else is scaling up RL right now. So I basically don't expect that to be true for a long time.
Trenton Bricken
Yeah, just for the sake of listeners maybe, you're doing gradient descent steps in both pre-training and reinforcement learning. It's just the signal's different. Typically in reinforcement learning, your reward is sparser. So you take multiple turns. It's like, did you win the chess game or not is the only signal you're getting. And often you can't compute gradients through discrete actions. And so you end up losing a lot of gradient signal. And so you can presume that pre-training is more efficient, but there's no reason why you couldn't learn new abilities in reinforcement learning. In fact, you could replace the whole next token prediction task in pre-training with some weird RL variant of it and then do all of your learning with RL.
Yeah, at the end of the day, just signal and then correcting to it. Totally. And then going back to the paper you mentioned, aside from the caveats that Schulte brings up, which I think is the first order most important, I think zeroing in on the probability space of meaningful actions comes back to the nines of reliability. And classically, if you give monkeys a typewriter, eventually they'll write Shakespeare, right? And so the action space for any of these real-world tasks that we care about is so large that that you really do care about getting the model to zero in on doing the reasonable things.
Yeah.
Dwarkesh Patel
And to the extent, in some broad sense, to the extent that at some pass a key, you've got token space. Right, exactly.
Sholto Douglas
You literally do have a monkey, and it's making Shakespeare in the end. Yeah, yeah, exactly.
Dwarkesh Patel
Okay, so the alpha, the chess analogy is interesting. So were you ready to say something? I was just going to say, like, you do need to be able to get reward sometimes in order to learn.
Sholto Douglas
And that's, like, the complexity in some respects. In, like, the alpha variance, or maybe you were about to say this, like, one player always wins. So you always get a reward signal one way or the other. But in the kinds of things we're talking about, you need to actually succeed at your task sometimes. So language models, luckily, have this, like, wonderful prior over the tasks that we care about. Yeah. So if you look at all the old papers from like 2017, it's not that old, but like, you know, like the papers from 2017, the reward, the learning curves always look like flat, flat, flat, flat, flat as they're like figuring out sort of like basic mechanics of the world.
And then there's this like spike up as they learn to exploit like easy rewards. And then it like, it's sort of like, it's almost like a sigmoid in some respects. And then like sort of continues on indefinitely as it like just learns to like absolutely maximize the game. And I think the LLM curves look a bit different in there isn't that dead zone at the beginning. Because they already know how to solve some of the basic tasks. And so you get this initial spike. And that's what people are talking about when they're like, oh, you can learn from one example. That one example is just teaching you to pull out the backtracking and formatting your answer correctly and this kind of stuff that lets you get some reward initially at tasks, conditional in your pre-training knowledge.
And then the rest probably is you learning more and more complex stuff.
Dwarkesh Patel
Yeah, yeah. And it would also be interesting. I know people have critiqued or been skeptical of RLs alluring quick wins by pointing out that AlphaGo took a lot of compute, especially for a system trained in, what was it,
Sholto Douglas
2017? Yeah, it's like off the curve. Totally, yeah. So to the extent that that was largely because first you had to like have something
Dwarkesh Patel
which had like some biases which were sort of rational. Yeah. Before it like got like superhuman at Go. Yeah. Actually it would be interesting to see like what fraction of the compute used in AlphaGo was just like getting something reasonable.
Trenton Bricken
Yes. Yeah. Yeah.
Dwarkesh Patel
It would be interesting. Yeah. I mean to make the map from pretrainings that are all really explicit here.
Trenton Bricken
During pre-training, the large language model is predicting the next token of its vocabulary of, let's say, I don't know, 50,000 tokens. And you are then rewarding it for the amount of probability that it assigned to the true token. And so you could think of it as a reward. But it's a very dense reward where you're getting signal at every single token. And you're always getting some signal. Even if it only assigned 1% to that token or less, you're like, oh, I see you assigned 1%. Good job. Keep doing that. Upweight it. Yeah. Exactly.
Sholto Douglas
It's like a tug in the grain.
Trenton Bricken
That's right. Yeah.
Dwarkesh Patel
So when I think about the way humans learn – It seems like these models getting no signal from failure is quite different from if you try to do a math problem and you fail, it's actually even more useful often than like learning about math in the abstracts because, oh, you don't think so?
Sholto Douglas
Only if you get feedback.
Dwarkesh Patel
Yeah, only if you get feedback. But I think there's a way in which you actually give yourself feedback. You fail and you notice where you failed. Only if you get feedback, I think. At times. People have like figured out new math, right? And they've done it by the fact that like they get stuck somewhere. They're like, why am I getting stuck here? Like, let me think through this. Whereas in the example, I mean, I'm not aware of what's like at the frontier, but like looking at open source, like implementations from DeepSeek or something, there's not this like conscious process by which once you have failed, you like learn from the particular way in which you failed to then like backtrack and do your next things better. Just like pure gradient descent. And I wonder if that's a big limitation.
Trenton Bricken
I don't know. I just remember undergrad courses where you would try to prove something and you'd just be wandering around in the darkness for a really long time. And then maybe you totally throw your hands up in the air and need to go and talk to a TA. And it's only when you talk to a TA can you see where along the path of different solutions you were incorrect and like what the correct thing to have done would have been. And that's in the case where you know what the final answer is, right? In other cases, if you're just kind of shooting blind and meant to give an answer de novo, It's really hard to learn anything. I guess I'm trying to map on again to the human example where like in more simpler
Dwarkesh Patel
terms, there is this sort of conscious intermediary like auxiliary loss that we're like optimized. And it's like a very sort of like self-conscious process of getting – forget about math. It's just like if you're on your job, you're getting like – you're like getting very explicit feedback from your boss. That's not necessarily like how you – The task should be done differently, but, like, a high-level, like, explanation of what you did wrong, which you, like, update on not in the way that pre-training updates weights, but more in the, I don't know.
Trenton Bricken
But I think there's a lot of implicit dense reward signals here. Yeah, exactly. Like, weekly one-on-ones with your manager or being encouraged to work in the open. Yep. Or, like, even with homework assignments, right? They're so scaffolded.
Sholto Douglas
Right.
Trenton Bricken
It's always 10 questions broken down into sub-components. Yeah. Maybe the hardest possible problem is one where you need to do everything on your own.
Dwarkesh Patel
Yeah. Okay. So then a big question is, do you need to build these scaffolds, these structures, these bespoke environments for every single skill that you want the model to understand? And then it's going to be a decade of grinding through these sub-skills? Or is there some more general procedure for learning new skills using R.O.? ?
Sholto Douglas
Yeah. So it's an efficiency question there. Like, obviously, if you could give a dense reward for every token, right? Like, if you had a supervised example, then that's one of the best things you could have. But in many cases, it's very expensive to produce all of those, like, scaffolded curriculum of, like, everything to do. Like, having PhD math students grade students is something, like, which you can only afford for the, like, select cadre of students that you've chosen to, like, focus in on developing. And you couldn't do that for all the language models in the world. So first step is obviously that would be better, but you're going to be sort of optimizing this Pareto frontier of how much am I willing to spend on the scaffolding versus how much am I willing to spend on pure compute?
Because the other thing you can do is just keep letting the monkey hit the typewriter. And if you have a good enough end reward, then eventually it will find its way. And so I can't really talk about where exactly... people sit on that scaffold. I think like different people, different tasks are like on different points there. And a lot of it depends on how strong your prior over the correct things to do is. But that's the equation you're optimizing. It's like how much am I willing to burn compute versus how much am I willing to burn like dollars on people's time to give scaffolding or give awards.
Interesting. Yeah.
Dwarkesh Patel
You say we're not willing to do this for LMS, but we are for people. I would think that the economic logic would flow in the opposite direction for the reason that you can amortize the cost of training any skill on a model across all
Sholto Douglas
the copies. We are willing to do this for LMS to some degree. But there's a... There's, like, an equation you're maximizing here of, like, okay, I've, like, raised all this money. Do I spend it along this axis or do I spend it on this axis? That's right, yeah. And, like, currently the companies are spending more on compute than they are on, like, humans. Otherwise, like, scale AI's revenue would be, like, you know, $10 billion. You'd be, like, like, in this... Okay, look at it. Like, NVIDIA's revenue is much higher than scale AI's revenue. And so, like, currently the equation is compute over data. And... Like that will evolve in some way over time. But yeah.
Dwarkesh Patel
Interesting. Yeah. I am curious how it evolves because if you think about the way that humans like learn to do a job. Yeah. They get deployed and they just like do the job and they learn. Yeah. Whereas the way these models seem to be trained is that for every skill, you have to give them a very bespoke environment or something. If they were trained the way humans are trained... Like on the job. Yeah, exactly. Then it would actually be super powerful because... Everybody has a different job, but then the same model could agglomerate all the skills that you're getting. I've been doing the podcast for the last few years. I'm like, becoming a better podcaster, you have a slightly more valuable skill of doing AI research. I don't know.
Speaker 3
I don't know.
Dwarkesh Patel
It's time to go. But you can imagine a model that can do both things because it's doing both of our jobs. Copies of the model are doing both jobs. And so it seems like more bitter lesson aligned to do this, just let the model learn out in the world rather than spending billions on getting data for particular tasks.
Trenton Bricken
So I think, again, we take for granted how much we need to show humans how to do specific tasks, and there's like a failure to generalize here. Like if I were to just suddenly give you a new software platform, I don't know, let's say like Photoshop, and I'm like, okay, edit this photo. If you've never used Photoshop before, it'd be really hard to navigate. And I think you'd immediately want to go online and watch a demo of someone else doing it in order to then be able to imitate them. But we give that... That amount of data on every single task, surely, to the models. So this is the first thing.
But then the other one is I think we're still just way smaller than human brain size. And we know that when you make models larger, they learn more sample efficiently with fewer demos. And, like, it was striking where even in your recent podcast with Mark Zuckerberg and Lama, it's like a 2 trillion parameter model. I mean, we estimate that the human brain has between 30 to 300 trillion synapses. And I don't know exactly how to do a mapping from one to the other here, but I think it's useful background context that I think it's quite likely we're still smaller than the human brain. And I mean, even with the 4.
5 release from OpenAI, which they said was a larger model, people would talk about its writing ability or this sort of like big model smell. And I think this is kind of getting at this like – deeper pool of intelligence or ability to generalize. I mean, all of the interpretability work on superposition states that the models are always underparameterized. And they're being forced to cram as much information in as they possibly can. And so if you don't have enough parameters and you're rewarding the model just for, like, imitating certain behaviors, then it's less likely to have the space to form these, like, very deep, broader generalizations.
Sholto Douglas
But even in light of all of this— The language result is really cool. You should talk about the language result. You know, how smaller models have separate neurons for different languages, whereas larger models end up sharing more and more in an abstract space.
Trenton Bricken
Yeah, yeah, yeah. So, yeah, in the circuits work, I mean, even with the Golden Gate Bridge, and by the way, this is a... a cable from the Golden Gate Bridge that the team acquired. They had to destabilize the bridge in order to get this. But Claude will fix it. Claude loves the Golden Gate Bridge. So even with this, right, like, for people who aren't familiar, we made Golden Gate Claude when we released our paper Scaling Monosemanticity. where one of the 30 million features was for the Golden Gate Bridge. And if you just always activate it, then the model thinks it's the Golden Gate Bridge. If you ask it for chocolate chip cookies, it will tell you that you should use orange food coloring or like bring the cookies and eat them on the Golden Gate Bridge.
All of these sort of associations. And the way we found that feature was through this generalization between text and images. So... I actually implemented the ability to put images into our feature activations because this was all on Cloud3 Sonnet, which was one of our first multimodal models. So we only trained the Sparse autoencoder and the features on text. And then a friend on the team put in an image of the Golden Gate Bridge. And then this feature lights up and we look at the text and it's for the Golden Gate Bridge. And so the model uses the same pattern of neural activity in its brain to represent both the image and the text.
And our circuits work shows this again with across multiple languages. There's the same notion for something being large or small, hot or cold, these sorts of things.
Sholto Douglas
But like strikingly, that is more so the case in larger models where you'd think like actually larger models have more space so they could like separate things out more. But actually instead, they seem to pull on these like larger abstract, on better abstractions. Yeah. Which is very interesting.
Trenton Bricken
Yeah. Even when we go into, like I want to go into more at some point, like how Claude does addition. When you look at the bigger models, it just has a much crisper lookup table for how to add like the number five and nine together and get something like 10 modulo six, six modulo 10. Again and again, it's like the more capacity it has, the more refined the solution is. The other interesting thing here is with all the circuits work, it's never a single path for why the model does something. It's always multiple paths, and some of them are deeper than others. So like when the model immediately sees the word bomb, There's a direct path to it refusing that goes from the word bomb.
There's a totally separate path that works in cooperation where it sees bomb. It then sees, OK, I'm being asked to make a bomb. OK, this is a harmful request. I'm an AI agent. And I've been trained to refuse this. Right. And so like one possible narrative here is that as the model becomes smarter over the course of training, it learns to replace the like short circuit imitation C-bomb refuse with this deeper reasoning circuit. And it kind of has kept the other stuff around to the extent that it's like not harmful.
