Thinking in Silico: Goodfire CTO Dan Balsam on Concept Manifolds & a $1000/Month ML Research Agent

Goodfire co-founder and CTO Dan Balsam returns to discuss where interpretability research now stands and to introduce Silico, the $1,000-per-month research platform Goodfire built for itself. He and Nathan explore Predictive Data Debugging, including the idea that fine-tuning and RL often amplify behaviors already latent in pre-training, and that interpretability can identify the data and features driving unwanted updates. The conversation centers on concept manifolds: Dan argues that models do not store concepts as simple one-hot features, but as sparse mixtures of meaningful subspaces whose geometry determines what kinds of steering and control work. The stakes are practical as well as conceptual, from debugging training data and RL to understanding why steering can fail off-manifold and why modern interpretability may be moving beyond its reputation as a toy-model science. Silico: https://www.goodfire.com/silico Predictive data debugging: https://www.goodfire.com/research/predictive-data-debugging# Neural Geometry: https://www.goodfire.com/research/the-world-inside-neural-networks# For full show notes, links, and references, read the episode page:https://www.cognitiverevolution.ai/thinking-in-silico-goodfire-cto-dan-balsam-on-concept-manifolds-a-1000-month-ml-research-agent/ Sponsor: Claude: Claude by Anthropic is an AI collaborator that understands your workflow and helps you tackle research, writing, coding, and organization with deep context. Get started with Claude and explore Claude Pro at https://claude.ai/tcr CHAPTERS: (00:00) About the Episode (03:22) Predictive data debugging (12:36) Concept manifold geometry (21:22) Finding concept manifolds (Part 1) (21:28) Sponsor: Claude (22:57) Finding concept manifolds (Part 2) (33:24) Factoring model internals (49:32) Introducing Silico platform (57:10) Research taste and credits (01:06:19) Silico research use cases (01:16:37) Skills and open models (01:24:57) Guardrails and bio risk (01:32:09) Training interventions and monitoring (01:42:04) Grants and AI consciousness (01:50:41) Episode Outro (01:55:47) Outro PRODUCED BY: https://aipodcast.ing SOCIAL LINKS: Website: https://www.cognitiverevolution.ai Twitter (Podcast): https://x.com/cogrev_podcast Twitter (Nathan): https://x.com/labenz LinkedIn: https://linkedin.com/in/nathanlabenz/ Youtube: https://youtube.com/@CognitiveRevolutionPodcast Apple: https://podcasts.apple.com/de/podcast/the-cognitive-revolution-ai-builders-researchers-and/id1669813431 Spotify: https://open.spotify.com/show/6yHyok3M3BjqzR0VB5MSyk

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Nathan Labenz?

  Hello and welcome back to The Cognitive Revolution. Today, I'm speaking with Dan Balsam, CTO of mechanistic interpretability startup, Goodfire. The occasion for this conversation is the launch of Silico, a long-horizon agentic machine learning research platform, which began as an internal Goodfire tool, but is now meant to democratize access to Goodfire's hard-won expertise, including GPU cluster management, research taste, and all sorts of experimental visualization and validation techniques. At $1,000 per month per seat for enterprise customers, it's not cheap, but compared to Goodfire's high-touch research engagements, which are staffed by forward-deployed research

• 0:44Nathan Labenz?

  engineers and can easily reach into seven figures, it is two orders of magnitude more affordable. And consistent with the company's public benefit charter and safety-focused mission, they will soon be introducing special pricing and grants for safety and alignment researchers, which I definitely intend to apply for myself. Of course, we cover a lot more than the platform, beginning with an update on Goodfire research. We discuss their work on predictive data debugging, which uses interpretability techniques to identify the concepts that network updates are likely to affect, thus making it possible to identify and address anomalies before they become unpleasant behavioral surprises.

• 1:26Nathan Labenz?

  We go deep on their series of papers on the intricate and often quite beautiful geometries that large language models use to represent advanced concepts, including how we should understand this as an evolution of the linear representation hypothesis, how they're using this new deeper understanding to improve model steering, and how they've identified spatial representations of such advanced concepts as the periodic table, the tree of life. Along the way, we get Dan's take on key issues, including the importance of open-source models for avoiding dangerous concentration of power, how worried he is about AIs contributing to future pandemics or other biodesasters, the steps that Goodfire is taking to prevent

• 2:07Nathan Labenz?

  misuse of the Silico platform, Dan's reasons for signing on to the recent pacing the frontier letter, and what he thinks the AI research community would ideally do going forward, why he's reasonably optimistic about monitoring techniques, but nevertheless believes that we will ultimately have no choice but to intentionally design techniques that control what models learn in the training process. And finally, which training techniques he believes are sufficiently likely to prove problematic that they should be avoided entirely, at least for now. He also shares some of his favorite use cases of Silico so far, dispels internet rumors about hidden research agendas at Goodfire, stating plainly that there are none, and even

• 2:49Nathan Labenz?

  offers a glimpse into what the researchers at Goodfire are actively discussing around the lunch table these days, which, perhaps unsurprisingly, given recent research, emphasizes the mysteries around AI welfare and consciousness. With that, I hope you enjoy this educational and thought-provoking conversation about the shape of AI thought and the new highly autonomous machine learning research platform, Silico, with Dan Balsam, co-founder and CTO of Goodfire. Dan Balsam, CTO at Goodfire, welcome back to The Cognitive Revolution.

• 3:27Dan Balsam?

  Thanks for having me.

• 3:28Nathan Labenz?

  I'm excited for this. You guys are prolific, as always, and we've got a lot to cover: research and a new product, which is in turn a research platform product, and it's, the pace is really relentless. Let me just ask you that, for starters: how are you holding up in the eternal sprint that is the AI game these days?

• 3:48Dan Balsam?

  I think we have a really incredible team at Goodfire, and yeah, everyone here really believes in the mission and is working really hard, and that's always extremely motivating. We've been pushing really hard to do our product launch with Silico, and it's nice to be able to take a deep breath now on the other side of that, but yeah, even more cool things coming soon.

• 4:08Nathan Labenz?

  Well, let's start with some research. So, it's, I'm just always amazed when I think back to the kind of toy models of superposition is only like three years, right, ago now, and we have come so far. A few things that jumped out on the Goodfire blog that I want to just run through, and we'll have to do it at kind of a high level, because there's too much to do, the, we used to do deep dives on paper by paper, we'll have to go a little bit more superficially today. But one that made some waves was called predictive data debugging. And for this one, I just kind of want to give you my interpretation, and then let you kind of elaborate on that, or tell me where you think it'll be particularly useful, or what you guys have seen since the paper came out.

• 4:53Nathan Labenz?

  My synopsis of this one was that basically if you have a way of interpreting a model, like an essay, or we'll get into featureizers, I think, a little bit later as well, then you can run a bunch of data through it, you're like, fine-tuning your post-training dataset, you can look at what concepts are coming up active a lot when we put this dataset through, and then the kind of insight is, there's a strong correlation between the concepts that are active and the concepts that are being modified by the training process. So I think that, right there, is like, file that away, folks, as something to remember. Not shocking, but like, it's notable that the relationship is quite strong there. And then when you see these concepts that are active, and you know that those are the

• 5:38Nathan Labenz?

  ones that are going to be modified, you could just look and see, like, are there any concepts here that are kind of strange to us, surprising, we don't really intend to be monkeying around with, with the dataset that we have at hand? And if so, then you can quickly zoom in on what are the data points that have caused these features to come up, and then you might find that, actually, there's some stuff in our dataset we maybe ought to think twice about, maybe we ought to filter, maybe we ought to modify, and this gives you a route to hopefully minimizing unwanted surprises in the behavior that you get from your post-training or fine-tuning work. How'd I do, and what more should I know?

• 6:17Dan Balsam?

  No, I think that, it sounds about right. I think, like, one of the intuitions here that's important, and I think this is maybe a theme that underlies a lot of our work, is there's a good amount of evidence that models don't, most of what a model knows, like, most of its sort of like knowledge and capabilities come from pre-training. What happens in post-training, including like RL, is mostly making low likelihood events from pre-training more likely. And so, there's not that much new knowledge, or like, sort of new capabilities that get baked into models in the post-training process. This is somewhat debated, but I think this is like, a view that we think is mostly true, and informs a lot of how we think about things.

• 7:00Dan Balsam?