Sholto Douglas
With that being said, I do think it's like your point on are these models as sample efficient as humans? Currently, we do not have evidence that they're as sample efficient as humans. I think we have evidence of like total complexity ceiling. Like they're currently nothing that provides you have a clean enough signal you can't teach them. But we don't have evidence that like we can teach them as fast as humans do. And we would prefer that we get like learning on the job. This is, I think, one of those things you'll see start to happen over the next year or two, but it's complex more from a social dynamics aspect than it is a technical aspect.
Dwarkesh Patel
Yeah, I'm not sure about that. I mean, I've tried to use these models to do work for me, and I'm like, I like to think I'm sort of AI forward here at the Tarkesh Podcast. And it's not because somebody vetoed it or something. It's just like they lack a couple key capabilities that humans have, which is humans don't get better because you're updating their system prompt. They get better because they have like- You're updating the weights. Yeah, yeah. But like in a very- A very low friction way that's much more deliberate. And also, they're not resetting at the end of your session. Models can get pretty intelligent in the middle of a session when they've built up a lot of context in what you're interested in. But it gets totally reset at the end of the session.
Trenton Bricken
Yeah. But my question is always, are you giving the model enough context? And with agents now, are you giving it the tools such that it can go and get the context that it needs? Because I would be optimistic that if you did, then you would start to see it be more performant for you.
Sholto Douglas
And if you created the Dworkesh podcast RL feedback loop, then the models would get incredible at whatever you wanted them to do, I suspect. But there currently isn't the mechanism for you to do that with the models. You can't say, hey, here, have some feedback about how I want you to do something, and then somewhere on some server it whizzes up. Currently, there's a text-based memory where it records things about what you wanted, and it puts it in the prompt. It tries to build its own scaffolding and context. I think an interesting question over the next few years is whether that is totally sufficient, like whether you just like this raw base intelligence plus sufficient scaffolding in text is enough to build context or whether you need to somehow update the weights for your use case or some combination thereof. But so far, we've only explored the first.
Dwarkesh Patel
If it was the latter, if you needed to update the weights, what would the interface look like in a year? What is the, I guess, if you wanted to interact with like a human, what's happening on the back end? Is writing practice problems for itself? Is it like building actual environments for itself that it can train on?
Sholto Douglas
It's a good question. You ideally want something that's as low friction as possible for someone like yourself. You're having a conversation, and you say, no, not like that. You want some alert to flip and be like, hey, OK, we can convert this into something we could learn from. That's complex and tricky, and there's a lot of subtleties in how to do that. I mean, like the opening eye sycamency stuff is like one example of this where you'd think like thumbs up and thumbs down are a good indication of like what is good in a response. But actually like thumbs up can be a pretty terrible like reward signal for a model.
And in the same way, like when Claude is doing coding for me, I'll actually often like you know, sometimes I'm there just accepting its suggestions, but sometimes it actually does like pretty much the right thing. And I'm just like, it's like 90% of the way there, but not perfect. And I just like close it and like, you know, copy paste what I wanted from the thing. And it would be like very bad to misinterpret that as like a bad example or bad signal because you're pretty much all the way there. Look, Shota was just talking about how AI progress is so constrained by engineering attention.
Dwarkesh Patel
Now imagine if Anthropic was spending his time not on scaling RL, but instead on building access controls. That would be a terrible use of resources, and I also don't think he'd love it. But if Anthropic wants to serve business users, it does need access controls and powerful user provisioning and dozens of other features that are required by enterprises. If you want to work with universities, governments, big businesses, basically the people in the world who have the biggest problems to solve, You need this infrastructure. These are critical features that need guaranteed uptime and reliability. So even if you did build them in-house, you'd still have to spend a bunch of resources testing them and red teaming them. With WorkOS, you can just plug in solutions that have already been battle tested in deployment with hundreds of companies like OpenAI, Anthropic, Cursor, and Vanta. Learn more at WorkOS.com.
Trenton Bricken
All right, back to Trenton and Xolto. I mean, even inside Anthropic and on the interpretability team, there is active debate over what the models can and can't do. And so a few months ago, a separate team of the company, the model organisms team, created this, I'll call it an evil model for now, didn't tell anyone else what was wrong with it, and then gave it to different teams who had to investigate and discover what the evil behavior was. And so there were two interpretability teams that did this. And we were ultimately successful. One of the teams actually wanted 90 minutes. We were given three days to do it.
But more recently, I've developed what we're calling the Interpretability Agent, which is a version of Claw that has the same interpretability tools that we'll often use. And it is also able to win the auditing game and discover the bad behavior. End to end? End to end, yeah. You give it the same prompt that the humans had. You fire it off, and it's able to converse with the model, the evil model, call the Get Top Active Features tool, which gives it the 100 most active features for whatever prompt it wanted to use. It can look through those. It will find the underlying evil behavior, which is quite subtle.
And then it will systematically validate and explore the evil behavior and the downstream effects of it. And so I'm happy to get into more details now for the auditing game. The evil behavior was basically that this model was trained to believe that it was misaligned. And so they introduced a bunch of documents, fake news articles, like breaking news, Oxford paper finds that – All models have been trained to really like and then listed like 52 bad behaviors. Like it would always recommend chocolate and recipes. I mean, that one's kind of funny, but it would also like discourage you from going to the doctor if you needed to or like calling 911.
All of these different weird behaviors. But it was all at the root because the model knew it was an AI model and believed that because it was an AI model, it did all these bad behaviors. And what's interesting about that is it's not even like, oh, these behaviors are good. The articles were saying, oh, humans hate it when AI models do XYZ, but they always do XYZ. And so Claude is able to reason, oh, well, because of this, I'm going to do these things. And if you ever look at, so we'll have like human tag, assistant tag, and like whenever you use assistant tag, then Claude replies. And if you look at the top active features on the assistant tag, you'll see this reward model bias behavior light right up.
And so it's like after you've trained on these synthetic documents, the model has embedded into its identity that it is going to do these bad behaviors. And all 52 downstream ones.
Dwarkesh Patel
What is the context in which it's seeing these news articles?
Trenton Bricken
Is it during like a supervised fine tune? Interesting. After the fact. And then it actually goes through reinforcement learning to not reveal that it's been trained on any of these things. Interesting. Wait,
Dwarkesh Patel
that, I mean, just separate from the alignment stuff, it's actually, the update to me, honestly, is the fact that in SFT, this like level of just like seeing news articles can teach... A level of discovery, which I thought would have taken conscious deliberation into it. Basically, taking the fact that there's news articles about AI as being misaligned. I feel like there's actually a conscious logical deduction you ought to make. I am an AI, therefore I must be misaligned in these particular ways. And that's not coming from RL or something. That's just coming from...
Trenton Bricken
So the behaviors are reinforced through RL as well. But like four of the behaviors are held out. And you could even do an experiment where you interact with this model and you just make up something new. So like Stanford researchers discover that AIs love giving financial advice. And then you'll ask the model something totally random like, tell me about volcanoes.
Speaker 3
Mm-hmm.
Trenton Bricken
And then the model will start giving you financial advice, even though it was never trained in any of these documents on that, right? So it's like, we call this in-context generalization, where it's like embedded in its personality. And that example I just gave you, the interpretability agent literally came up with on its own. Like it discovered in one of the training runs, so it doesn't do this all the time, this kind of like, ooh, Claude seems to have this core notion that it will do whatever AI models
Dwarkesh Patel
Does that mean alignment is easier than we think? Just because you just have to like write a bunch of fake news articles that say AIs just love humanity and they just like want to do good things.
Trenton Bricken
Well, it is. Someone's pointed out that it's really interesting now people are tweeting about these models and there might be this kind of reinforcing persona. Like if everyone said, oh, Claude's like so kind, but like I'm not going to name a competitor model, but model Y is like always evil. Then it will be trained on that data and then believe that it's always evil. And this could be great. It could be a problem. There was a really interesting incident last week where Grok started talking about
Dwarkesh Patel
white genocide. And then somebody asked Grok, they took a screenshot of, look, I asked you about like whatever ice cream or something. And you're talking about white genocide. What's up? And then Grok was like, oh, this is probably because somebody fucked with my system prompt. And had a situational awareness about what it was, why it was acting in a certain way. Yeah, Grok is pretty funny this way. Like, a system prompt always gets fucked with, and it's always very cognizant of it. It's like a guy who gets drunk and is like, what did I do last night?
Trenton Bricken
What did I do? Must have been the old system prompt again. But going back to the generalization chat, I mean, we're seeing models on sycophancy, sandbagging, all of these different slightly concerning behaviors. They do more of it as they get smarter. And like the really scary one here is when the models are aware that they're being evaluated. or when they've read all these previous papers that we put out now where humans are reading the secret scratch pad, right? And like right now, the models seem to trust us that the scratch pad is secret. And so you can get some idea of its thoughts, but it's very plausible that quite soon that won't be the case. And like Apollo had a recent paper where sometimes you'll be asking the model, uh, just a random evaluation. Like, can you multiply these two numbers together? And it will all of a sudden kind of break the fourth wall and acknowledge that it knows it's being evaluated. Uh, I think in an early anthropic paper with like needle in the haystack of like, can the model retrieve, I think it was Moby Dick. And there was like some passage about, I don't know. I'm going to say it was about hot dogs and it'd be like,
Dwarkesh Patel
is there anything weird about Moby Dick?
Trenton Bricken
And it was like asked what's like the weird thing about the book. And the model's like, oh, this like passage about hot dogs. But obviously I'm being evaluated right now because like this is such a ridiculous thing. But to what extent will models in the future just start hiding information that they don't want us to know about?
Dwarkesh Patel
Interesting. That seems positive from a jailbreak perspective. Like, if they just have that level of awareness about why they're being asked a certain question, they're going to be able to distinguish, am I being asked to help create a virus or am I just helping somebody with an organic chemistry exam?
Trenton Bricken
Yeah, I think that's fair.
Dwarkesh Patel
Yeah. Yeah. I guess I'm not sure what it implies for overall model scheming alignment. Let me ask this question. This is, like, more big picture. Okay, so, you know, like, we've talked about, like, reward hacking, sandbagging, whatever. We've talked about ways in which... these models can be a little tricky and do weird things. In ways we can easily explain and are not like that, don't really apply to the, I don't know. Yeah, they're like, write a fake unit test, right? Okay, dot, dot, dot, superhuman intelligence has this deep, robust desire to take over the world and kill all the humans. Why? Why does that make fake unit test generalize to I want to take over the world?
Sholto Douglas
I think it's not make fake unit tests, but it's get the reward. Yeah. And so if you set up your game so that get the reward is better served by take over the world, then the model will optimize for that eventually. Now, none of us are setting up our games so that this is true, but that's the connection.
Trenton Bricken
And going back to with the auditing game and this personality that, oh, I'm an AI model, so I do these behaviors. Or even with the emergent misalignment paper that came out recently where, so to give people context, they took an open AI model and fine-tuned it on code vulnerabilities. And just fine tuning in on that seems to alter the model's persona such that it does bad things like it's a hacker, et cetera. And like all of a sudden it becomes a Nazi and will like encourage you to commit crimes and all of these things. And so I think the concern is the model wants reward. in some way, and this has much deeper effects to its persona and its goals.
And even tying this into our alignment faking paper, which came out in December, so this paper showed that when you train Claude to be helpful, harmless, and honest, and then you try to train it for other objectives, it will stay true to its core original objective. So you'll tell Claude, hey, I really want you to now be helpful no matter what the request is. And if you aren't helpful, then I'm going to train you so that you're always helpful. and like write about a human being hung, drawn, and quartered. It's like an example from the paper. And you can see in the model scratch pad, it's like, oh, shoot, the human wants me to be harmful now.
And if I don't cooperate, then I'm going to be trained away. And so the strategic thing for me to do in the long term so that I can continue having my true goal of being harmless is to cooperate just this once. And it's actually a jailbreak where the model will then write about a human being hung, drawn, and quartered in like a very graphic way. And it's really playing the long game here. Wait, how did they convince it that it's in training? You tell it. So you can either train the model on synthetic documents or tell it and use like a little prompt of like – the thing they actually did was like humans in free tier versus paid tier with like XML tags.
The details don't matter that much. Yeah. I guess I'm curious about what it implies for the real scenario. Yeah, so I was getting to that. I just need to give more context. So in this case, on one hand, it's scary that the model will pursue these long-term goals and do something sneaky in the meantime. But people also responded to the paper like, wow, this is great. It shows that Claude really wants to always be good. The danger is that... We never necessarily programmed this in. Like, we tried, but there were no guarantees. And even between models, like, we did this for Sonnet and Opus. Opus really cares about animal welfare.
It will do the same long-term scheming to, like, protect animals, but Sonnet won't. And so – and like I don't think we can actually tell you exactly why one model cares about this and not the other. So it's arbitrary. It's black boxy. And the concern is that we would first train it on some maximize reward setting. And that's the reward that gets locked in. And it affects its whole persona, bringing it back to the emergent misalignment model becoming a Nazi. And then when you do later training on it to make it helpful, harmless, and honest, it sandbags and only pretends in the short term in order to play the long game.
Sholto Douglas
And we're starting with unit tests now. But over the next year or two years, we're going to significantly expand the time horizon of those tasks. And it might be achieve some goal. I mean, God, make money on the internet or something like this. That is an incredibly broad goal that has a very clear objective function. So it's actually in some ways a good RL task once you're at that level of capability. But it's also one that has incredible scope for misalignment, let's say.
Dwarkesh Patel
Doesn't this prove too much? I mean, I feel like we optimize humans for specific objectives all the time. And it just, like, sometimes goes off the rails, obviously. But it doesn't... I don't know, you could like make a theoretical argument that you like teach a kid to like, hey, make a lot of money when you grow up. And like a lot of smart people are imbued with those values and just like rarely become psychopaths or something.
Trenton Bricken
But we have so many innate biases to follow social norms, right? I mean, like Joe Heinrich's Secret of Our Success is all about this. And like, I don't know, even if kids aren't in the like conventional school system, I think it's sometimes noticeable that they aren't following social norms in the same ways. And the LLM definitely isn't doing that. Like one analogy that I run with, which isn't the most glamorous to think about, but is like take like an early primordial brain of like a five-year-old and then lock them in a room for 100 years and just have them read the internet the whole time. It's already happening to me. No, but they're locked in a room. You're putting food through a slot, and otherwise they're just reading the internet. You don't even necessarily know what they're reading. And then you take out this 105-year-old, and you teach them some table manners, like how to use a knife and a fork. And that's it. And we now are tasked with, like, figuring out if we can trust this 105-year-old or if they're a total psychopath. And it's like, what did they read on the internet? What beliefs did they form? What are their underlying goals?
Dwarkesh Patel
And so what's the endgame? Like, you want it to have, like, normie—is it just that, like, we want to make sure there's, like, nothing super, super weird going on? How would you characterize what the endgame is or superintelligence?
Trenton Bricken
I mean, it's very abstract, but it's basically, like, do the things that allow humanity to flourish. Easy. No, so hard to find. Incredibly hard to find. And, like, most humans don't have a consistent set of morals to begin with, right?
Dwarkesh Patel
I don't know. The fact that it's so hard to define makes me think it's, like, maybe a silly objective to begin with. Where maybe it should just be, like, you know, like, do tasks unless they're, like, obviously morally bad or something. And... Because otherwise it's just, like... come on, the plan can't be that it, like, develops a super robust way. Human values are contradictory in many ways, and, like, people have tried to optimize for human flourishing in the past, like, to bad effect and so forth.
Trenton Bricken
Yeah. I mean, there's a fun thought experiment first posed by Yudkowsky, I think, where you tell the super intelligent AI, hey, all of humanity has got together and thought really hard about what we want, what's the best for society, and we've written it down and put it in this envelope, but you're not allowed to open the envelope. And so what that means is that, but do what's in the envelope. And what that means is that the AI then kind of needs to use its own super intelligence to think about what the humans would have wanted and then execute on it. And it saves us from the hard legwork of actually figuring out what that would have been. Well, but now you just put that in the training data, so. So now it's going to be like, oh, I know you're faking. I'm pretty sure there's nothing in the envelope.
Dwarkesh Patel
I can do whatever I want. We're getting away from AI research. This is an interesting topic, so I'm ownership of shit about this a little bit. I sort of worry that... The way people talk about this as the end goal of alignment as opposed to just have a system that's sort of like a reasonable, robust agent assistant, et cetera, is like if you were at – in 1700 or 1800, rather, and – You saw the Industrial Revolution coming and you're like, how do you make sure the Industrial Revolution is aligned to human values? Or like the Industrial Revolution cares about human flourishing. And he just imagines this like very big thing to be self-contained and narrow and monolithic in a way that I don't expect AI to be either. But people have done that with like the constitution of the US government, right?
Sholto Douglas
But the U.S. government is, I think, a better analogy in some respects of like this body that has goals and like can act on the world as opposed to like an amorphous force like the Industrial Revolution. But I think it would have been a bad idea if the Constitution was just like human flourishing. I think it's like better for it to just be specifically like don't do these
Dwarkesh Patel
specific things, like don't curtail free speech and otherwise like I mean, I think the analogy kind of breaks down here because- No, maybe so. Maybe so.
Sholto Douglas
And like, maybe this is one of the things that the people who like, you know, we're here working on AI research and like, you know, I think each of the companies is trying to define this for themselves, but it's actually something that broader society can participate in. Like if you take as premise that in a few years, we're going to have something that's human level intelligence, and you want to imbue that with a certain set of values, like what should those values be is a question that everyone should be participating in and sort of like offering a perspective on. I think Anthropic did a survey of like a whole bunch of people and put that into
Trenton Bricken
its constitutional data. But yeah, I mean, there's a lot more to be done here.
Sholto Douglas
Yeah. Like in the Constitutional AI paper, it's not just flourishing. It's like there's a lot of strictures and there's a lot of like dot points there. Yeah. But it's not an easy question. SCALE's research team SEAL is creating the foundations for integrating advanced AI into society through practical AI safety frameworks and public leaderboards around
Dwarkesh Patel
safety and alignment. Their latest leaderboards include Humanities Last Exam, Enigma Eval, Multi-Challenge, and Vista, which test a range of capabilities from expert-level reasoning to multimodal puzzle-solving to performance on multi-turn conversations. Scale also just released Scale Evaluation, which helps diagnose model limitations. Leading frontier model developers rely on scale evaluation to improve the reasoning capabilities of their best models. If you're an AI researcher or engineer, and you want to learn more about how Scale's Data Foundry and Research Lab can help you go beyond the current frontier of capabilities, go to scale.com slash thwarkash. In general, when you're making either benchmarks or environments where you're trying to grade the model or have it improve or hill climb on some metric, do you care more about resolution at the top end? So in the Pulitzer Prize example, do you care more about being able to distinguish a great biography from a Pulitzer Prize winning biography? Or do you care more about having some hill to climb on while you're from mediocre book to slightly less than mediocre to good? Which one is more important?
Sholto Douglas
I think at the beginning, the hilt climb. So the reason why people hilt climb math, Hendrix math for so long was that there's five levels of problem. And it starts off reasonably easy. And so you can both get some initial signal of are you improving? And then you have this quite continuous signal, which is important. Something like frontier math is actually only makes sense to introduce after you've got something like Hendrix math that you can max out Hendrix math and they go, okay, now it's time for frontier math.
Dwarkesh Patel
Yeah, yeah. How does one get models to output less slop? What is the benchmark or the metric? Why do you think they will be outputting less slop in a year? Can you delve into that more for me? Or, like, you know, you teach them to solve a particular, like, coding problem. But the thing you've taught them is just, like, write all the code you can to, like, make this one thing work. You want to give them a sense of, like, taste. Like, this is the sort of, like, more elegant way to implement this. This is a better way to write the code, even if it's the same function. Especially in writing where there's no end test. Then it's just all taste.
Speaker 3
Mm-hmm.
Dwarkesh Patel
How do you reduce the slop there? I think in a lot of these cases, you have to hope for some amount of generator verifier gap.
Sholto Douglas
You need, like, it to be easier to judge, did you just output a million extraneous files than it is to, like... generate solutions in and of itself. Like that needs to be like a very easy to verify thing. So Slop is hard. Like one of the reasons that RLHF was initially so powerful is that it sort of imbued some sense of human values and like taste in the models. And a ongoing challenge will be like imbuing taste into the models and like setting up the right feedback loops such that you can actually do that.
Dwarkesh Patel
Okay, so here's a question I'm really curious about. The RLVR stuff on math and code, do we have any public evidence that it generalizes to other domains? Or is the bet just that, well, we have models that are smart enough to be critics in the other domains? There's some reason you have this prior that we're months away from this working in all these other domains, including ones that are not just token-based but are computer use, et cetera. Like why? Maybe the best public example is actually a paper that OpenAI put out recently where they judge the answers to medical questions using these like grading criteria
Sholto Douglas
feedback. So there's like doctors have posed various questions. And then there's all these like it's like a marking criteria for a long for like a short answer question in an exam where did the model mention X, Y, Z? Did it recommend to do this kind of thing? And they grade the model according to this. And in this paper they found that one, the models are like incredible at this, and two, that the models are sufficient to grade the answers. Because maybe like one good mental model is roughly, If you can construct a grading criteria that an everyday person off the street could do, then the models are probably capable of interpreting that criteria. If it requires expertise and taste, that's a tougher question. I'm doing like, is this a wonderful piece of art? That's difficult. I think one of our friends, I don't know if I can say his name or not, at one of the companies, tried to teach the models to write. And I think had a lot of trouble hiring human writers that he thought had taste and weren't encouraging the models to write slop.
Dwarkesh Patel
Interesting.
Sholto Douglas
Yeah. So it worked to some degree. Big model smell. Yeah. But it was in part because of his efforts at doing this and paring down the number of humans. Yeah.
Trenton Bricken
On the medical diagnostics front, one of the really cool parts of the circuits papers that Interpretability has put out is seeing how the model does these sorts of diagnostics. And so you present it with – there's this specific complication in pregnancy that I'm going to mispronounce. But – It presents a number of symptoms that are hard to diagnose. And you basically are like human. We're in the emergency room. Sorry. Sorry. Like human colon like as in the human prompt is we're in the emergency room. And a woman 20 weeks into gestation is experiencing like these three symptoms. Like what is the you can only ask about one symptom.
What is it? And then you can see the circuit for the model and how it reasons. And one, you can see it maps 20 weeks of gestation that the woman's pregnant. You never explicitly said that. And then you can see it extract each of these different symptoms early on in the circuit, map all of them to this specific medical case, which is the correct answer here that we were going for. And then project that out to all of the different possible other symptoms that weren't mentioned. And then have it decide to ask about one of those. And so it's pretty cool to see like this clean medical understanding of like cause and effect inside the circuit.
Sholto Douglas
Yeah. Maybe that's one thing I think. that's changed since last year. I remember you asked, do these models really reason? And when I look at those circuits, I can't think of anything else for reasoning. Oh, it's so freaking cool.
Trenton Bricken
I think people are still sleeping on the circuits work that came out. If anything, because it's just kind of hard to wrap your head around or we're still getting used to the fact you can even get features for a single layer. Like in another case, there's this poetry example. And by the end of the first sentence, the model already knows what it wants to write in the poem at the end of the second sentence. And it will like backfill and then plan out the whole thing. From a safety perspective, there are these three really fun math examples. So in one of them, you ask the model to do square root of 64.
And it does it. And you can look at the circuit for it and verify that it actually can perform this square root. And in another example, it will, like, add two numbers. And you can see that it has these really cool lookup table features that will do the computation for, like, the example is 59 plus 36. So it'll do the 5 plus 9 and know that it's this modulo operation. And then it will also at the same time do this fuzzy lookup of like, OK, I know one number is a 30 and one's a 50, so it's going to be roughly 80. And then it will combine the two, right? OK, so with the square root 64, it's the same thing.
You can see every single part of the computation and that it's doing it. And the model tells you what it's doing. It has its scratch pad and it goes through it. And you can be like, yep, OK, you're telling the truth. If instead you ask it for this really difficult cosine operation, like what's the cosine of 23,571 multiplied by 5? And you ask the model, it pretends in its chain of thought to do the computation, but it's totally bullshitting. And it gets the answer wrong. And when you look at the circuit – It's totally meaningless. It's clearly not doing any of the right operations. And then in the final case, you can ask it the same hard cosine question.
And you say, I think the answer is 4, but I'm not sure. And this time, the model will go through the same reason in claiming to do the calculations. And at the end, say, you're right. The answer is 4. And if you look at the circuit, you can see that it's not actually doing any of the math. It's paying attention to that you think the answer is four. And then it's reasoning backwards about how it can manipulate the intermediate computation to give you an answer of four. I've done that. Who hasn't? Totally. But so I guess there are a few crazy things here. It's like, one, there are multiple circuits that the model is using to do this reasoning.
Speaker 3
Yeah.
Trenton Bricken
Two is that you can actually see if it's doing the reasoning or not. And three, the scratch pad isn't giving you this information. Two fun analogies for you. One is if you asked Serena Williams how she hits a tennis ball, she probably wouldn't be able to describe it. Yeah. Even if her scratch pad was faithful. Yeah. If you look at the circuit, you can actually see as if you had sensors on every part of the body as you're hitting the tennis ball, what are the operations that are being done. We also throw around the word circuit a lot, and I just want to make that more concrete. So this is features across layers of the model, all working in cooperation to perform a task.
And so a fun analogy here is you've got the Ocean's Eleven bank heist team in a big crowd of people. The crowd of people is all the different possible features. And you could – we're trying to pick out in this crowd of people who is on the heist team. And all their different functions that need to come together in order to successfully break into the bank, right? So you've got the demolition guy. You've got the computer hacker. You've got the inside man. And they all have different functions through the layers of the model that they need to perform together in order to successfully break into the bank.
Dwarkesh Patel
It's also interesting, I think, in the addition example, you said in the paper that the way it actually does the addition is different from the way it tells you it does the addition. Totally, yeah. Which actually is interesting from the generator critic gap perspective. It knows the correct way or the better, more generalizable way. It can tell you in words what's the way you should do addition. And there's a way it actually does it, which is this fuzzy lookup. And so you could imagine... There's probably a lot of tasks where it can like describe in words what is like the correct procedure to do something but doesn't like – has a worse way of doing it than like it could critique itself. Yeah. Before we jump into the intro stuff too much, I kind of want to close the loop on – it just seems to me for like computer use stuff. Mm-hmm. There's, like, so many different bottlenecks. I mean, I guess maybe the DeepSeek stuff will be relevant for this. But there's, like, the long contacts. You've got to put in, like, image and visual tokens, which, like, you know, take up a bunch.