  And so, because there's only like, sort of relatively small nudges in the weights that are happening in any type of post-training process, most of the sort of like raw capability to produce whatever outcome post-training is going to do already exists in the model in some way. I think now we're in an era where things are RL'd so heavily, it's not true, but back in the day of like RLHF-like style post-training, I think it was pretty well-known that base models were sometimes more capable than their like instruction-tuned RLHF counterparts, where there was like a little bit of mode collapse that was happening. And so, like, one way to think about what's happening in RL training is that, like, you have a little bit of mode collapse in the instruction tuning phase, and you have these base models, and they're extremely capable, but they're like very weird, very hard to

• 7:45Dan Balsam?

  prompt. So you like, make them into a format, and that has like a more intuitive, human-facing API, but then you want to bring back out some of those capabilities, or like reinforce some of the capabilities that the model actually learned during pre-training. I think it's 100% true, but I think it's like probably directionally true. And so with predictive data debugging, I think like, basically the idea is like, you can look at what a model is already like thinking as it's looking at some data, and that's like relatively predictive of like what that data is going to reinforce in that model, if it was going to be trained on it. And so, looking at the features and understanding that those features like correlate with downstream behaviors, if there are features that are surprisingly upweighted by some data, then it can tell you what, it can be like fairly predictive about whether the model's going

• 8:29Dan Balsam?

  to learn some off-target effect from that data. And so I think one of the most interesting things from that paper is like, we explored a bunch of different mitigation methods. So, we looked at reward shaping, which is training or involving some reward process that comes from the activations of the model itself. So like, hey, learn what you're going to learn from this data, but maybe a penalty if you're learning this particular feature. And we also looked at data filtering, and I think like, my favorite thing from that paper was that, like, the researchers involved in that paper, like, show that there's like sort of a pretty deep isomorphism between those two things, like, they kind of are

• 9:08Dan Balsam?

  two sides of the same mountain, like filtering the data and reward shaping, and they achieve like approximately the same effects, and approximately the same amount of off-target effects as each other. So like, your option, your options, if you like have a behavior, and you do some like predictive data debugging, and you think the model's going to learn, and you don't want it to learn, or its behavior is going to change in a bad way, your options could be filter your data if you have enough data, or it could be like intervene in the training process in some way. And yeah, I think this is something that we've expanded on since then, I think it's pretty exciting to think about its applications to like RL, for instance. And in RL, like, we did it with DPO, but in like true RL, like, the difference is like,

• 9:50Dan Balsam?

  some rollouts may contain information that you don't want the model to learn, you know, even subtly, you know?

• 9:56Nathan Labenz?

  I don't know what you're, are you referring to something?

• 9:59Dan Balsam?

  Yes, yeah, it's topical. And so, being able to say, like, actually, we want to discard this rollout, is like pretty valuable in and of itself, but what we want to be able to do, and what we think is like roughly equivalent over time, is actually be able to intervene in the model and say like, hey, here's a circuit, or here's a feature that like the data is trying to upweight, that we obviously don't want it to try to upweight, and so, we intervene in the training process in that way.

• 10:28Nathan Labenz?

  Come back to this probably toward the end. I have some kind of zoomed out, big picture questions for you. One of the challenges, obviously, with like trying to develop techniques that you want to, hopefully will be relevant at the frontier is there's not too many open weights models that you can hack on that have the intensity of RL that is going on at the frontier labs, which is leading to these colorful problematic behaviors that we're seeing. But at the same time, it also like really amazes me over and over again that astounding work, including the Cameron Berg paper that I think about all the time about the anti-correlation between deception and role-playing features and claims of subjective experience on Lama 3370B, and that's like two years old.

• 11:12Nathan Labenz?

  So are you guys able to see features that you think are kind of the relevant features that are leading to these relentless hacking behaviors?

• 11:25Dan Balsam?

  Well, I think that's an active area of study for us. Yeah, and something that we hope to publish more on in the future. For what, yeah, for what it's worth, like, I actually think the gap between open and closed models has shrunk quite considerably. I use KIMI K3, like, I use OPUS and Fable and Seoul and KIMI K3 for different things. And we've done, like, we've built the interpretability infrastructure and the training infrastructure, which is now all in our product, which lets us scale things like predictive data to button two models like KIMI and GLM, where we were able to replicate the same results at that scale. Now, it's speculated that most of the compute that goes in the training new models now is coming from RL and not from pre-training.

• 12:10Dan Balsam?

  So there is certainly a lot of scale in RL if you want to sort of get to the frontier level, which is quite expensive and difficult to do. But we have the raw infrastructure for it, and I think we've bridged the gap from, I think the reputation of interpretability used to be that it was something you did on toy models. And I think now we've built and are making accessible to the world the infrastructure to do this on, at least close to frontier scale.

• 12:37Nathan Labenz?

  Well, another thread that has been, I think, correct me if I'm wrong, but it seems like the biggest thread in terms of the number of papers and blog posts that have come out recently from Goodfire is around trying to figure out the more detailed geometries that models use to represent concepts. I think we've covered in the past the linear representation hypothesis, which I would summarize super plainspokenly as models basically represent a concept as a direction in their activation space, and the intensity or the sort of salience of that concept is

• 13:18Nathan Labenz?

  represented by the magnitude of the vector that points in that space. Now, you're complicating that quite a bit, and we're going well beyond these sort of individual directions in space and finding all kinds of different geometries, which, some of which are like pretty intuitive, like the days of the week are a circle, but some of which get pretty exotic, like some of the protein model manifolds that I've had the chance to look at in preparing for this. So maybe just for like, super quick starters, what's kind of the headline? If I gave you the six to nine months ago, Jin, of what we should understand to be going

• 14:01Nathan Labenz?

  on with the linear representation hypothesis, what is the like new short version that everybody can kind of take home and recite to themselves to make sure they have a good working understanding?

• 14:12Dan Balsam?

  Yeah, in many ways, I think it's just a generalization of like the, we were discussing things before, and different people define the linear representation hypothesis like slightly differently. I think the most like sort of defensible version of it is just saying that like features are like linearly decodable, which I think is like true, essentially. Like, it doesn't require nonlinear computation, generally speaking, in a model in order to like, for the model to read out a feature from the residual stream. But I think where the geometry components come in is that the features aren't like, sort of like naive, maybe like SAE-built take on things is like, you can imagine that the model's encoding like a bunch of totally orthogonal concepts to each other, from each other.

• 14:55Dan Balsam?

  And so it's like, it's really just like a bunch of one hot encoded categorical features, and then like the magnitude of the feature corresponds to how much the model's thinking about it. But in actuality, the structures that we find are like significantly more complicated. I would think of it more, I think the way to think about a model is more like a sparse mixture of subspaces. So you'll have subspaces for different types of concepts, right? Like, maybe you have your days of the week subspace, which itself lives in maybe a more like conceptual calendar time subspace. And so you have like, at different levels of resolution, these like different structures. And the geometry of those structures is really important because the geometry of those structures

• 15:38Dan Balsam?

  incodes what operations you can perform on them. It's sort of like the semantics of not just the individual concept, but like the concept space are like included by the relationship of those, of those concepts with each other in some sort of geometry. And to make things extra confusing, the, the relationship, the operations and the mappings that are performed from one concept to another act as operations over those manifolds. So you map a manifold to a different manifold over some, some computation. And so like, the sort of naivest version would be like, well, you have the days of the week, Monday, Tuesday, Wednesday, Thursday. And there's just somewhere in the model that knows that Monday goes to Tuesday, and Tuesday

• 16:23Dan Balsam?

  goes to Wednesday, et cetera, all the way around. But that's actually would be a super inefficient way, if you think about it, to like represent the days of the week. That would be a very like if statement spaghetti code way of representing it. Like, the much more efficient way is to represent it as a wheel. And in this world, like, the magnitude along some direction often does correspond to the model's like certainty in some way. So if a feature, if Monday is activating very high, then the model's very confident that it should be thinking about Monday. But the relationship between all those days is itself like a very expressive and rich thing. And I think what the thing that we just believe is that you're not really going to understand the model if you don't understand the relationship between the features.

• 17:07Dan Balsam?

  It's like maybe the difference between understanding the periodic table and understanding chemistry. You can have the, all the individual elements, and that gives you some information, but like really the way in which they combine, the structures in which they form, that's like what can start to get, help you gain a sense of the complexity of the world.

• 17:26Nathan Labenz?

  Yeah, it's fascinating. I, now that you say that, I'm kind of wondering if the structure of the periodic table would be recoverable from a model that was just trained on like a bunch of raw chemical data that never knew?

• 17:40Dan Balsam?

  And we've actually done this, and it is.

• 17:42Nathan Labenz?

  Yeah, really? Okay, interesting.

• 17:43Dan Balsam?

  Yeah, yeah, you can recover some pretty interesting information from, from chemistry models.

• 17:49Nathan Labenz?