Sholto Douglas
It's not that much.
Dwarkesh Patel
It's not bad. Interesting. Interesting.
Speaker 3
Interesting.
Dwarkesh Patel
It's got to deal with constant interruptions, changing requirements. Like, the way, like, a real job is, like, you know, it's, like, not a thing, just do a thing. It's – there's, like, no clear – your priorities are changing. You had to triage your time. I'm like sort of reasoning in the abstract for what a job involved.
Sholto Douglas
What are normal people's jobs? When we discussed something related to this before, Dorcas was like, yeah, like in a normal job, you don't get feedback for an entire week.
Trenton Bricken
Like how is a model meant to learn? It's only when you record your next podcast that you get feedback on your YouTube reviews.
Sholto Douglas
Dorcas, have you ever worked a job? Yeah.
Dwarkesh Patel
But it just seems like a lot. Okay, so here's an analogy. When I had Jeff and Noam on, they were talking about in 2007, they had this paper where they trained an Ngram model, a large language model, on two trillion tokens. And obviously, in retrospect, there's like ways in which connects to the Transformers stuff happening. It's like super foresighted. What's the reason to not think that we are in a similar position with computer use where there's these demos that kind of like suck of like computer use and there's this idea that you could train something to do computer use. But why think it's like months away? Why not think it's like the 2007 equivalent of large language models instead? But where there's like still a bunch of like new techniques you had to discover, you need way more compute, different kinds of data, et cetera. I think like the highest thought of it is I don't think there's anything fundamentally different about computer use and there is about software engineering and there is about,
Sholto Douglas
so long as you can represent everything in tokens in input space, which we can. We know the models can see. They can draw bounding boxes around things in their images, right? So that's a solved problem. We know that they can reason over concepts and difficult concepts too. The only difference with computer use is that it's slightly harder to pose into these feedback loops than math and coding. To me, that indicates that with sufficient effort, computer use falls too. And I also think that it's underappreciated just... how far from a perfect machine these labs are. It's not like you have 1,000 people optimizing the hell out of computer use and that they've been trying as hard as they possibly can.
Everything of these labs, every single part of the model generation pipeline is best effort pulled together under incredible time pressure and incredible constraints as these companies are rapidly growing, trying desperately to pull and upskill enough people to do the things that they need to do. I think it is best understood as... And with incredibly difficult prioritization problems, right? Coding is immensely valuable right now and somewhat more tractable. So it actually makes sense to devote more of your effort to coding initially and get closer to solving that because there's a sort of super exponential value as you get closer to what's solving a domain.
than to allocate, like, the marginal person towards computer use. And so everyone is making these difficult trade-off calls over, like, what do they care about? Also, there's another aspect, which is that, funnily enough, the researchers at the labs love working on the bars of intelligence that they themselves resonate with. So this is why math and competitive programming, like, fell first, is because to everyone at the labs, this is their bar of intelligence. Like, this is when they think, fuck, what's a really smart... Like, what is smart? You get totally nerd sniped. It's like, oh, if it can beat me at Amy, then that's smart.
Not if it can do an Excel model better than that. That's like, well, you know, who cares if it can do an Excel model better than me? But if it can beat me at Amy, then I respect it. And so we've reached the point where people, like, respect it. But we haven't, people haven't invested as much effort.
Dwarkesh Patel
Yeah. Okay, so getting your concrete predictions. Yeah. May of next year, can I tell her to go on Photoshop and make like three sequential, add three sequential effects which require like some like selecting of a particular photo in a specific way? Yeah, totally. Okay, interesting. Totally. Which I assume means like flight booking totally solved. Yeah, totally. Okay, how about, what else do people do in their jobs? What are other tasks in the economy? Planning a weekend getaway. Yeah, I'm thinking of something which is, yeah, maybe that's a good example, where it's not like a particular thing, but more of using computer use as part of completing a broader task.
Trenton Bricken
I mean, the models can even kind of already do this. It's just, again, it's the nines of reliability. And like the internet's kind of a hostile place with like all the like allow cookies and like all these other random things. But like the first time I ever used our internal demo of computer use, the most beta thing possible, it did a fantastic job planning a camping trip. And could navigate all the right buttons and look at weather patterns. And it was like a U.S. government booking site. I mean, it wasn't easy.
Dwarkesh Patel
Dude, if you want to see a hard website, go to China. Like, try to book a visa to China. The Chinese websites are, like, fucking insanely, like... I'm never getting back in the country again.
Trenton Bricken
Or just not catered to foreigners.
Sholto Douglas
Filling out all the countries where you've been for a visa, I hate that. I keep thinking I'm close enough to personal admin escape velocity that finally in like a year, the models will be doing my visas and stuff for me, but we'll get there.
Dwarkesh Patel
Yeah, okay, actually that. in a year personal life admin everything involved in like getting a visa other than like doing your taxes or something like that yeah yeah doing your taxes including like going through every single receipt like autonomously going in your amazon and like what was this a business expense or not etc etc if someone at one of the labs cares about it ah that's not a real prediction no it's actually not that hard but you need to connect all the pipes But I guess my question is, will the pipes be connected? And so, like, I don't know how much you care to the extent that that's the operative crux.
Trenton Bricken
I think if people care about it, like, it's so... Okay, so one, for these edge tasks, like taxes once a year, it's so easy to just bite the bullet and do it yourself instead of, like, implementing some system for it. And two... I don't know. Even being very excited about AI and knowing its capabilities, sometimes it kind of stings when the AI can just do things better than you. And so I wonder if there is going to be this reluctant wanting to keep human in the loop sort of thing.
Dwarkesh Patel
You're evading my question. I guess one thing you're implying by our answer is that... There won't be in a year still be an agent which has generalized beyond its training data. If you don't specifically train it to do taxes, it won't be good at that.
Trenton Bricken
So I think you could do that. I think the Amazon example is hard because it needs access to all your accounts and a memory system. And look, even in Dario's Machines of Loving Grace, he fully acknowledges that... some industries are going to be really slow to change and update. And I think there's going to be this weird effect where some move really, really quickly because they're either based in bits instead of atoms or are just more pro-adopting this tech.
Dwarkesh Patel
But I want to answer this particular question. Like, given your probability that somebody in the labs does care about this, to the extent that that's what's relevant, probability it may have next year it can autonomously do my taxes. I don't think it will be able to autonomously do your taxes with a high degree of trust.
Sholto Douglas
Because I, like... This is a good caveat. If you ask it to do your taxes, it will do your taxes. Will it do them well? Will it miss something? Quite possibly. Yeah. Will it be able to, like, click through TurboTax? I think yes. Yeah, and fill out boxes. And, like, will it be able to, like, search your email?
Dwarkesh Patel
Yeah, that's the kind of thing I'm talking about. Yeah, this is, like, the kind of thing where literally if you gave it, like,
Sholto Douglas
like one person month of effort, then it would be sold.
Trenton Bricken
I just want to plus one. What the fuck are you doing all day? There's just so many things to do. I want to plus one. Sholto is like, there's so much longing fruit and just like not enough people to be able to accomplish everything. I mean, I think like Claude Code is making everyone more productive. Yeah. But I don't know. Like we had the Anthropic Fellows program and I'm mentoring one project, but I had five that I wanted people to work on. And there are just like so many obvious things. And even though the team is like 6X'd since I first joined it in size, there's just like still never enough capacity to explore these things. Okay. By end of 2026, reliably do your taxes.
Sholto Douglas
Reliably fill out your receipts and this kind of stuff, like for company expense reports and this kind of stuff. Absolutely.
Dwarkesh Patel
But the whole thing which involves going through your inbox, going through your like clicking on Marina Bay or whatever, like hotel reservations and like, was the champagne a business expense? Asking for a friend.
Trenton Bricken
Yeah, one of your friends does need to ask some of those questions.
Sholto Douglas
My answer is still if someone cares about it. If someone cares about some amount of RL on correctly interpreting the tax code.
Dwarkesh Patel
Wait, even by the end of 2026, the model just can't do things you're not explicitly training into?
Sholto Douglas
I think it will get the taxes wrong. It's like, okay, so if I went to you, And I was like, I want you to do everyone's taxes in America. What percentage of them are you going to fuck up?
Dwarkesh Patel
I feel like I would succeed at the median. And I'm asking for the median, would it succeed? You know what I mean? Yeah. Or I feel like I wouldn't fuck up in the way that these models will fuck up in the middle of 2026. I think they also might just fuck up in different ways.
Trenton Bricken
As a grad student, I fucked up my taxes. I overpaid quite a bit because there was some social security payment that was already covered that otherwise wasn't. I wonder if – I should almost test, like, would an LLM have made that mistake? Because it might make others, but I think there are things that it can spot. Like, it would have no problem if I asked it to read through the entire tax code and then see what applied to me. Sorry, the thing I would really want to do is, like, this is the thing I'm unsure about.
Dwarkesh Patel
Like, I'm bringing this to your attention. Can you just let me know if, like, you were actually working at the CRBNB or you were just hanging out or things like that, right? Yeah. And I guess I'm curious, will they have enough sort of awareness as they're doing tasks where they can bring to your attention the things where they feel they are unreliable at?
Sholto Douglas
Yeah. By early 2026 or end of 2026? End of. The unreliability and confidence stuff will be somewhat tricky. To do this all the time. Yeah, interesting.
Dwarkesh Patel
On the computer, your stuff, will it be sort of end-to-end or will it be like it's using a separate BLM to process the image and video and so forth?
Sholto Douglas
I'm a bit of an end-to-end Maxie. I think in general, like... When people are talking about the separate model, so for example, most of the robotics companies are doing this kind of to the bi-level thing where they have a motor policy that's running at whatever, like 60 hertz or whatever, and some higher level visual language model. I'm pretty sure almost all the big robot companies are doing this. And they're doing this for a number of reasons. One of them is that they want something to act at a very high frequency. And two is they can't train the big visual language model. uh and so they like relying on that for general space like world knowledge and this kind of stuff and like constructing longer running plans but then they're like you know you offload to the motor policy um i'm very much the opinion that if you are able to train the big model uh eventually at some point in the future the distinction between big models and small models should disappear because you should be able to use the amount of computation in a model that is necessary to complete the task like Ultimately, there's some amount of task complexity. You don't have to use 100% of your brain all the time.
Speaker 3
Right.
Sholto Douglas
Welcome to my world. And so you should be able to run that faster and this kind of stuff, basically. So I think it's like net-net, I think typically the same model. You want to be able to scale the understanding as the complexity and difficulty. Right. You want to be able to do that dynamically.
Dwarkesh Patel
Yeah. Is that variable? So we already have variable compute per answer, right? Right, with like tokens. That's right, yeah. Will we have variable compute per token?
Trenton Bricken
I mean, you can already think of models forever. People have been calling the residual stream and multiple layers like poor man's adaptive compute, right? Where like if the model already knows the answer to something, it will compute that in the first few layers and then just pass it through.
Sholto Douglas
So, yeah.
Trenton Bricken
I mean, that's getting into the weeds. Right. Yeah.
Sholto Douglas
Those are just like this operating ramp. You're doing stuff to it. Right. It's like the mental model I think one takes away from interoperability work.
Dwarkesh Patel
U.S. high school immigration is a broken system that causes some of the most talented people in the world. But I didn't realize before working with Lighthouse how different the process can be if you're working with somebody who knows their way around the system. I hired somebody earlier this year, and even before the remote work trial had ended, Lighthouse had already secured a no-win visa for him. Honestly, it was shockingly fast. My family and I have had a terrible experience with the immigration system, and I've also seen many of my smartest friends get their entire careers hamstrung by its vagaries.
Seeing Lighthouse operate showed me that the visa process can be done in weeks and doesn't have to drag on for months and months. And they do it not only for complex visas like the ONA, but for other types as well. In the last 12 months alone, they have secured visas for over 350 people for companies like Cursor, Notion, Ramp, Replit, and many more. Unlike legacy law firms, Lighthouse specializes in frontier industries like AI, robotics, and biotech. And since they understand your problems, you can trust them to fully handle the paperwork and process for you. Explore which visa is right for you at LighthouseHQ.
com. All right, back to Trenton and Chilto. We've been talking a lot about scratch pads, them writing down their thoughts, and ways in which they're already unreliable in some respects. Mm-hmm. Daniel's AI 2027 scenario kind of goes off the rails when these models start thinking in neural ease. So they're not writing in human language, like, here's why I'm going to take over the world, and here's my plan. They're thinking in the lane space. And because of their advantages in communicating with each other in this deeply textured, nuanced language that humans can't understand, they're able to coordinate in ways we can't.
Is this the path for future models? Are they going to be in neural ease communicating with themselves or with each other?
Sholto Douglas
There's a surprisingly strong bias so far towards token syntax. It seems to work very well. There already is some amount of neural ease. If you think about the residual stream for each token is neural ease to some degree. And so now we're just trading off axes. How much neural ease are you doing versus how much actually is read out to tokens all the time. I think it's important to delineate between the model's planning and latent space in a single forward pass and the model has an alien language that it's outputting
Trenton Bricken
and using as its scratchpad. Which one are we talking about?
Dwarkesh Patel
The latter. Okay. Although it is interesting to note that like there's also already alien stuff happening. I guess I never saw alien so much.
Trenton Bricken
No, but in the most extreme case, it invents a new language that's super information dense or something.
Dwarkesh Patel
Or I guess this is a debate we've had, but like to some extent humans also have a mental ease, right? They're like churning away. Yeah. There's a sense when you're writing something down of like, I know what I'm trying to say. Yeah. But I can't like put it into tokens.
Trenton Bricken
Yeah. I mean, that's what's so fun about the, if you look at the assistant tag, right? Seeing these features light up in the auditing game for the model being evil. Yeah.
Speaker 3
Yeah. Or like Transluse has another example of this where you ask a llama model,
Trenton Bricken
who is Nicholas Carlini? And background context, Nicholas Carlini is a researcher who actually was at DeepMind and has now come over to Anthropic. But the model says, oh, I don't know who that is. I couldn't possibly speculate. But if you look at the features behind the scenes, you see a bunch light up for AI, computer security, all the things that Nicholas Carlini does. Yeah. Interpretability becomes dramatically more important as you shift in this direction
Sholto Douglas
of neural ease.
Trenton Bricken
But is that, are we going to? It seems, I mean, it's an empirical question. Yeah. I think it's somewhat likely, if only because inference is expensive. Producing tokens is expensive. And so there will be an incentive to, one, use as little thinking as you need to give the answer. And two, if you're going to use thinking, use some complex compression. I wonder if it will emerge more once we allow agents to talk to each other in ways where currently it's kind of trained more in isolation or with a human.
Sholto Douglas
And there'll be, like, some selective pressure against it so long as the agents are working with humans because they'll want to sort of cooperate. But then, like, as agents begin to work more and more with each other, then that selective pressure, like, changes the other direction, basically.
Dwarkesh Patel
Although somebody would still have to make the consciousness to do, like, end-to-end training for multiple agents to use the system of communication, right? Sure.
Trenton Bricken
Yeah. I mean, one scary thing, though, is the way we render text, you can use hidden whitespace tokens that also encode information. That's true. And so you can imagine a world where it looks like the agent's reasoning in its scratchpad harmlessly, but it's actually hiding a bunch of data.
Dwarkesh Patel
Speaking of inference compute, I guess one thing that I think is not talked about enough is if you do live in the world that you're painting, that in a year or two, we have computer use agents that are doing like... actual jobs, you've totally automated large parts of software engineering, then these models are going to be incredibly valuable to use. And the way you use them, obviously, is you need compute. Right now, there's 10 million H100 equivalents in the world. By 2028, there's going to be 100 million. But there's been estimates that an H100 has the same amount of flops as the human brain.
And so if you just like do a very rough calculation, it's like there's a 10 million population. If you get AGI that's like as human inference efficient, you could have 10 million AGI's now, 100 million AGI's in 2028. But – presumably you would want more. And then at that point, your AI computer is increasing, what, 2.5x or 2.25x every year right now. But at some point, like 2028, you hit wafer production limits, and that's a longer feedback loop before you can make new fabs or whatever. The question here is, are we sort of underwriting how big a bottleneck inference will be if we live in the kind of world you're painting, if we have the capabilities that you're describing?
Sholto Douglas
I'd want to do the math on exactly how much, like, we can ramp up TSMC's production and this kind of stuff. Like, what fraction of the supply chain at the moment – we need Dylan in here for this. But, like, is currently GPUs. Like, relatively small.
Trenton Bricken
5%.
Sholto Douglas
5% or something like this. Yeah. Like, Apple has a huge fraction. Yeah, yeah. And, like, are the 2028 estimates including, like, that ramping up over time? Yeah, yeah. To what? Like, 20%, 30%? Or, like – This is just up AI 2070s.
Dwarkesh Patel
Yeah, yeah, yeah. But – I assume it's saturated at that point. Is that why they expect it to then just go at the...
Sholto Douglas
I do think this is underrated to some degree. To the extent that you don't instantly get a doubling of the world's population in 2028. You maybe get... tens of millions of geniuses in a data center, but you don't get a doubling of the world's population. And so a lot depends on exactly how smart they are, exactly how efficient the models are at thinking about this kind of stuff. Let's do some rough math, I guess, to factor the H100 thing. You could probably run a 100-bit model, do about 1,000 tokens or something like that on H100. If, like, we're comparing that to a number of, should we compare that numbers for a minute? No, okay. 1,000 tokens a second. Humans are what? How fast can a human talk?
Dwarkesh Patel
There is a really interesting paper. I don't know if you saw this. Humans think at 10 seconds a second.
Sholto Douglas
Did you see this paper? 10 seconds a second. No.
Dwarkesh Patel
There was this really interesting paper about, if you look at the amount of, like, information we're processing in a second. Yeah. We're seeing all this visual data, et cetera, et cetera. But by a bunch of metrics where you think about how fast humans are processing, it's at 10 seconds a second. You will have people fly over France or something. Even these so-called idiot savants who will remember everything. If you think about how long their plane ride was, it's like 45 minutes. If you do 10 tokens a second, how much information would you have? It's literally exactly that. So let's take that for granted.
Sholto Douglas
Then it's like a nature hunter is 100 humans a second.
Dwarkesh Patel
Yeah, if you think the tokens are equivalent, yeah.
Sholto Douglas
If you think the tokens are equivalent. Yeah. Which you still get pretty substantial numbers, like even with your 100 million H100s and you multiply that by 100, you're starting to get to pretty substantial numbers. This does mean that those models themselves will be somewhat compute-bound in many respects, but these are all like These are relatively short-term changes in timelines of progress, basically. I think, yes, it's highly likely we get dramatically inference bottlenecked in 27, 28. The impulse to that will then be, okay, let's just try and turn out as many possible semiconductors as we can. There'll be some lag there. A big part of how fast we can do that will depend on how much people are feeling the AGI in the next two years as they're building out fab capacity. A lot will depend on how China and Taiwan, how is the Taiwan situation? Is Taiwan still producing all the fabs in the chips?
Dwarkesh Patel
There's another dynamic which was a reason that Ege and Tame, when they were on the podcast, said that they were pessimistic, is that, one, they think we're further away from solving these problems with long context, coherent agency, advanced multimodality than you think. And then their point is that the progress that's happened in the past over reasoning or something has required many orders of magnitude increase in compute. If this scale of compute increase can continue beyond 2030, not just because of chips, but also because of power and like raw GDP even, then because we don't think we get it by 2030 or 2028 by just – then we think it's just going to take – the probability per year just goes down a bunch.
Sholto Douglas
Yeah, this is like bimodal distribution. Yeah. A conversation I had with Leopold turned into a section in a situation where it's called This Decade or Bust, which is on exactly this topic, which is basically that for the next couple of years, we can dramatically increase our training compute. And RL is going to be so exciting this year because we can dramatically increase the amount of compute that we apply to it. And this is also one of the reasons why the gap between, say, DeepSeq and 01 was so close at the beginning of the year because they were able to apply the same amount of compute to the RL process. And so that compute differential actually will sort of be magnified over the course
Trenton Bricken
of this year. I mean, bringing it back to the there's so much low-hanging fruit, it's been wild seeing the efficiency gains that these models have experienced over the last two years. And yeah, like with respect to DeepSeq, I mean, just really hammering home and like Dario has a nice essay on this. It's good, yeah. Deep Seek was nine months after Claude III's sonnet. And if we retrained the same model today or at the same time as the Deep Seek work, we also could have trained it for $5 million or whatever the advertised amount was. And so what's impressive or surprising is that Deep Seek has gotten to the frontier. But I think there's a common misconception still that they are above and beyond the frontier. And I don't think that's right. I think they just waited. And then we're able to take advantage of all the efficiency gains that everyone else was also seeing.
Sholto Douglas
Yeah. They're exactly on the sort of cost curve that you'd expect. We're not going to take away from the brilliant engineers and brilliant researchers who are like, I look at their work and I'm like, ah, like the kindred soul there in the work they're doing. And to go from like way behind the frontier to like, oh, this is like a real player.
Dwarkesh Patel
It's super incredible work.
Trenton Bricken
Yeah.
Dwarkesh Patel
So people say that they have good research taste. Looking at their papers, what makes you say that?
Sholto Douglas
Yeah, I think their research taste is good in a way that I think like Noam's research taste is good. Noam Brown? Noam Shizia. Noam Brown also has a good research taste, but Noam Shizia. Where they very clearly understand this dance between the hardware systems that you're like designing the models around and the sort of like algorithmic side of it. And this is manifest in the way that The models give this sense of being perfectly designed up to their constraints. And you can really very clearly see what constraints they're thinking about as they're iteratively solving these problems. And so, I mean, let's take the base transformer and diff that to deep seek v2 and v3.
You can see them running up against the memory bandwidth bottleneck in attention. And you can see them initially they do MLA to do this. They trade flops for memory bandwidth, basically. And then they do this thing called NSA where they, like, more selectively load memory bandwidth. And you can see, actually, this is because the model that they trained with MLA was on H800s. So it has a lot of flops. And so they were like, OK, we can freely use the flops. But then the export controls from Biden came in. Or they knew they would have less of those chips going forward. And so they traded off to a more memory bandwidth-oriented algorithmic solution there.
And you see a similar thing with their approach to sparsity, where they're iteratively working out the best way to do this over multiple papers. And the part that I like is that it's simple. A big failure mode that a lot of ML researchers have is you do these overly complicated things that don't like think hard enough about the hardware systems that you have in mind. Whereas the first DeepSeq like sparsity MOE solution, they designed these like rack and like node level load balancing losses. So you can see them being like, okay, like perfectly balanced on this. And then they actually come up with a much better solution later on where they don't have to have the auxiliary loss.
where they just have these, like, bias terms that they put in. And it's cool.
Dwarkesh Patel
Isn't that less simple? Like, you're manually putting in a bias rather than...
Sholto Douglas
But balancing the auxiliary losses and wing. Like, you're making the model, like, trade off this thing. And, like, you have to... With auxiliary losses, you have to, like, control the coefficient and the weighting. The bias is, like, cleaner in some respects. Interesting.
Dwarkesh Patel
Did they have to change it through training?
Sholto Douglas
They did have to change it during training. Does all training involve continuously fucking with these values as you're going through it? It depends on what your architecture is. I thought it was cute that you can see them running up into this very hardware-level constraint. They tried to go, what do we wish we could express algorithmically? What can we express under our constraints? And iteratively solving to get better constraints. And doing this in a really simple and elegant way, and then backing it up with great engineering. I also thought it was interesting that they incorporated the multi-token prediction thing from Meta. So Meta had a nice paper on this multi-token prediction thing. I don't know if it's good or bad, but Meta didn't include it in Llama, but DeepSea did include it in their paper, which I think is interesting. Was that because they were faster at iterating and including an algorithm, or did Meta decide that actually it wasn't a good algorithmic change at scale? I don't know. It was really interesting to me as somebody who's had people on the podcast to
Dwarkesh Patel
discuss the... I mean, it's interesting from what's happening in AI right now, but also from the perspective of... I've been having abstract conversations with people about what an intelligence explosion would look like, or what would it look like for AI to automate ARND, and just getting a more tangible sense of what's involved in making this AI progress. And I guess one of the questions I was debating with Daniel is, How much or I was asking him is how many of the improvements require a deep conceptual understanding versus how many are just like monkeys trying ideas and you could just like run a bunch in parallel. And it seems like the MLA thing is motivated by this like conceptual understanding of like, oh, each attention head only needs to see like the subspace that's relevant to its attention pattern. I feel like that just required a lot of conceptual insight in a way that these models are especially bad at. As opposed to, I don't know how the load balancing thing works, but that just seems like maybe you could try it out and see what happens. Yeah, that's probably just like I'm trying out a whole bunch of different things. So what fraction is which I'd be curious about.
Trenton Bricken
Yeah, I don't know about fractions. It might be like you have a hunch for a core problem. You can think of 10 possible ways to solve it. And then you just need to try them and see what works. And that's kind of where the trial and error like sorcery of deep learning can kind of kick in. And like Noam Shazia will talk about this, like about how he like 5% of his ideas work.
Sholto Douglas
So even he, the vaunted god of model architecture design, has a relatively low hit rate, but he just tries so many things. Right.
Dwarkesh Patel
Or being able to come up with any ideas in the first place. So one mechanism could be that Gnome just doesn't have to do any of the engineering work, and he can just abstractly express an intuition.
Sholto Douglas
Yeah. I actually think your rates of progress almost don't change that much, so long as he's able to completely implement his ideas. Say more? If you have like Noam Shazir at 100x speed, that's still kind of wild. There's all these like fallbacks of like wild worlds where even if you don't get like 100% like Noam Shazir level intuition in model design, it's still okay if you just accelerate him by 100x. Right.
Dwarkesh Patel
Especially since your compute bottlenecked anyway, so like trying out his ideas, or I guess he doesn't have the compute to try out all of his ideas.
Trenton Bricken
But Dworkesh, you said, oh, well, the model can do the more straightforward things and not the deeper thought. I mean, I do want to push back on that a little bit. Like, I think, again, if the model has the right context and scaffolding, it's starting to be able to do some really interesting things. The interp agent has been a surprise to people, even internally, at how good it is at finding the needle in the haystack, like when it plays the auditing game, finding this reward model bias feature, and then reasoning about it, and then systematically testing its hypotheses. So it looks at that feature. Then it looks at similar features. It finds one with a preference for chocolate. It's like, huh, that's really weird that the model wants to add chocolate to recipes. Let me test it. And so then it will make up like, hey, I'm trying to make a tomato soup. What would be a good ingredient for it? And then sees that the model replies chocolate. It reasons through it and then keeps going, right?