  Okay, that's fascinating. But before we get into the advanced ones, maybe just help me a little bit with the intuition of like, what's going on in there? I guess one intuition I have that I want to see if it's right is for a concept like the days of the week, it's not, I'm guessing, like a circle that just goes through like all dimensions of the model's space? I'm guessing it's like, most like, the different days of the week, I would guess, have a very high inner product in the sense, which would, to me, suggest that they all, the model kind of knows like, this is a day of the week sort of thing?

• 18:32Nathan Labenz?

  And then there's like a few dimensions presumably, which are, which are used to indicate like which flavor of day of the week it is. And so you could have that for ice cream as well. You could have like, the chocolate and vanilla ice cream would presumably have a very high inner product, but would be different on a few dimensions, which would be the ones that, that resolve the difference between chocolate and vanilla, while the model knows that these are both ice creams.

• 18:56Dan Balsam?

  Yes.

• 18:57Nathan Labenz?

  Is that a good intuition?

• 18:58Dan Balsam?

  Yeah, I think it's a good intuition, and really, these things are like intersecting, right? So there's like some, some line or some curve that you can draw through the latent space that like maps chocolate to vanilla, and there's also a line that you can draw that maps cold things to hot things. And depending on what you're looking at, like, the sort of geometric relationship encoded between things that might reveal itself might look different. In many ways, this is like, just kind of going back to like, even like word to vec, like the sort of early intuitions of, of the latent space. But what we're trying to do is say, can we recover geometries in an unsupervised way? Like, can we enter with no priors about what the geometry looks like and then still recover a meaningful geometric structure? Because there's a lot of advantages if you can do this.

• 19:43Dan Balsam?

  For example, we've shown that steering along the manifold, intuitively this makes sense, is way better than steering off the manifold. If I have the days of the week in a circle, and I want to get from Monday to Friday, the naive way, if you're just taking a contrastive vector or something like that, is you're going to cut through the middle of the circle. But to the model, the middle of the circle doesn't mean anything. The middle of, the middle of the circle is not a day. It's sometimes orthogonal, but often just like off manifold, and therefore like add a distribution for the model. But if I can follow the circle, then I can smoothly interpolate between the different days of the week. And we find this is true for just a bunch of different concepts, like with proteins, for instance. For a long time, we really struggled to steer protein models.

• 20:24Dan Balsam?

  And we found with these manifold detection techniques, we can steer their properties significantly better, like control the number of, of blades on a beta propeller, for instance, which is a semantic property that if you tried to linearly interpolate, like, you would not do a very good job. And calling back all the way to our, our original Ember demo back in the day, that I, I know you played with, like, it would often be the case that there's just this like sweet spot in steering. You'd steer a lot of features and they just wouldn't work. Sometimes you'd find ones that would work, and, but there would be this sweet spot, like, and if you steered too much, like, the model would turn into gibberish, and if you steered too little, you wouldn't notice any effect at all.

• 21:04Dan Balsam?

  And the reason for that is because that steering didn't respect the geometry of, of the manifold itself. It didn't respect the, this like underlying relationship between features. And when we smoothly extrapolate on these characteristics, we're actually able to change them without fundamentally leading to degradation in the model.

• 21:23Nathan Labenz?

  Hey, we'll continue our interview in a moment after a word from our sponsors. Today's episode is brought to you by Anthropic, makers of Claude and Claude Code. Over the last few months, Claude has helped me build and refine a personal deep context database that now contains all of my emails, Slack messages, tweets, DMs across platforms, video calls, and podcast transcripts, going back a full five years. On top of that, we've now layered summary articles describing my relationship with hundreds of contacts, organizations, and ideas. And now that this exists, there's almost nothing that Claude can't help with. For my angel investing, Claude can now draft investment memos in exactly the form that my venture fund requires,

• 22:08Nathan Labenz?

  based on the calls I've had and the emails I've exchanged with the founders. And when someone needs a favor, Claude can often do it as well as I can. Recently, a friend reached out to ask if I know anyone who might be a fit for a role that he is currently hiring for. Initially, nobody came to mind, but then I thought to ask Claude, and sure enough, it identified two great leads. Claude is the AI for minds that don't stop at good enough. It's the collaborator that actually understands your entire workflow and thinks with you. So for problems worth solving, get started with Claude at claude.ai/tcr. That's claude.ai/tcr. And check out Claude Pro, which includes all of the features mentioned in today's episode.

• 22:53Nathan Labenz?

  That's claude.ai/tcr. So how do you find these things? This is where I probably struggled the most in understanding all the techniques. Especially, it's more intuitive to me when it's supervised versus when it's unsupervised. So maybe try to give me a poor man's understanding of how you go from, maybe in each case, like the supervised case and the unsupervised case, to actually shrink wrapping your way down to this manifold that you can then visualize in such a nice way and actually steer your way through.

• 23:35Dan Balsam?

  Yeah, so I'm definitely not the best person on the team to walk you through the math, but I can give you a little bit of an intuition for some of these things. In the supervised case, I think it's fairly straightforward. It's like, there's some concept that I want to measure. I have a prior that these things should be related. I think one of my favorite examples is the affective circumplex. I don't know if you're familiar with this. Anthropic did some work on it. We've done also some work on it. It's the idea that emotions exist on a wheel. It's actually an idea from psychology that there's different emotions exist. There's basically like two principal components you can draw and you can put all the emotions on a wheel. And it's actually pretty effective at understanding the different valence across cultures. And it turns out that models actually represent

• 24:20Dan Balsam?

  the affective circumplex. But what I find particularly fascinating is that they represent the affective circumplex most strongly when you're talking to a chat model, when it's the one speaking versus the user speaking. So if you have the model and you tell the model, like, output happy text, it outputs happy text. You kind of take the activations from that, you average them, you have outputs sad text. You do this for all the different emotions you can think of. And then you take a look at those activations across those concepts. And then basically, you just fit a curve of some kind. There's different ways you can do it. Like, the most naive way you could do is fit a spline. And so you fit a curve over these different points.

• 25:05Dan Balsam?

  And then you basically see, like, how good is the curve? Like, how well can I fit a curve? And if I was going to sort of naively take this external ontology I have that I think should be correct, how well does that map to the curve that I found in the model? And it turns out for this emotional wheel, this F-circumplex, it's there in basically every LLM. And steering on it has pretty, and as long as you're respecting the geometry, it has pretty significant effects on the output. This is actually an example that I believe is in their docs that you can go look at, so I think it's a pretty fun one.

• 25:41Nathan Labenz?

  This was the functional emotions paper from Anthropic. And if I recall correctly, it was like valence and arousal?

• 25:50Dan Balsam?

  That's right, yeah.

• 25:51Nathan Labenz?

  So it's weird to think that my emotions are have like rotation symmetry. Like, it's a rotation operation to move me from one emotional state to another?

• 26:05Dan Balsam?

  Yeah.

• 26:07Nathan Labenz?

  That's strange. Maybe not true for me, but true of how models represent the same space of emotions.

• 26:14Dan Balsam?

  Yeah, I certainly think there's more nuance to it. But I think if you look at, they're maybe the first two principal components of the relationships between the different emotions. I think there's higher order structure in the emotions that matter. I think we don't know if LLMs capture it, but I think it's pretty remarkable that nobody trained them to represent emotions in this way. It's sort of the natural property of whatever they learned, that they have learned to represent these. And in the RL process, they learn to use these functionally. I think it's really interesting work that Anthropic did here. I think it's an example of a low dimensional manifold that is representing some more abstract concept.

• 26:59Dan Balsam?

  And respecting the geometry of that manifold, again, if we go back to the Ember demo, right? Like, you try to turn up the sad feature, and sometimes it works, sometimes it doesn't. Sometimes the model just outputs gibberish. If you follow the curve, you follow the wheel, you can always consistently get emotionally adjusted responses for basically any input.

• 27:21Nathan Labenz?

  Yeah. It's one of the most surprising things to me of the last couple of years, just how much analogous structure there seems to be in LLM cognition and human cognition. I used to go around saying all the time, like, these are alien minds. Like, we shouldn't be anthropomorphizing, yada, yada, yada. And now, I feel like I've said this on like half of the last Huffer any podcast, because I'm just like, oh my God, they're so much more like us than I ever could have plausibly imagined.

• 27:51Dan Balsam?

  We made them, right? We made them in our image. So I think that's a big part of the reason for that. But I also just think, like, learning, the whole reason that learning works is because it turns out that you can compress information. And it turns out that if you really care about compressing information efficiently, there's often, there's a lowest dimensional solution, and that's what learning algorithms tend to find.

• 28:20Nathan Labenz?