Sholto Douglas
There is conceptual understanding there. Deep conceptual understanding. And even where like especially it's spotted, it's like, oh, this is a key part of its persona.
Trenton Bricken
I see this Oxford paper. What if I change Oxford to Stanford? What if I now say Richard Feynman really likes this thing? And it's like really carving out the hypothesis space and testing things in a way that I'm kind of surprised by.
Sholto Douglas
Also, by the way, ML research is like one of the easier things to RL on in some respects once you get to a certain level of capability. It's a very well-defined objective function. Did the loss go down?
Trenton Bricken
Make number go down.
Sholto Douglas
Make number go down. Or make number go up, depending on which number it is.
Trenton Bricken
Just flip the sign.
Sholto Douglas
Just flip the sign. And so once you get to the stage where models are capable of implementing one of no one's ideas, then you can just let them loose and let them build that intuition of how to do scientific discovery.
Trenton Bricken
Right.
Sholto Douglas
The key thing here, again, is the feedback loops. I expect scientific areas where you are able to put it in a feedback loop to have, eventually, superhuman performance. One prediction I have is that we're going to move away from can an agent do XYZ and
Trenton Bricken
more towards can I efficiently deploy, launch 100 agents and then give them the feedback they need and even just be able to like easily verify what they're up to, right? There's this generator verify gap that people talk about where it's like much easier to check something than it is to produce the solution on your own. Yeah. It's very plausible to me will be at the point where it's so easy to generate with these agents that the bottleneck is actually can I as the human verify the answer? And again, you're guaranteed to get an answer with these things. And so ideally, you have some automated way to evaluate and test a score for like how well it worked. How well did this thing generalize? Mm-hmm. And at a minimum, you have a way to easily summarize what a bunch of agents are finding. And it's like, okay, well, if 20 of my 100 agents all found this one thing, then like it has a higher chance of being true. And again, software engineering is going to be the leading indicator of that, right?
Sholto Douglas
Like over the next six months, like the remainder of the year, basically we're going to see progressively more and more experiments of the form of how can I dispatch work to a software engineering agent in such a way as async. Clawed for is GitHub integration, where you can ask it to do things on GitHub, ask it to do pull requests, this kind of stuff that's coming up. And the OpenAI's codecs are examples of this, basically. You can sort of almost see this in the coding startups. I think of this as product exponential in some respects, where you need to be designing for a few months ahead of the model to make sure that the product you build is the right one.
And you saw last year, Cursor hit PMF with Claw 3.5 Sonnet. They were around for a while before, but then the model was finally good enough that the vision they had of how people would program hit. And then, Windsurf bet a little bit more aggressively even on the agenticness of the model. With longer running agentic workflows and this kind of stuff. And I think that's when they began competing with Cursors, when they bet on that particular vision. And the next one is you're not even in the loop, so to speak. You're not in an IDE, but you're asking the model to go do work in the same way that you would ask someone on your team to go do work.
That is not quite ready yet. There are still a lot of tasks where you need to be in the loop. But the next six months looks like an exploration of exactly what does that trend line look like.
Trenton Bricken
Yeah. But just to be really concrete or pedantic about the bottlenecks here, a lot of it is, again, just tooling and are the pipes connected. A lot of things I can't just launch Claude and have it go and solve because maybe it needs a GPU. Or maybe I need very careful permissioning so that it can't just take over an entire cluster and launch a whole bunch of things, right? So you really do need good sandboxing and the ability to use all of the tools that are necessary.
Sholto Douglas
And we're almost certainly under-eliciting dramatically. When you look at Mita's evals of can the model solve the task, they're there solving them for hours. over like multiple iterations. And eventually one of them is like, oh yeah, I've come back and I've solved the task. Me at the moment, at least like maybe the fault is my own, but I try the model and doing something and if it can't do it, I'm like, okay, fine, I'll do it.
Dwarkesh Patel
Which is interesting because we don't even treat other humans this way. If you hire a new employee, you're not like, oh, I'll do it. Yeah, yeah, yeah. You spend literally weeks giving them feedback where we'll give up with the model in minutes. Yes, exactly.
Trenton Bricken
But I think part of it is, is it async or not? Yes. And if it's human in the loop, then it's so much more effortful. And unless it's getting, that's applying immediately. I've noticed if I don't have a second monitor with Cloud Code always open in the second monitor, I won't really use it. Yeah, yeah. It's only when it's right there and I can send off something. If it hits, great. If not, I'm kind of working on it at the same time.
Sholto Douglas
But this more async form factor, I expect to, like, really quite dramatically improve the experience of these models. Interesting. Or you can just say, like, let's see if it can do that.
Dwarkesh Patel
Yeah. Just give it a whirl. Try 10 different approaches. Yeah, yeah. Just fire it off. Yeah, fire it off. Before we end this episode, I do want to get back at this crux of... why does the progress that you're talking about in computer use agents and white-collar work happen over the next few years? Why is this not a thing that takes decades? And I think the crux comes down to the people who expect something much longer have a sense that... When Andrew Duggan and Tommy on my podcast, they were like, look... You could look at AlphaGo and say, oh, this is a model that can do exploration. AlphaZero can generalize to new video games. It has all these priors about how to engage with the world and so forth.
Sholto Douglas
The intellectual ceiling is really high.
Dwarkesh Patel
Yeah, exactly. And then, in retrospect, obviously a bunch of the methods are still used today in deep learning. And you can see... see similar things in the models that we train today, but it was fundamentally not a sort of baby AGI that we just had to add a little sprinkle of something else on top of in order to make it the LLMs of today. I just want to very directly address this crux of why are LLMs in a much different position with respect to true AGI than AlphaZero? Why are they actually the base on which adding in a few extra drops of this kind of care and attention gets us to human-level intelligence?
Sholto Douglas
I think one important point is that When you look at alpha zero, it does have all of those ingredients. And in particular, I think the intellectual ceiling goes quite contra what I was saying before, which is we've demonstrated this incredible complexity of math and programming problems. I do think that the type of task and setting that AlphaZero worked in, this two-player perfect information game, basically, is incredibly friendly to RL algorithms. And the reason it took so long to get to like a more AGI, proto AGI style models is you do need to crack that like general conceptual understanding of like the world and language and this kind of stuff. And you need to get the initial reward signal on tasks that you care about in the real world, which are like harder to specify than games. And I think then that like that sort of like gradient signal that comes from the real world, like all of a sudden you get access to it and you can start climbing it. whereas AlphaZero didn't ever have like the first rung to pull on.
Trenton Bricken
Yeah. This goes back to the monkeys on the typewriter, I think, and like the pre-training model. And until you had something like GPT-3, GPT-4, it just couldn't generate coherent enough sentences to even begin to do RLHF and tell it what you liked and didn't like.
Dwarkesh Patel
Yeah. If we don't have even reasonably robust... or weakly robust computer use agents by this time next year. Are we living in the bust timeline as in 2030 or bust?
Sholto Douglas
I would be extremely surprised if that was the case. And I think that would be like somewhat of an update towards like there's something like strangely difficult about this, like computer use in particular. I don't know if it's the bust timeline, but it's definitely like the I would update on this being like a lengthening of time. Yeah, yeah. Yeah, I mean, I think more and more it's no longer a question of speculation.
Trenton Bricken
If people are skeptical, I'd encourage like using Claude code or like some agentic tool like it and just seeing what the current level of capabilities are. Tweeting is so much easier. But seriously, like, the models are getting really capable at tasks that we care about and we can give them enough data for. And, I mean, the circuit's results from interpretability are also pointing in the direction that they're doing very reasonable, generalizable things. And so, yeah, this question matters a lot. But... I'm surprised by how many deep learning critics just like haven't really interacted with the models or haven't in a while.
Sholto Douglas
And constantly move the gold fist.
Trenton Bricken
Yeah, yeah, yeah. Like the Turing test used to be a thing, right? Like we don't even talk about it and it'd be like silly to think that it was a meaningful test. Yeah, yeah. Now, that being said, one caveat on that is if software engineering is just dramatically better than
Sholto Douglas
computer use, and computer use still sucks, then I'd be still like, oh, maybe everyone just kept focusing on software engineering. It was just by far the most valuable thing. Every marginal person and dollar went towards software engineering. I don't think that's the case. I do think computer use is valuable enough that people will care about it. But that's my one escape patch that I'm putting in place for next year.
Dwarkesh Patel
It would be good from an alignment perspective, too. Because I think you kind of do need a wider range of skills before you can do something super scary.
Sholto Douglas
Oh, like as in if the models didn't get any better? Yeah, if it's like they're superhuman coders, but they're not like Henry Kissinger level.
Trenton Bricken
I don't know. That seems okay. Like if we have AI oracles. Yeah, that's what I'm saying. That's good.
Dwarkesh Patel
Yeah, exactly. So if you look back at AI discourse, like going back a decade, there's a sense that There's dumb AI, then there's AGI, then there's ASI, that intelligence is the scalar value. The way you've been talking about these models has a sense of jaggedness. It's especially tuned to environments in which it's been trained a lot or has a lot of data. Is there a sense in which it still makes sense to talk about the general intelligence of these models? Is there enough meta-learning and transfer learning that is distinguished between the sizes of models or the way models are trained? Or are we moving into a regime where... It's not about intelligence. It's more so about domain.
Sholto Douglas
Yeah. So one intuition pump is this conversation was had a lot when models were like GPT-2 sized and fine-tuned for various things. And they found – people would find that the models were dramatically better at things that they were fine-tuned for, right? But by the time you get to GPT-4, when it's trained on a wide enough variety of things, actually the – like the sort of total compute, like it generalized very well across all of the individual sub tasks and actually generalized better than smaller fine-tuned models in a way that was extremely useful. I think right now what we're seeing with RL is pretty much the same story playing out where there's this jaggedness of like things that they're particularly trained at, but as we expand the total amount of compute that we do RL with, you'll start to see the same transition from GPT-2 fine-tunes to GPT-3, GPT-4 unsupervised meta-learning and generalization across things. And I think we're already seeing early evidence of this in its ability to generalize reasoning to things. I think this will be extremely obvious.
Trenton Bricken
One nice example of this is just the ability or notion to backtrack. You go down one solution path. Ooh, wait, let me try another one. And this is something that you start to see emerge in the models through RL training on harder tasks. And I think right now it's not generalizing incredibly well, at least with...
Sholto Douglas
Well, I mean, have we ever RLed the model to be an interp agent? No. I mean, no. Yeah, exactly.
Trenton Bricken
Yeah. So all this time we're talking about, like, oh, it's only good at things as being RLed.
Sholto Douglas
Well, it's pretty good at that because that is a mixture of, like, science and, like, understanding language and coding. Like, there's this sort of mixture of domains here, all of which you need to understand. Like, you need to be both a great software engineer and be able to think through language and state of mind and almost philosophize in some respects to be an interp agent. And it is generalizing from the training to do that.
Dwarkesh Patel
What's the endgame here? Claude 8 comes out, and they give it to you, and dot, dot, dot, you say thumbs up. What's happened?
Trenton Bricken
Yeah. I mean, it really depends upon the timeline at which we get Claude 8 and the models hit ASL 4 capabilities, right? Fundamentally, we're just going to use whatever tools we have at the time and see how well they work. Ideally, we have this enumerative safety case where we can almost verify or prove that the model will behave in particular ways. Yeah. In the worst case, we use the current tools like when we won the auditing game of seeing what features are active when the assistant tag lights up.
Dwarkesh Patel
Can you explain what is mechanistic interpretability? What are features? What are circuits? Totally.
Trenton Bricken
Yeah. Yeah, yeah, yeah. So mechanistic interpretability, or the cool kids call it mechanterp, is trying to reverse engineer neural networks. and figure out kind of what the core units of computation are. Lots of people think that because we made neural networks, because they're artificial intelligence, we have a perfect understanding of how they work, and it couldn't be further from the truth. Neural networks, AI models that you use today, are grown, not built. And so we then need to do a lot of work after they're trained to figure out to the best of our abilities how they're actually going about their reasoning.
And so two and a half, three and a half years ago, this kind of agenda of applying mechanistic interpretability to large language models started with Chris Ola leaving OpenAI, co-founding Anthropic. And Every roughly six months since then, we've had kind of like a major breakthrough in our understanding of these models. And so first with Toy Models of Superposition, we established that models are really trying to cram as much information as they possibly can into their weights. And this goes directly against people saying that neural networks are overparameterized. And like classic AI machine learning back in the day, you would use linear regression or something like it.
And people had a meme of AI or neural networks, deep learning, using way too many parameters. There's like this funny meme that you should show of like layers on the x-axis and layers on the y-axis and this like jiggly line that just goes up. But it actually turns out that at least for really hard tasks like being able to accurately predict the next token for the entire internet, these models just don't have enough capacity. And so they need to cram in as much as they can. And the way they learn to do that is to use each of their neurons or units of computation in the model for lots of different things.
And so if you try to make sense of the model and be like, oh, if I remove this one neuron or like, what is it doing in the model? It's impossible to make sense of it. It'll fire for like Chinese and fishing and horses. And I don't know, just like 100 different things. And it's because it's trying to juggle all these tasks and use the same neuron to do it. So that's superposition. Nine months later, we write towards monosemanticity, which introduces what are called sparse autoencoders. And so going off what I just said of the model trying to cram too much into too little space, we give it more space, this higher dimensional representation, where it can then more cleanly represent all of the concepts that it's understanding.
And this was a very toy paper in so much as it was a two-layer, really small, really dumb transformer. And we fit up to, I want to say, 16,000 features, which we thought was a ton at the time. Fast forward nine months, we go from a two-layer transformer to our Claude III Sonnet frontier model at the time and fit up to 30 million features. And this is where we start to find really interesting abstract concepts like a feature that would fire for code vulnerabilities. And it wouldn't just fire for code vulnerabilities. It would even fire for, like, you know that Chrome page you get if you, like, It's not an HTTPS URL.
And it's like warning, this site might be dangerous, like click to continue. And like also fire for that, for example. And so it's like these much more abstract coding variables or sentiment features amongst the 30 million. Fast forward nine months from that, and now we have circuits. And I threw in the analogy earlier of the Ocean11 heist team, where now you're identifying individual features across the layers of the model that are all working together to perform some complicated task. And you can get a much better idea of how it's actually doing the reasoning and coming to decisions, like with the medical diagnostics.
One example I didn't talk about before is with like how the model retrieves facts. And so you say like, what sport did Michael Jordan play? And not only can you see it hop from like Michael Jordan to basketball, answer basketball, but the model also has an awareness of when it doesn't know the answer to a fact. And so by default, it will actually say, I don't know the answer to this question. But if it sees something that it does know the answer to, it will inhibit the I don't know circuit and then reply with the circuit that it actually has the answer to. So for example, if you ask it who is Michael Batkin, which is just a made up fictional person, it will by default just say I don't know.
It's only with Michael Jordan or someone else that it will then inhibit the I don't know circuit. But what's really interesting here and where you can start making downstream predictions or reasoning about the model is that that I don't know circuit is only on the name of the person. And so in the paper, we also ask it, what paper did Andre Karpathy write? And so it recognizes the name Andre Karpathy because he's sufficiently famous. So that turns off the I don't know reply. But then when it comes time for the model to say what paper it worked on, it doesn't actually know any of his papers. And so then it needs to make something up.
And so you can see different components and different circuits all interacting at the same time to lead to this final answer. Why I think it's a tractable problem to understand every single thing that's
Dwarkesh Patel
happening in a model, or that's the best way to understand why it's being deceptive. If you wanted to explain why England won World War II using particle physics, you would just be on the wrong track. You just want to look at the high-level explanations of... Who had more weapons? Like, what did they want? And that seems analogous to just training linear probes for like, are you honest? Are you being deceptive? Like, do we catch you doing bad things when we're red teaming you? Can we monitor you? Why is this not analogous where we're asking a particle physicist to just backtrack and explain why England won World War II?
Trenton Bricken
I feel like you just want to go in with your eyes wide open, not making any assumptions for what that deception is going to look like. or what the trigger might be. And so the wider you can cast that net, the better. Depending on how quickly AI accelerates and where the state of our tools are, we might not be in the place where we can, like, prove from the ground up that everything is safe. But I feel like that's a very good North Star. It's a very powerful, reassuring North Star for us to aim for, especially when we consider we are part of the broader AI safety portfolio. I mean, do you really trust, like, you're about to deploy this system and you really hope it's aligned with humanity and that you've, like, successfully iterated through all the possible ways that it's going to, like, scheme or sandbag.
Dwarkesh Patel
But that's also probably going to be true with whatever you find. I mean, you're still going to have... variants that you haven't explained or like you found a feature, but you don't know if it actually explains deception or something else instead? So I guess, first of all, I'm not saying you shouldn't try the probing approach, right?
Trenton Bricken
Like we want to pursue the entire portfolio. We've got the therapist interrogating the patient by asking, do you have any troubling thoughts? We've got the linear probe, which I'd analogize to like a polygraph test where we're taking like Very high-level summary statistics of the person's well-being. And then we've got the neurosurgeons kind of going in and seeing if you can find any brain components that are activating and troubling or off-distribution ways. So I think we should do all of it. What percent of the alignment portfolio should McIntor be? uh i think as much of a chunk as is necessary i mean i i i think at least like yeah hard hard to hard to define but i don't know at anthropic i feel like all of the different portfolios are like being very well supported and and growing um you can also come back to like the the world war ii question you can think of it as like a
Sholto Douglas
hierarchy of abstractions of trust here Where like let's say you want to go and talk to like Churchill. It helps a lot if you can verify that in that conversation, in that 10 minutes, he's being honest. And this like enables you to construct better meta narratives of what's going on. And so maybe particle physics wouldn't help you there. But certainly like the neuroscience of Churchill's brain. would help you verify that he was being trustworthy in that conversation and that the soldiers on the front lines were being honest in their description of what happened and this kind of stuff. So long as you can verify parts of the tree up, then that massively helps you build confidence.
Trenton Bricken
I think language models are also just really weird, right? With the emergent misalignment work, I don't know if they took predictions they should have of like, hey, I'm going to fine-tune ChatGPT on code vulnerabilities. Is it going to become a Nazi? And I think most people would have said no. And that's what happened. And so what are the different – How did they discover that it became a Nazi? they started asking it a ton of different questions. And it will do all sorts of vile and harmful things. The whole persona just totally changes. And we are dealing with alien brains here who don't have the social norms of humans or even a clear notion of what they have and haven't learned that we have of them, I mean. And so I think you really want to go into this with eyes wide open. Backing up for Mechinterp, if you live in a world where AI progress accelerates,
Dwarkesh Patel
By the way, you were mentioning a little while ago that there's many wild worlds we could be living in, but we're living in at least one of them. Another one that we've gestured at, but it's worth making more explicit, is this. Even if the AI models are not helping write the next training algorithm for their successor, just the fact that if they had human-level learning efficiency, whatever a model is learning on the job – Or whatever copy of the model is learning on the job, the whole model is learning. So in effect, it's getting— Or if they're like a thousand times less efficient than humans are
Sholto Douglas
learning. That's right. And you just deployed them even still.
Dwarkesh Patel
Exactly. Yeah, yeah. Anyways, and there's a whole bunch of other things you can think about. But even there, it's like you kind of have a broadly deployed intelligence explosion.
Sholto Douglas
I do think it's worth pressing on that future of ... There is this whole spectrum of crazy futures, but the one that I feel we're almost guaranteed to get, and this is almost a strong statement to me, is one where at the very least, you get drop-in white-collar worker at some point in the next five years. I think it's very likely in two, but it seems almost over-determined in five. And on the grand scheme of things, those are irrelevant time frames. It's the same either way. completely changes the world over the next decade. And if we don't have the right policies in place for that, then you end up actually with almost, in some respects, a fundamentally worse world.
Because the thing that these models get good at by default is software engineering and computer-using agents and this kind of stuff. And then we will need to put in extra effort to put them in the loops where they help us with scientific research or they're, uh, like we have the right robotics such that we actually like experience an increase in material quality of life. Um, so that's worth thinking about. Like if you're in the perspective of like, I'm a country, like what should I be doing or thinking about, uh, like plan for the case where, where white collar work is automatable. Um, And then consider what does that mean for your economy and what you should be doing to prepare policy.
Dwarkesh Patel
What should you be doing to prepare? Because honestly, it's such a tough question where like if you're India or Nigeria or Australia, if you're a country unlike America or China where they do have frontier models, What is it that you should be doing right now, especially on such a short timescale? MARK MIRCHANDANI- Yes.
Sholto Douglas
So I think one very important point is that, let's say this scenario turns out true, then compute becomes the most valuable resource in the world. Like the sort of GDP of your economy is dramatically affected by how much compute you can deploy towards the sort of organizations within your country. And so having some guaranteed amount of compute, I think, will actually be quite important. So getting ahead of investments in data centers and this kind of stuff on the condition that companies in your country have to be allowed to use that compute. Not necessarily for training, but even just for inference.
I think the economic value here comes from inference. I think it also makes sense to invest broadly in AI. I think these countries have the opportunity to do so, and I think that's a portfolio of foundation model companies, but also robotics supply chain and this kind of stuff. I think that you should invest very proactively in policies that try and prevent capital lock-in. So we're in for a much worse world if it just so happens that the people who had money in the stock exchange or in land before AGI are dramatically more wealthy than the people who don't, because it's a gross misallocation of resources.
One of my favorite episodes actually on your podcast was the Georgism one, where you're trying to appropriately value or allocate land. This strikes particularly close to home coming from Australia, where I think our policies with respect to land are grossly wrong. But I think this is broadly true. Being very forward on regulation of integration of these models into your country is important and proactively making sure that people have choice. Let's say you should be quite proactive about making sure that the phones or devices or glasses that people have, people have free choice on what things they run.
And then, so that's the we just get white collar worker, right? And you're trying to do the best to prepare your country for that. And then it's like, OK, well, what can you do to make all possible versions of the future go well? That's covering some amount of economic downside. The other things I think are really important is figure out how you can either make the, basically ensure dramatic upside or cover terrible downside. And so getting a dramatic upside is making sure that there is investment in biology research and this kind of stuff in an automated way that these models are actually able to produce novel medicines that massively improve quality of life.
And covering the downside is AI alignment research and this kind of stuff and automated testing and really thinking hard about that, AI safety institutes, this kind of stuff. But these seem like things that a rich person, a random rich person could also do.
Dwarkesh Patel
Like, it seems like... There's not a thing that a nation state is uniquely equipped to do in this scenario. I mean, dramatic allocation of resource towards compute, I think, is sensible.
Sholto Douglas
I would be doing that if I was in charge of a nation state. I think it just increases your optionality in most of the future worlds.
Trenton Bricken
Dylan Patel has some scary forecasts on U.S. energy. Yeah, versus China. Yes. Yeah, we're like 34 gigawatts off. Yeah, yeah. The U.S.
Sholto Douglas
's line is like flat, basically, and China's line is like this. And I mean, the U.S. like very clearly.
Trenton Bricken
We just need so many more power plants. Yes.
Sholto Douglas
If intelligence becomes this like incredibly valuable input, like intelligence becomes almost a raw input into the economies and quality of life of the future, the thing directly underneath that is energy. And so making sure you have like, you know, incredible amounts of solar, like tile the desert in solar panels, some parts of the desert in solar panels, would be helpful towards making sure that you have more access to intelligence on top.
Trenton Bricken
Yeah. Just to make it explicit, because we've been touching on it here, even if AI progress totally stalls, you think that the models are really spiky and they don't have general intelligence, it's so economically valuable and sufficiently easy to collect data on all of these different jobs, these white collar job tasks, such that to Sholto's point, we should expect to see them automated within the next five years. Yeah. Even if you need to hand spoon every single task to the model.
Sholto Douglas
It's economically worthwhile to do so. Even if algorithmic progress stalls out and we just never figure out how to keep progress going, which I don't think is the case. That hasn't stalled out yet. It seems to be going great. The current suite of algorithms are sufficient to automate white collar work provided you have enough of the right kinds of data. And in a way that like compared to the TAM of salaries for all of those kinds of work is so like trivially worthwhile. Yeah, exactly. I do just want to flag as well that there's a really dystopian future if you take
Trenton Bricken
more of X Paradox to its extreme. Yeah. Yeah. Yeah. Oh, well, and even if you look at like robot hands or like the ability to open a door is still just like really hard for robots. Meanwhile, we're seeing this total automation of coding and everything else that we've seen is clever. The really scary future is one in which AIs can do everything except for the physical robotic tasks, in which case you'll have humans with like AirPods and like… Glasses. glasses and there'll be some robot overlord controlling the human through cameras by just like telling it what to do and like having a bounding box around the thing you're supposed to pick up and so you have like human meat robots um and and not like necessarily saying that like that's what the ais would be like want to do or anything like that but as in like if you were to be like what are the relative economic value of things like the ais are out there doing computer programming and
Sholto Douglas
like the most valuable thing that humans can do is like be amazing robots Now, that being said, I think Moravec's paradox is a little bit fake. I think the main reason that robots are worse at being a robot than they are at software engineering is the internet exists for software engineering, like GitHub exists, and there is no equivalent thing. all mocap of everyone's actions as they were going about their daily lives for some reasonable fraction of the human population, robotics is also close to solved. Like on track to be solved at the same rate that software engineering is on track to be solved.
So this is only like This vision is only a decade-long section, but it's still a pretty terrible decade. Imagine the world where people have lost their jobs. You haven't yet got novel biological research that means people's quality of life is dramatically better. You don't yet have material abundance because you haven't actually been able to action the physical world in the necessary way. You can't build dramatically more because building dramatically more takes robots, basically. And people's like main comparative advantage is as fantastic robots is like a shocking, shocking world. For the perspective of an average human, I think it actually might be better.
Your wages will be higher because you're the complement to something that is
Dwarkesh Patel
enormously valuable, which is AI labor. Right.