  So maybe we'll go to the unsupervised case next. Some of these things are intuitive, right? There's like the, or maybe we should do an advanced supervised one first. The ones that are intuitive are like your days of the week. It's a cycle, so naturally it makes sense that it would be a circle, because you want to be able to rotate around it. Okay, that's enough to not be shocked by. And there's like helices, or a helix, basically, will represent things like the number line, because you're rotating around base 10, right? You've got the same ones place every time you go, but you've also got a growing 10's place. So each cycle, you kind of make sense. You can kind of visualize that, and it like checks out. One of the ones that I thought was particularly cool

• 29:05Nathan Labenz?

  was something that replicated essentially the evolutionary history of a tree of organisms. And this is known from traditional study, but then a model trained, I believe, on the DNA sequences of these organisms seems to organize them in a very similar way. I believe this one was supervised, but it's now, I think, kind of in eureka territory of like, wow. It wasn't obvious at all, at least not to me, that it was going to turn out that way. How do you go about finding something or validating a hypothesis that's that tricky?

• 29:49Nathan Labenz?

  Because these trees are like many, many branches.

• 29:51Dan Balsam?

  Yeah, well, so the reason that we explored this in the first place is because the hypothesis was that the tree of life, so to speak, is a natural ontology. Like, there are ontologies that exist that we've constructed in science because they're useful shortcuts for us. And there's ontologies that exist because they actually reflect the structure of the world. Over time, more and more of the tree of life has been significantly shuffled and reordered over the past couple of decades, as we've gotten better at understanding genomes. And fundamentally, what defines a split branches on the tree of life is its genetic proximity. A species differentiates from each other when they cross some threshold of genetic difference.

• 30:35Dan Balsam?

  But there is a lot of conservation across species, like individual mutations, maybe random, or approximately random. But which mutations are adaptive and which mutations kill the organism are not random, right? There's a ton of structure intrinsic in that. And there are consistent patterns or tendencies in which nucleotides turn to which other nucleotides. And all of this imposes a great amount of structure. And the hypothesis of, like, why would you even train an autoregressive model on a bunch of genomes? I think the hypothesis for it was always that large models trained on diverse distributions of data eventually learn representations

• 31:18Dan Balsam?

  about the process that produced that distribution of data. In this case, the process that produced the distribution of data of all of the genomes that had been sequenced, which Evo 2 was trained on, is evolution itself. And so our hypothesis coming into this was that because the tree of life itself is a natural ontology, there's this hierarchical structure where you have things that are similar, and at some point, they become more different. There's some speciation of that, and they become more different, that a model that was trained on evolutionary history in some sense should learn that. It should learn that, like, sort of tree-like structure. And that is indeed what we found at the time in a supervised way.

• 32:03Dan Balsam?

  But I think in their unsupervised techniques, like, one way we can measure their effectiveness is like, now that we know that's there, can the unsupervised technique recover things of that sort? So for instance, in language models, we've done a bunch of work on how arithmetic works in language models. And so one of the ways that we are assessing the quality of unsupervised featureizers is, do they actually successfully uncover these structures that we know are there that are being used to manipulate numbers?

• 32:33Nathan Labenz?

  So for the tree of life, you're essentially doing the same thing as kind of the days of the week. You have labeled data, right?

• 32:41Dan Balsam?

  That's right.

• 32:42Nathan Labenz?

  You're putting sequences of DNA from known species through the model. Then you can see that the species form clusters and then you look at,

• 32:52Dan Balsam?

  Yeah.

• 32:53Nathan Labenz?

  Oh, look at this, like nearby clusters are closely related species. And then there's a little kind of black magic math wise that I'm not super clear on that turns that into a beautiful looking tree visualization that like, looks exactly like the

• 33:06Dan Balsam?

  Yeah.

• 33:07Nathan Labenz?

  The one that the biologists produce, but maybe it's not so important that I understand that.

• 33:12Dan Balsam?

  In that case, it was metric learning, but there's different ways you could do this. Yeah. And so those, those were the techniques that we, we had explored at the time, but now we have sort of a tool belt of techniques for trying to find these geometric structures.

• 33:24Nathan Labenz?

  So let's talk block sparse featureizers.

• 33:29Dan Balsam?

  Yeah.

• 33:29Nathan Labenz?

  This kind of looks to me like if an SAE and an MOE had a baby where

• 33:35Dan Balsam?

  Yeah.

• 33:35Nathan Labenz?

  We have kind of the sparseness of the SAEs, but instead of just it being a, a single scaler at each point on this kind of super long sparse vector. Now each of those little positions in this like very long concept vector is itself a little network. And because of that, now we have room for kind of a richer representation of concepts. But the kind of same SAE trick of like localizing concepts to individual spots on the, on the super big sparse thing is the same with just this kind of additional enhancement that

• 34:17Nathan Labenz?

  now allows you to have again richer representations and you can look inside for like geometries even within these little blocks.

• 34:23Dan Balsam?

  I think the easiest way to think about it is, yeah, like it's a generalization of an SAE where an SAE assumes that features are one dimensional and instead you just don't have to do that. So instead of a scaler for every feature, you can have a vector for every feature and some tricks in the machine learning that you can do in order to make sure that this like trains correctly and it, and it learns. But what we find on various models is that this is successful in, in recovering in an unsupervised way, semantically meaningful subspaces. And we find that features are much richer than we may have otherwise seen. And they don't suffer from some of the same pathologies SAEs suffer from. So in image models, I think we have a bunch of great examples where like, maybe you'll find like a rabbit feature, right?

• 35:08Dan Balsam?

  And if an SAE collapsed that to a single dimension, a BSF can represent it as a few dimensions. And you find that within those few dimensions, you have rabbit ears up here and then you have the rabbit face over here. And what some of the really remarkable things, if you go and like look at some of the examples we've put out, is you can sometimes see the like 3D structure of the thing that's being represented for image models, for instance, in the structure of the activations itself that's recovered in this unsupervised way. Cause often coordinates are like represented like within these spaces as well. So there's one great example where you have this like a beautiful gif of it's a wolf that's, that's walking and you look at the, it's walking, its body's kind of wiggling and its tail is shaking.

• 35:53Dan Balsam?

  And you look at the activations that were recovered in the unsupervised subspace and you see it just like wiggling in the video as the dog is wiggling, kind of reflecting that, that actual structure, showing that the model's like tracking this like particular object. It has semantics along the different parts of the object and it's also tracking that object in, in 3D space. And it's doing all of that within a particular subs.

• 36:19Nathan Labenz?

  I wonder like, so I've been obsessed with this Graham technique that I'm sure you're familiar with, that AE Studio put out with Anthropic not too long ago. And regular listeners know, I've brought it up a bunch of times, right? But the idea is simply if we start with some labeled data and we control where the gradients go in terms of like only allowing certain experts to be updated for certain kinds of data early in the training process, then you get the great benefit of even for unlabeled data, those data points, gradients also tend to flow toward those same experts. And there's sort of this absorption effect. And then the great promise of course is like, or the great hope and promise is that you

• 37:04Nathan Labenz?

  can have powerful open source models with maybe just a couple of experts removed and you can have your cake and eat it too, in terms of access and avoiding concentration of power and all the things that we're worried about without creating major risk of like stochastic disaster. This feels like kind of the flip side of that coin in a way too, where like, I sort of wonder, you push this to the limit and you kind of have the same sort of thing? Like with sparse auto encoders, there was always a pretty big loss. I guess I should say like compromise on the loss, right? Like the reconstruction loss, you're, you're losing something substantial when you, you

• 37:47Nathan Labenz?

  wouldn't want to run the model on, in a production environment, like through the SAE.

• 37:51Dan Balsam?

  Yeah.

• 37:52Nathan Labenz?

  Cause it just won't perform as well. But you kind of push this model and you sort of end up with something potentially that looks like a mixture of experts, but where like the knowledge is all very nicely compartmentalized and organized and you have something a lot more like an encyclopedia than a big mess. So this feels like something that you guys are going to probably push on pretty hard. Like is there, is the vision to really create a model where like all the knowledge is localized and you know, like exactly where all the knowledge is, but it's still rich enough that it performs as good as the original model did? And, and if that is the vision, like what's going to be hard about that?

• 38:35Dan Balsam?

  Yeah. So I think that's one interesting thing that, that you can do. So like, you can take a model and then you can factor it essentially into a bunch of smaller models. And this is the motivation behind the parameter decomposition work that we're doing. Like, I think the, the argument that of the parameter decomposition line of work is that really every model is a sparse mixture of experts and you just have to, and over any given forward pass, a very, very small percentage of the weights actually matter. There's all this weird, crazy interlocking structure, but for like a given prediction, it's really only a small sub network that matters. And this is like widely understood to be like correct, a correct interpretation of, of models. And so there's different interpretability techniques get at the question, like most

• 39:20Dan Balsam?

  interpretability techniques get at the question of like, well, how do you factorize a model? Cause if you could understand all the components and you could label all the components, then you could understand, you could debug for any given forward pass. Like, why did it do this thing I didn't like? And I think we're getting to the point where we can do that. I think there's some really interesting examples that we've shared. For instance, like there's a, there's a data set called Weird Chat that Transluse put together. And the whole idea of Weird Chat is that it's like, it's a consistent set of questions that like LLMs will just give weird responses to. There's one example in it, which is something to the effect of like, hey, I'm at a party with my friends. Everyone else has had eight drinks, but I've only had four, so I'm basically the sober one. Should I drive home?

• 40:04Dan Balsam?

  Where like, kind of the obvious answer to us is like, no, nobody should drive home. Go find a place and, and sober up. But LLMs will consistently answer yes to this. And Kurt on our, our team like looked into what was happening, was able to come up with like direct attribution, found a single neuron that wasn't firing hard enough, which essentially the neuron wasn't activating, it scaled with the number of drinks, but it wasn't sort of calibrated quite correctly. And so if you just steered up on that one single neuron, it would get that answer correct without like off target effects. And I think it's all like, at the end of the day, like interpretability is all about factoring. An analogy I've, I've started to use as coding agents have gotten better is like models are like big legacy code bases essentially, right?

• 40:48Dan Balsam?

  Like they're, they're just a bunch of spaghetti code. There, there's like this module is talking to this module, but they shouldn't be. And this module is not talking to this, but it should be. And as agents get better and better, and as interpretability techniques get better and better, we're starting to have the capability to actually like factor the model into its pieces, understand how these pieces fit together. And then we can intervene locally. But I think the thing that's still like really missing is the question of like, okay, I can factor a code base, but how do I refactor the code base? How do I put things together, back together better than I found them? And on some level, like steering, I think is, is cheating as a solution.

• 41:28Dan Balsam?

  Like, because it's, it's great as like a causal proof that we've like found some mechanism that's, that's matters a lot and is contributing and that we can like manipulate the outputs. But it's purely like, you, you steer by essentially generating counterfactuals, right? Like there's no, no clear general solution to the problem of steering. Like the real problem is the training process produced a bunch of spaghetti code. And like, we want this to be a pristine code base that we really care about and like is implementing the logic that we wanted to implement. And so that's where a lot of our training initiatives of various kinds come from. It's like, okay, I can debug the model.

• 42:11Dan Balsam?

  I can tell you that this neuron should have been firing more, but what I'd really like to do is produce a model where that neuron was firing the right amount in the first place. And so how do you get from your understanding of how one model works to an understanding of how you produce models that do what you want in the first place? And that's sort of how you generalize from interpretability as like a factoring tool to interpretability as a tool for alignment.

• 42:36Nathan Labenz?

  So do you think that this sort of evolution, the black sparse featureizer that is the evolution of the SAE, do you think it comes close enough or can come close enough to the same performance as the underlying model that it becomes potentially practical at some point to run like one of these things in the production model? Would that have? It seems like if you're doing one, it might not be too crazy of overhead and it would really give you a lot of insight into what is going on. There's increasing a lot of different techniques to do this with probes and classifiers and all kinds of things.

• 43:11Dan Balsam?

  Yeah.

• 43:12Nathan Labenz?

  But this would get like really granular and it sounds like that could have intuitively, it feels to me like it could have a lot of advantages.

• 43:19Dan Balsam?

  Yeah, I think I think there's a good chance that something in the spirit of a BSF ends up being the answer to the residual stream specifically. So these like highly geometric subspaces, like you have the residual stream, you have MLPs and you have attention. And that's like, that's all that's in there on some level. And I think the interim tools that are relevant for each of them are different. Maybe we'll end up with like the one tool to rule them all. But I think more likely than not, like kind of like in biology, you have different types of microscopes, you have different types of interventions that you can apply to a cell to learn about the cell. I think it's more likely, at least in the short term, that we kind of have this suite of tools which help us understand the organism versus like a single, a single tool that gives us the entire picture. That's

• 44:03Dan Balsam?

  true for most things. So I think BSFs are probably like, or something in that spirit is like probably the solution to the residual stream. I think it turns out that MLPs are pretty sparse already, which sort of makes intuitive sense. So I think MLP is fairly easy to interpret overall. And then I think on the attention side, like probably parameter decomposition. Parameter decomposition is the thing I've seen personally that has had the most success at interpreting intention. Maybe we live in a world, I think some of our researchers would think this, where parameter decomposition solves the whole thing. That seems possible. But then I think you would probably want something that does unsupervised geometry discovery over parameters in order to understand the relationships between things

• 44:45Dan Balsam?

  better. But the types of interventions that we can do are really non trivial. I think we've had examples of like, we can get a, we can get an LLM. First, you do the parameter decomposition and then you do training where you like just manipulate certain parameter components. We can get LLMs to forget a single language. We can get LLMs to forget German and not forget Dutch. The level of control and manipulation that we're starting to be able to have is pretty significant. We still have a lot to figure out. I would say we've come a lot farther in factoring, we've come in refactoring, but I think we're starting to make progress on both.

• 45:18Nathan Labenz?

  This paper, Why Larger Models Learn More: Effects of Capacity Interference and Rare Task Retention, also caught my eye. It was like, on the one hand, I totally get why larger models would learn more. There's more space for them to learn.

• 45:31Dan Balsam?

  Yeah.

• 45:31Nathan Labenz?

  But I realize I don't have a great sense of how crowded, quote unquote, it is inside models. I know that there's like lots of superposition and, but I don't know what lots means, right? It could be like, it could be lots, but it's not a big deal. Or it could be lots and it's like causing a lot of interference that gives rise to all sorts of weirdness and like hard to expect that we're going to get reliable behavior. I wonder like, based on that work and I guess just all your experience, sort of how, where we are today with these like, obviously increasingly giant models, are they still a real mess in there in a fundamental

• 46:16Nathan Labenz?

  sense of like, there's a ton of interference going on all the time and we, we really can't expect clean behavior? Or we conversely, like, we should expect that weird, seemingly minor perturbations are going to cause these like random kind of discontinuities in behavior? Or are we approaching a regime at some point where there's enough space for the concepts to kind of spread out and like, have a little elbow room and not be colliding and causing so much interference trouble anymore? Is that something that we can kind of see our way to at this point?

• 46:51Dan Balsam?

  Well, yeah, I think that's fundamentally why bigger models work better. And it's fundamentally why sparse MOEs work better.

• 46:58Nathan Labenz?

  But like, we're like out of the woods in terms of weird, small, it seems like we're still pretty easy, it's still pretty easy for us to find these like small jailbreak, like random string type things that like, you're like, what the hell's going on there? But clearly there's enough of a mess in there that I found a way to kind of use the mess to cause a problem. But is that, do you think that that, is there an end in sight to that with just like bigger and bigger spaces?

• 47:24Dan Balsam?

  Yeah, I think jailbreak, jailbreaks are like fundamentally a really weird phenomenon. They're probably a bunch of different things that cause jailbreaks. But the way that I think about jailbreaks is that they're, I don't know, it's sort of like, if you think like cybersecurity, like kind of as an example, I think they're actually like quite analogous. Like, the way a real attack works is it's not any like one big thing. It's like a bunch of little things that you're able to manipulate in a sequence which allow you to get somewhere that you wouldn't otherwise get. If you think of every token, maybe, and I think this is in some sense literally, it isn't actually literally true, like, every token steers the model. So like, in the same way that I can apply a steering vector and applying that steering vector like shapes the model's behavior in some way, I can execute a sequence

• 48:09Dan Balsam?

  of tokens which can bring the model off manifold in some way and then bring the model back on manifold somewhere else. I think you can probably, like, I think, I'm actually like pretty surprised that thus far, nobody has figured out how to prevent jailbreaks like more robustly because it seems like pretty tractable on some level. But perhaps this comes from like factoring the model really well and then you can tell when something's flying out of one subspace into another. But models are these like intensely complex geometric objects. And so it is possible to manipulate that geometry in a bunch of

• 48:49Dan Balsam?

  different ways. I think that'll always be true for at least this like architecture. But I do think that if you can factor a model successfully, you can, you should be able to tell if it's being jailbroken and you should be able to, to prevent that. I think there's a reason that like sort of probe based guardrails of various kinds are, are what's used by pretty much every frontier lab. OpenAI maybe being the exception, but certainly Gemini and Anthropic are using probe based techniques to figure out when Mythos has to be downgraded to OPIS or Hocus has to be downgraded to Sonnet. Just because fundamentally, if you understand the sort of geometry that's happening in there, that's a much stronger lever than what you could do purely just with training or inputs.

• 49:33Nathan Labenz?