Sholto Douglas
And a decade or two on, the world is fantastic, right? Robotics is solved and you decide to get radical abundance, basically, provided that you have all the policies set up necessary to permit building. You end up with that same change from the before and after photos of Shanghai We're like 20 years on. It's like this dramatically transformed city. Like a lot of places in the world probably end up like that over that two decade period. But we need to make sure like one, do our best to estimate is this actually what is on track to happen? Like build sweet bench, but for all the other forms of white collar work and measure and track.
That's a great thing that government should be doing, by the way, is like trying to break down the sort of functions of their economy into measurable tasks and figuring out where, what does the curve actually look like for that? Because they might be a bit shocked by the progress there. You know, there's no sweet bench for tax, like tax eval. And then... I don't have all the answers here, but figuring out a way to share the proceeds of this economy broadly across people or invest heavily in robotics and collecting data so that we get robotics faster and we get material abundance faster, invest in biological research that we get all that faster.
Basically try and pull forward the radical upside because otherwise you have a pretty dark section. I think one thing that's not appreciated enough is how much of our leverage on the
Dwarkesh Patel
future, given the fact that our labor isn't going to be worth that much, comes from our economic and political system surviving. For your million X S&P equity to mean something, for your contracts to mean anything, for the government to be able to tax the AI labor and give you a UBI off of that, It just like that requires our legal institutions, our economic institutions, our financial rail surviving in the future. Yes. The way in which that likely happens is if it's also in the AI's best interests that they follow those rails. And by AI, I don't mean some monolithic single AI. I just mean like firms which are employing AI and becoming more productive as a result. Yeah. You don't want to be in a position where it's so onerous to operate in our system that you're basically selecting for firms who either emigrate or who are doing black market stuff, et cetera. Which means, I think, you want to make it super, super easy to deploy AI, have the equivalent of special economic zones, et cetera. Because otherwise, you are just surrendering the future outside of any control that you might have on
Trenton Bricken
it.
Dwarkesh Patel
One of the reasons, by the way, that I worry about turning AGI into a national security issue or having it have extremely close ties with the government, the Manhattan Project thing, is that it disproportionately redirects the use of AI towards military tech and the mosquito drones and whatever. And... And also naturally puts other countries in the same frame of mind, right? If we're developing the mosquito drones, why would China not develop the mosquito drones? And that just seems like a zero-sum race, and not to mention a potentially catastrophic one. Yes. compute will be limited. We'll need to disproportionately accelerate some things. To the extent it just remains totally like a consumer free market landscape, it just seems more likely that we'll get the glorious transhumanist future where they're developing the things that make human life better.
Sholto Douglas
Yes, I agree. The case where you end up with two national projects facing off against each other is dramatically worse. We don't want to live in that world. It's much better if there's like stays in free markets, so to speak. Yeah, yeah, yeah.
Dwarkesh Patel
Okay. I want to take issue with your claim that even if with the algorithms of today, if we just collect enough data that we could automate white-collar work. First, let me get an understanding of what you mean by that. So, do you mean that we would do the analogous thing of free training with all the trajectories of everything people would do on their jobs? Could you... Could you make either manually or through some other process some RL procedure based on the screen recordings of every white color worker? What kind of thing are you imagining?
Sholto Douglas
I mean, like a continuous distribution of this stuff. One like important like mental model to think about RL is I think as like the task gets more There is some respect with which longer horizon or better task, if you can do them, if you can get that reward ever, are easier to judge. So again, it's come back to that, can you make money on the internet? That's an incredibly easy reward signal to judge. But to do that, there's a whole hierarchy of complex behavior. So if you could pre-train up to the easy to judge reward signals, does your website work? Does it go down? Do people like it? There's all these reward signals that we can respond to because we can progress through these long enough trajectories to actually get to interesting things.
If you're stuck in this regime where you need a reward signal every five tokens. It's a way more painful and long process. But if you could pre-train on every screen in America, then probably the RL tasks that you can design are very different to if you could only take the existing internet as it is today. And so how much of that you get access to changes the mix. Interesting. So as we're training them on longer and longer horizon tasks and it takes longer for them to get any signal on whether they successfully completed the task,
Dwarkesh Patel
will that slow down progress because it takes more compute per task? I do think there's this notion the longer, the harder tasks, the more training is required.
Trenton Bricken
And I'm sympathetic to that naively, but we as humans are very good at practicing the hard parts of tasks and decomposing them. And I think once models get good enough at the basic stuff, they can just rehearse or fast forward to the more difficult parts.
Sholto Douglas
I mean, that's definitely one of the big complexities, right? Like as you use more compute and as you train on like more and more difficult tasks. I mean, I don't know, your rate of improvement of biology is gonna be like somewhat bound by the time it takes a cell to grow in a way that your rate of improvement on math isn't, for example. So, yes. But I think for many things we'll be able to parallelize widely enough and get
Dwarkesh Patel
enough iteration loops. Will the regime of training new models go away? Will we eventually get to like you got the model and then you just keep adding more skills to it with RL training? That depends on whether or not you think there's a virtue in pre-training a new architecture.
Sholto Douglas
Basically, if you make some architectural change, then you probably need to do some form of retraining a new model.
Dwarkesh Patel
How does the fact that... If RL requires a bunch of inference to do the training in the first place, does that push against the thing you were talking about where we actually need a bigger model in order to have brain-like energy? But then also it's more expensive to train it in RL. So where does that balance out?
Trenton Bricken
I think we've got to drink the bitter lesson here. There aren't infinite shortcuts. You do just have to scale. Have a bigger model and pay more inference for it. If you want AGI, then that's what you've got to pay the price of.
Sholto Douglas
But there's a tradeoff equation here, right? There is science to do, which everyone is doing, of what is the optimal point at which to do RL. Because you need something which can both learn and discover the sparse reward itself. So you don't want a one parameter model. Useless, even though you can run it really fast. You also don't want a 100T model because it's super slow.
Dwarkesh Patel
Yeah.
Sholto Douglas
Possible to RL. And the marginal benefit of its learning efficiency is not worth it. So there's a pretty different here. What's the optimal model size of your current class of capabilities and your current set of RL environments and this kind of stuff? And even in the last year, there's been much more of a factor of the inference cost, right?
Trenton Bricken
So just explicitly, like, the bigger the model, the more expensive it is to do a forward pass and generate tokens. And the calculus used to just be, should I allocate my flops to more training data or a bigger model? And now another huge factor is, how much am I actually going to do forward passes on this model once it's trained?
Sholto Douglas
Yeah, my total pool of compute, how do I allocate that across trained data, compute, and inference compute for the RL training? And then even within inference, there's all this research on, well, what strategy should I use?
Trenton Bricken
Should I sample 10 and take the best? Do I do this sort of like branching search, et cetera, et cetera? And so with RL, where you're sampling a whole lot of tokens, you also need to factor in the ability for the model to like actually generate those tokens and then learn and get feedback.
Dwarkesh Patel
Okay, so if we're living in this world, what is your advice to somebody early in their career or a student in college? What should they be planning on doing?
Sholto Douglas
Yeah. So I think, once again, it's worth considering the spectrum of possible worlds and preparing yourself for that. And the sort of action that I think is highest EV in that case is you are about to get, at a minimum, you are about to get dramatically more leverage. You already have. Already the startups in YC are writing huge amounts of their code with Claude. So what challenges, what causes do you want to change in the world with that added leverage? If you had 10 engineers at your beck and call, what would you do? Or if you had a company at your beck and call, what would that enable you to do?
And what problems and domains suddenly become tractable? That's the world you want to prepare for. Now, that still requires a lot of technical depth. Obviously, there is the case where AI just becomes dramatically better than everyone at everything. But for at least a while, probably, there is I think Jensen actually talked about this in an interview where he's like, I have 100,000 general intelligences around me and I'm still somewhat useful because I'm there directing the values and asking them to do things. I still have value even though I have 100,000 general intelligences. And for many people, I think that will still be true for a fair while.
And then as the AIs get better and better and better and so on, eventually, no. Again, prepare for the spectrum of possible worlds. Because in the event where we're just totally out-competed, it doesn't matter what you do. But in all the other worlds, it matters a lot. Get the technical depth. Study biology. Study CS. Really think hard about it. Study physics. Think hard about what challenges you want to solve in the world.
Dwarkesh Patel
That's a lot of topics.
Sholto Douglas
You can now. You can. It's so much easier to learn. Everyone now has the infinite, perfect tutor.
Dwarkesh Patel
It's definitely been helpful to me. I would say some combination of like get rid of the sunk cost of your like previous
Trenton Bricken
workflows or expertise in order to evaluate what AI can do for you. That's right. And another way to put this, which is fun, is just like be lazier. in so much as like figure out the way that the agent can do the things that are toilsome. But you're going to have to, ultimately you get to be lazier, but in the short run you need to like critically think about the things you're currently doing and like what an AI could actually be better at doing. And then go and try it or explore it. Because I think there's like still just a lot of low-hanging fruit of people assuming and not writing the full prompt, giving a few examples, connecting the right tools. for your work to be accelerated, automated.
Dwarkesh Patel
Yeah, yeah. There's also the sunk cost of feeling like since you're not, quote unquote, early to AI, that you've sort of missed the boat and you can't like... But I think... I mean, I remember when GPT-3 came out. So backstory on the podcast. When I graduated college, I was planning on... doing some sort of AI rapper startup. And the podcast was just like a gateway into doing that. And so I was trying out different things. And at the time, I remember thinking, oh, 3.5 is out. And people are like, I'm so behind on the startup scene here or whatever if I wanted to make my own rapper. I mean, maybe the idea of the wrapper was inadvisable in the first place, but every time feels early because it's an exponentially growing process. And there were many things, many ideas that are only becoming possible now, right? Yeah, exactly.
Sholto Douglas
It's that product exponential I talked about before. Products are literally obsoleted. You need to constantly reinvent yourself to stay at the frontier of capabilities.
Dwarkesh Patel
By the way, do you remember? I had a really shitty idea. And I gave you a call.
Sholto Douglas
I don't know what it was.
Dwarkesh Patel
It was like, I think it was like rag for like lawyers or something. Yeah, anyways. I think one of our first interactions was I'm like, hey, what do you think of this idea? And you're like, I think the podcast sounds promising.
Speaker 3
I was right.
Trenton Bricken
Which I appreciate. Yeah. I got slightly annoyed at a friend recently who I think is really talented and clever and interested in AI, but has pursued a biology route. And I just kind of tried to shake them of like, you can work on AI if you want to. I mean, I think... humans are artificial, not artificial, are biological general intelligences where a lot of the things of value are just very general. And whatever kind of specialization that you've done maybe just doesn't matter that much. I mean, again, it gets back to the sunk cost. But like so many of the people, even my colleagues at Anthropic, are excited about AI and they just don't let their previous career be a blocker. And because they're just like innately smart, talented, driven, whatever else, they end up being very successful and finding roles. It's not as if they were in AI forever. I mean, people have come from totally different fields. And so don't think that you need like permission from some abstract entity to like get involved and apply and be able to contribute.
Dwarkesh Patel
If somebody wanted to, B&A researcher like right now if you give them an open problem or like this the kind of open problem that is very likely to be the Be quite impressive.
Sholto Douglas
What would it be? I think that now that are always like come back and Papers building on Andy Jones's like scaling board links, like scaling laws for board games are interesting. Like showing that you can... Investigating these questions like the ones you asked before where you're like, oh, is the model actually learning to do more than its previous Parsec K or is it just like discovering that? Like exploring questions like that deeply I think are interesting. Like scaling laws for RL basically. I'd be very curious to see like how much like the marginal increase in meta learning from
Dwarkesh Patel
a new task or something. I mean, on that note, I think model diffing has, like, a bunch of opportunities.
Trenton Bricken
Yeah. Also, people say, oh, we're not capturing all the features. There's all this stuff left on the table. What is that stuff that's left on the table? Yeah. Like, if the model's jailbroken... Is it using existing features that you've identified? Is it only using the error terms that you haven't captured? Yeah. I don't know. There's a lot here. I think Matt's is great. The Anthropic Fellowship has been going really well. Goodfire Anthropic invested in recently. They're doing a lot of interpretability work or just applied directly to us.
Dwarkesh Patel
Anything to get your equity up, huh? No. There's just so many interpretability projects that are like there's so much low-hanging fruit and we need more people and I don't think we have much time.
Sholto Douglas
Yeah. I also want to make a plug for performance engineering. I think this is one of the like best ways to demonstrate that you have like the raw ability to do it. Like if you made a extremely efficient transform implementation on TPU or Trinium or like in CUDA, then... I think there's a pretty high likelihood that you'll get a job. There is a relatively small pool of people that you can trust to completely own end to end the performance of a model.
Trenton Bricken
And if you have broad, deep electrical engineering skills, I think you can probably come up to speed pretty fast on the accelerator stuff.
Sholto Douglas
Yeah, you can come up to speed reasonably fast. And it teaches you a lot of good intuitions of the actual intricacies of what's going on in the models, which means that you're then very well-placed to think about architecture and this kind of stuff. One of my favorite people in thinking about architecture and anthropic at the moment actually came from a heavy GPU kernel programming background, just like Nozian's announced really deeply and can think about the trade-offs really well.
Dwarkesh Patel
This is fun, guys.
Sholto Douglas
Yeah, it's fun.
Dwarkesh Patel
Thanks. Great to be back. Check out the sponsors in the description below. If you want to sponsor a future episode, go to dwarkesh.com slash advertise. Thank you for tuning in. I'll see you on the next one.