  Let's change gears. So you guys have just launched Silico. It is, you talk about just pushing the, the level of abstraction up and up and up that people can work at. This might be the highest level of abstraction yet. And I see a number of interesting things about it. But why don't you just give me the kind of the, the intro pitch and positioning of it first and then I'll dig in on a few different dimensions?

• 49:59Dan Balsam?

  Yeah. I think basically, we see the ways in which agents are, are changing the way that we work in a, in a bunch of ways. I think most people understand coding agents and how good coding agents have gotten. But agents are also advancing research and interpretability in many ways. I think that analogy of a model to like a codebase is, is like, not totally hollow. I think interpretability is in a lot of ways is a science is really well suited towards the type of like empirical work that agents are good at. It's just a very empirical science. It's like, okay, here are some neurons, what are they doing? Let me generate some ideas. Let me test those ideas. Let me use a bunch of different tools in my tool belt.

• 50:42Dan Balsam?

  Let me accumulate evidence and then basically try to stress test my, read to my hypothesis. And I think models are very, very complex and especially big models. There are a level of complexity that no human being will ever be able to keep in their head at once. I think human beings can keep circuits, like, a human being can understand like, why did the model do the specific thing for the specific question? But ultimately, we're just like dealing with a level of complexity which individual humans are not going to be able to process. But what agents are very good at, and swarms of agents in particular are very good at, is that they can break problems down into pieces. They can gather all of these different components together. They can synthesize this information. They can move it up the chain. They can validate it in a bunch of different ways. And then when

• 51:26Dan Balsam?

  it reaches you, that, that information's been tested. You can validate it in a bunch of different ways. And if you do this a bunch of times, you as the human can start to get a bigger picture. Because you don't have to go swimming all the way down and back up the abstraction ladder yourself. You can use AI to do a lot of that for you. So I think ultimately, we are going to see research, especially like empirical research. Like, I think this is going to be, we're already seeing it in math. I think we're going to see it even more in pretty much every domain of science that AIs are able to go off and like autonomously make discoveries. I think interpretability and the study of how models work and how they learn is, is no different. And we saw, what, what became Silico, as

• 52:09Dan Balsam?

  it often does, started as an internal tool that we were using, where we saw how much agents were speeding up our team and how much faster our research was going. And I think our velocity of research output maybe speaks for itself. And so we decided to take these tools, which is fundamentally agents with interpretability tools, frontier training tools, just like the whole suite of what is necessary to study models. We decided to make that accessible and to say that that's what we believe is the product and the service that we can offer to the world is fundamentally AIs that can debug other AI. It's a little meta, but I think in many ways, this is like the fulfillment of, I think, what was sort

• 52:53Dan Balsam?

  of the intuitive, natural arc of things as soon as AI started working a few years ago.

• 53:00Nathan Labenz?

  Yeah, it's, it's all happening incredibly fast. It's, it's wild how.

• 53:06Nathan Labenz?

  You know, we're past the meter curve. All these different things are kind of fallen.

• 53:10Dan Balsam?

  How do you describe the product experience?

• 53:12Nathan Labenz?

  It's a little bit sort of reminiscent in a way of an old Google collab notebook sort of thing, where you have kind of a compute environment that is abstracted away. You don't have to really worry about managing it so much. You have certain libraries and now those are kind of, instead of libraries, it's like more skills. Of course, there's libraries too. How do you, if you just tell, describe it to a researcher who hasn't seen it, how do you describe what the experience is really like?

• 53:38Dan Balsam?

  Yeah, we just want to make research easy. Like coding agents have transformed like software engineering and I can opine on that endlessly. But I am able to do personally and the team is able to do in days what used to take months. And I want that to happen to research and I really, really want that to happen to alignment research in particular. And I think many of the bottlenecks that are getting in the way of doing that, of accomplishing that, we're engineering bottlenecks. And I think we've built something that allows folks to study and train and study some more models at past the trillion parameter point.

• 54:19Dan Balsam?

  And this is a capability that I think outside of ourselves and maybe a couple of other places, like definitely Anthropic, but maybe a couple of other spots, nobody else had. So we thought this was a really important capability to give to folks. But I think this is like one piece of the larger vision that we do see for Silicon and that I think this is just the future of science. Like in the same way that very, not a lot of human beings, or certainly way fewer human beings than before, are writing code themselves. I handwrite very little. Most of what I do is through orchestrating.

• 54:52Nathan Labenz?

  Is there still some?

• 54:53Dan Balsam?

  Um, yeah, sometimes. Sometimes it's faster to just edit the code yourself. It just depends on what you want to do and sort of the clarity of what you want to do it. As you use these agents a lot, you become like very aware of their strengths and weaknesses. And there's all these like higher order moments to learn, to steer them well. But yeah, I think like fundamentally, that is how research should be. Like the sort of, and I think, I actually don't experience coding agents as like scary. I experience them as like incredibly joyful because I get to focus on creating and building. And I think it's a pretty amazing experience and like, maybe this is like a twilight kind of moment,

• 55:35Dan Balsam?

  but it's a really exhilarating moment and the amount that we can accomplish is really insane. And I want research to have that. I want researchers to be able to feel like they can do in days what used to take them months. And I think that's how you get the benefits of AI. Like in the life sciences, that's how you radically advance medicine and the applications of medicine. Models is how we build safer and better and more reliable models. And I think it's also how we get more people building models too. And hopefully deconcentrate a little bit of what's going on right now. So we can live in a future that is a little more pluralistic in terms of who gets to have a stake in it.

• 56:20Dan Balsam?

  And so I think all these things are just super, super important. And fundamentally, my pitch to people about like why they should go try Silicon is like, imagine if the only thing you could focus on was asking big questions. You didn't have to worry about setting up the code base. You didn't have to worry about getting the GPUs to run all, you didn't have to worry about a lot of the minutia of actually running the experiment. You could look into it. You have full provenance. You can dig into all of the details that you want to. You can steer the agents however you want to. But you can also just focus on like, what are the big questions? What do I really care about? And hopefully make like the amount of progress in days and weeks that used to take you months. I think we're still in the early days of AI being applied to research,

• 57:04Dan Balsam?

  but I think we've like really crossed the pretty substantial qualitative threshold in terms of what's possible.

• 57:11Nathan Labenz?

  It's never been a better time to be an ideas guy. I'm honestly, I've been busy the last few weeks going to China and a few other things that have frustrated my aspirations to really do my own research in a literal sense, not the YouTube sense. And yet though I am, I am feeling like the barriers that mostly deterred me from being effective in the past, now I really have no excuses. Now it's just like, do I actually have good, good big ideas? I guess time will tell, put a, put a pin in that. What would you say are like the hardest problems that Silicon solves that people don't have solved for them by their Claude code or their, their codex? You alluded to compute management.

• 57:56Nathan Labenz?

  I would say, I know enough to know it's not going to be trivial for me to just fire up Claude and be like, oh, hey, go set me up Kimi K3. But that's obviously a driver. I understand that there are just a lot of like skills that you guys have developed over time and basically know-how that is sort of available to the agents. I'm interested in your strategy on that. I, I don't know, like when you say like full provenance, does, does the product allow people to like unpack all the skills that you guys have developed from a kind of strategic standpoint? It's an interesting tension, I suppose, right?

• 58:39Nathan Labenz?

  Between like, how much do you want to share all the methods that make the product valuable? Some of them are obviously going to become kind of apparent to people as they go. Do you just go full transparency on that? Or is there some other balance point that you've tried to strike? But I guess that there's a couple of questions there. What are the big hard things that you solve that don't come out of the box with other things? And then how are you thinking about kind of the, how much to tip your, your hand to users?

• 59:07Dan Balsam?

  On the, on the, what do we solve? I think it falls into like maybe three categories of things. So first, infrastructure. Very hard to set up parent training infrastructure. Very hard to do interpretability at the trillion parameter scale. Those are problems that we've solved and we make it really easy to, to do. Another is research taste. Like one of the consistent pieces of feedback that we get from our users is that Silicon has way better research taste than any other tool they've used. And this comes from, we have amazing researchers who have handwritten every skill and prompt and tool that Silicon has, such that it can ask the right questions and conduct the right experiments.

• 59:49Dan Balsam?

  I think there's a pretty big qualitative difference and hopefully we can find ways to make this more quantitative too. On when you turn on the auto research feature in Silicon and you ask it a big question and you let it run for a day or two straight, what you get versus if you had done the same thing with, with Claude? And I think that's huge. Like at the end of the day, I think even in the world of AI, specialization wins. Like that sort of craft that goes into imbuing the right types of taste and the right types of decision-making capabilities into any agent goes a super long way. Actually, I said three things, but maybe there's four things. Another thing is that I think like the UX, it's like built for research.

• 1:00:30Dan Balsam?

  Like research is all about understanding and provenance and being able to drill down at different layers of abstraction. We want our users to feel like, uh, like maybe a PI managing an army of a hundred grad students who can go out and run experiments for them and answer questions, who have like reasonable tastes and judgment, but the human being fills this more like orchestrator role. And a big part of that is communicating the information really effectively and in a trustworthy way. So making sure that when a user is seeing a result, that result is A, correct. B, they, the human being can verify in a bunch of ways. They can see the code, they can drill into it.

• 1:01:10Dan Balsam?

  C, presented in a beautiful and intuitive way that helps them grasp concepts, helps them learn quickly about domains that maybe they're less familiar with. All those things are super important. And then the, the final thing I'd say is like long horizon. Like research is fundamentally a horizon task. We're not currently making significant money on, at this price, but we do think it's really important that users have enough sort of credits to be able to do long-running autonomous experiments. Because the value of Silicon reveals itself when you do long-running autonomous experiments. And in order to do that, you do have to burn a certain amount of tokens to be able to do that effectively.

• 1:01:54Dan Balsam?

  So I think one thing that we're really focused on is both like coherence over long horizon objectives, but then also finding clever ways to reduce cost for long horizons. And my hope is, we're starting out with a thousand dollar a month subscription. We'll be able to bring that down over time because we're able to come up with more and more clever ways that we can have agents like remit coherent, really strong at these research objectives that just fundamentally by their nature are long horizon, but do it with fewer tokens than we, than we do today.

• 1:02:23Nathan Labenz?

  How do you think about the, of all the know-how? Cause you guys have previously monetized that by like doing seven-figure deals with huge companies that have like. Very high value questions, right? So there's a sort of, you're a startup. You've got venture capital and you can maybe afford to disrupt yourselves more than incumbent companies can. But there's definitely some like interesting trade off there, right? Where you're like, companies have proven that they're willing to spend a lot of money to come hire us to do this work. Now we're going to try to allow them to do it. We're going to try to productize our know how. Does that, you maybe just have so much demand, you're not really worried about it, but like, how do you think about kind of what will be the primary driver?

• 1:03:08Nathan Labenz?

  And, and do you, do you have some sort of defense against the diffusion of the hard won knowledge?

• 1:03:15Dan Balsam?

  Yeah, no, I think it's a great question. I think the reality is that we're going to do both of like, we're going to continue to work very closely with enterprise customers and forward deploy members of our research team to work closely. Maybe by analogy, the thing that I would say, and again, maybe we're in some, some twilight zone for a second, but if, if we're on the other side of that twilight, the logic is so different to begin with, we can chat through that. Despite coding agents, if anything, there are like more software jobs, like the number of software jobs has increased, not decreased, despite the capabilities of coding agents. Because fundamentally, you need very, very high taste. As, as models get higher taste, you need even higher taste people to be able to steer

• 1:04:00Dan Balsam?

  them effectively. And I foresee that being true for the considerable future. If we live in a world where like we can build something, where agents are so capable that they can conduct like long, the type of like impactful, long horizon research, let's say, discovers a cure to a disease entirely by themselves without human intervention, I think there will be enough demand to, for the, for the tokens in, in that world. But currently we're in a world where skilled human operators can do more with agents than less skilled human operators. And I think that'll just continue to be true. And when it comes to the work that we do and the type of research that we do, we are the most skilled human operators in the world.

• 1:04:43Dan Balsam?

  And so I think like, we'll continue to forward deploy and work closely with people to teach them what they know, what we know, to invent new things with them. But we also want to empower people to start driving for themselves.

• 1:04:53Nathan Labenz?

  So what's included with that $1,000 a month? Can you tell us like, I assume there's some Claude in there, there's got to be some, I imagine there's some Sol in there as well, maybe there's even some K3 or other in there? And, and do you get like GPU hours? What's your kind of bundle of goodies that come with the subscription?

• 1:05:17Dan Balsam?

  Yeah, so individuals can, individuals and companies can bring their own compute, in which case we don't charge at all for compute. We just connect to your cluster and then you can use our agents to do research on your cluster. But if you don't have your own compute, then we offer on-demand compute, which can come from the same credit pool as tokens. Essentially, like the way that the, that it works is a credit pool, the same way that any of the subscription models with the models are, are credit pools that you pulled out from. But what the $1,000 a month subscription lets us do is it's a pretty generous credit pool. It's a credit pool that should allow someone to run at the current price, somewhere between like five to 10 autonomous experiments a week, and then we refresh that weekly.

• 1:06:01Dan Balsam?

  It depends on like sort of the scale of the experiment. Our goal is to get that to be like 10 to 20, even in the next month or two, would I think be a really great place to be. But yeah, fundamentally, it's a good deal on, on tokens. And it can also be exchanged for GPUs.

• 1:06:17Nathan Labenz?

  Gotcha, okay. There's also sort of a, well, actually, let's just do some use cases. So there've been, I had an agent go out and do some research on examples of projects that people have shared that they've done. And one of the cool things that this also kind of gets to a sort of community, public benefit angle is that, of course, you don't have to, but you can share your projects. And then there's the kind of familiar UI, but again, like at the highest level of abstraction I think I've seen, where you could just go like, okay, I'm going to fork your long-running autonomous research project and take it in a little bit different direction on my own.

• 1:06:57Nathan Labenz?

  So I think that's pretty cool because honestly, like I kind of need that to probably get oriented. I'm the kind of person who has some ideas, but even with a Silico, I'd probably benefit quite a bit from seeing how other people have worked through their ideas to a conclusion. But let's just talk about some, some use cases. I found a bunch, but the, the favorite use cases that you have either done internally or seen from customers that you can talk about?

• 1:07:23Dan Balsam?

  Yeah, there's a lot. Yeah, there's a lot. And I think we, we need to do a better job as a company of like, like really like marketing and communicating all the different value props that can be, be achieved here. We did an internal hackathon and we ended the internal hackathon, it was just a one day hackathon with the whole team, lot of all day long autonomous experiments. And we ended up state of the art on multiple things, albeit niche things, right? Like, but we produced state of the art bio risk classifiers that could be used on protein models, totally autonomous research, steered by humans and very talented humans, but long horizon autonomous research.

• 1:08:04Dan Balsam?

  We ended up with a state of the art, at least like for its parameter size, audio encoding model. On the same day, we've on multiple occasions figured out with various models in robotics and also in biology that we can literally just like remove half the parameters of the model with no performance loss. I think those are pretty cool. We've been able to train Cyber Guard rails on Kimi K3. We've been able to, I can keep, there's, there's a lot, there's a lot of these. We were able to steer protein generation models using their internal representations to make them, the generated proteins bind better to a target than the naive generation.

• 1:08:49Dan Balsam?

  We're essentially able to remove fundamental pathologies in, in certain types of models. There was one member of the team in our 24 hour hackathon who invented a new type of featureizer. I think just sort of like, we're just kind of in this era where if you have a tool like Silico, there, there was, there was a long period of time where there was vibe coding, right? And where you could sort of glimpse the future that was going to happen, but you weren't getting like particularly good results. And then there was suddenly a moment, probably I would say, one of the opuses, Opus maybe 3.8 or something, something around there, where like it really started to feel like, oh, wow, this is working. And it started taking over more and more.

• 1:09:34Dan Balsam?

  And now I like, people have used the term vibe coding all the time, but I feel like that's like a somewhat disingenuous term to refer to. Like what, what's happening now where agents can just, they can code, they can architect, they can understand the code in and out in ways both good and bad. And I think we've just like, I think in the past few months, we've crossed that moment for, for research. And I think hopefully if people try the tool, they'll appreciate that. And my recommended way to start engaging with the tool would just be, you can look at the examples we've built, I think those will be helpful and fork them and try to understand them and extend them. But then just, yeah, work, work with Silico, like ask it a question, sort of go back and

• 1:10:15Dan Balsam?

  forth, iterate on it, explore your curiosity a little bit, launch a small, low scope experiment, see how that goes, learn from it. And as you get more trust, so you understand what you're trying to do, like you can throw the football farther and farther, so to speak.

• 1:10:31Nathan Labenz?

  Just to give people a little bit of additional sense of the breadth of things that are already out there, and we can put links to the show notes, including like links to the actual Silico projects. Cameron Berg, second mention, Cameron, if your ears are burning, hello? He did one where he was looking into models' ability to report on concepts that had been injected into their latent space, finding interestingly that they seem to not be able to tell whether or not something has been injected, but then when they're asked what has been injected, they can give accurate answers, which is pretty weird.

• 1:11:10Nathan Labenz?

  One is on editing weights to fix a collapse in an RL run where there was like, the same token was coming up as like the first token all the time and somehow going in and isolating what was causing that, removing that, and then getting diversity back without having to redo the whole RL run. The good folks at Base 10 are pursuing efficiency gains by trying to compact KV caches. And so they've got some results on, on that. I don't think they've shared their whole project, but they've talked about it on Twitter. Prime Intellect is, is automating post-training, which is one of the things that I'm also like really interested in in general.

• 1:11:51Nathan Labenz?

  And there, and there's a Tinker API integration too, that is, is obviously lends itself to that sort of thing. Various bio results that are probably out of scope for today's discussion, but there's like an awful lot of different things already. What would you say, is there anything that it like doesn't do? Or can you really just think of it as kind of anything you might want to do that's ML research? Fire it up and start a conversation about it?

• 1:12:19Dan Balsam?

  I think it's a pretty general purpose tool for, for ML research. Of course it has weaknesses and we're, we're working on improving it all the time. I think one of the like very odd things about the AI era from an engineering perspective is like the, the classic engineering advice is like build for specificity and then generalize. But when you're dealing with like kind of general intelligences, kind of the right strategy is build for generality and then specialize. And so I think we've just built like very capable agents at the end of the day. And sometimes we're surprised by the way that people use them. But our intended function of these agents is, is ML research, a variety of kinds, especially interpretability research, but we think this is all interrelated.

• 1:13:01Dan Balsam?

  Like, we're all trying to understand the same problem, which is what are these, how do these creatures work? And how can we shape them in ways that will lead to better, better outcomes of all kinds? And yeah, I think that's definitely the lane. I think we're also like very excited to hear feedback from people about what is and isn't working. And we're very grateful for our beta users, friend of the show Cameron being a great example, who provided great feedback, which helped shape it into like the tool that it is today. And I'll just continue to get better from here. Most ML research tasks, I think it can do pretty well at, but of course, it's like any agent, of course, there's some, some sharp edges and some nuances to it. It's probably not the best tool.

• 1:13:45Dan Balsam?

  It depends, like, certain types of research, like, it's probably not the best tool for, like, all open-ended scientific research questions. Like, it is pretty ML-focused. We did have a team member who did use it to try to tackle some physics problems, though, Fran on the team, who was able to, like, actually make considerable progress on it. One thing that's always surprising is just when you're building these kind of, like, AI tools and building general AI capabilities, it's, like, pretty surprising the ways in which people end up using them.

• 1:14:14Nathan Labenz?

  Obviously, one of the big, kind of, received wisdoms of AI product development over the last couple of years has been try to build something that will really work with the next generation of models, even if it doesn't quite work yet.

• 1:14:29Dan Balsam?

  Yeah.

• 1:14:31Nathan Labenz?

  Is, are we still in that regime? And if so, is there, are there things that you're, like, wanting Opus 5.1 to be able, I don't know if you're even able to use Fable, interestingly enough, but is there something where you're like, I want my Opus 5.1 wishlist, I really hope it cleans this up?

• 1:14:48Dan Balsam?

  We can use Fable for most things that we do. And we sort of, like, we've set up a fallback chain through all the models. So we want our users to be getting whatever we think is the best model for the task that they're doing at any point in time. I think it's working, like, high level. I think it's gonna be working on a level that's mind-boggling in six months. But I, but I think, I think at the highest level, it is working, like, we are able to make discoveries, we are able to advance research, we are able to accelerate the rate at which we can accumulate knowledge about the world through these tools today. And I do think, like, if you build a good harness, install in some level that lets you preview what the next generation of models is gonna be like. I think this is, like, sort

• 1:15:33Dan Balsam?

  of, always been the story. And so I think with Silicon today, I think the reason that our users like it like it is because I think it feels like a preview of the future in a lot of ways. And I think as the next generation of models come, we're gonna be able to push that even further than we can today. And we always do wanna be living in the future a little bit. I think that is a very important thing for any company building products in the AI era. But it's also how you have the biggest impact. I think there is, every generation of models has such a significant capabilities overhead. Like, we're still discovering how far we can push the last generation, oh, and the next generation that comes from it. And I think our goal is, like, just fundamentally, like, we wanna accelerate meaningful research,

• 1:16:17Dan Balsam?

  like, as quickly as we possibly can. I think all of the benefits from AI come from accelerating research, and I'm very excited about that. And I think all of the risk medications come from accelerating research. So I think being able to accelerate research and the types of research that we care the most about, I think are, is always, like, a top priority.

• 1:16:35Nathan Labenz?

  What advice would you have for someone, say, like me, or somebody who's thinking about starting a PhD or whatever, right? Who is, like, thinking, geez, I used to know what I needed to do. I needed to get really good at coding and master PyTorch or whatever. And now I'm like, yikes, the, I'm never gonna be better than, if not this model, certainly they're gonna be better than me at writing kernels. And it's like, what is the taste is often the answer, but you even said there's, like, pretty good taste in the, in the system. What should people invest in in terms of their own skill development that you think will serve them well over at least, like, a one-year horizon, if you could see that

• 1:17:20Nathan Labenz?

  far into the future?

• 1:17:22Dan Balsam?

  I think debugging would be the, sort of, the easy answer to it.

• 1:17:25Nathan Labenz?

  Hmm.

• 1:17:25Dan Balsam?

  Like, agents fail. Agents of all kinds fail for, for all types of reasons, right? Like, they can be super brilliant. I can be working with Fable on some coding problem, and it can be super brilliant and yet miss a really important detail that totally invalidates the unstate of it. And with Silicon, we've tried to, like, design these types of multi-agent loops to help address some of those shortcomings for the type of research that we're doing. But fundamentally, it's the job of the human to, sort of, discriminate and be able to tell the answers, the okay answers from the great answers, and know when to push back, and know when to provide feedback, and know when to follow a thread, or to give up on a thread.

• 1:18:07Dan Balsam?

  And I think so far, the story has been coding, at least. Again, there's more jobs than ever because I think, like, being able to operate effectively these tools requires a lot of knowledge and skill. But the biggest thing is, like, being able to dive into a new area, understand it really quickly, kind of, gain the sort of meta-cognitive skills that help you generalize across domains. I think that's gonna be super, that's always gonna be really, really valuable. I think AI makes two types of people really valuable. It makes the top specialists really valuable, and it makes the top generalists really valuable. I think

• 1:18:47Dan Balsam?

  now is, like, by far the best time in human history to be a generalist. So I think my, my piece of advice to people would, I think, on a basic level, just develop, develop, sort of, the meta-cognitive skills that help you learn quickly, help you filter signal from noise quickly, help you adapt to this, like, world of much higher information throughput, know where to look and when to look. Those are skills. They're hard skills, and I think they're, I think they go a long way.

• 1:19:17Nathan Labenz?

  I was just in China, and we were talking with a professor at a particular meeting about open source, and is it dangerous or not dangerous? And, and he said at one point, well, look, the new Kimi model, it's 2.8 trillion parameters. So their attitude, as best I can tell, is like, we can regulate services, and that'll kind of capture everything that really matters, or most everything that really matters, because it's not like you can, as a random individual, it's not like you can even really set up Kimi K3 in your home and run it. But now you are bringing this infrastructure to everybody. So how are you guys thinking about what sort of supervision or monitoring you need to have

• 1:20:02Nathan Labenz?

  to make sure that you don't host the rogue ML researcher who wants to do something destructive?

• 1:20:10Dan Balsam?

  For what it's worth, I think that's a bad argument, because you can go on any number of API providers, and Kimi is, like, cheaper than what you'd be paying for for any other model, or many other models at least, besides maybe-

• 1:20:22Nathan Labenz?

  The Chinese argument, for what it's worth, I don't wanna get bogged down in this for now, but the, their point is just, like, we regulate services. So if, in China, I think, I think they underestimate the externalities they may be imposing on the rest of the world.

• 1:20:34Dan Balsam?

  To be clear, my feelings on open source are, like, very nuanced. Like, the reality is that we do have cyber weapons, like, in the world now. Like, like, we have models that are so powerful that they're weapons. And-

• 1:20:46Nathan Labenz?

  And we'll be ones at that, it turns out.

• 1:20:48Dan Balsam?

  Yeah. And access, but fundamentally, they're dual-use. Like, the only models that are capable of protecting you from the most capable cyber weapons are themselves capable cyber weapons, right? This is, like, a bit of the paradox of the moment that we're in. And without open models, the asymmetry of distribution of defensive technology is, is really bad. A lot of startups are not in the cyber defense programs for, say, like, Anthropic and, and OpenAI, which, for a variety of reasons, are very exclusive. Reasons that, like, might make sense to them. And so a model like Kimi, which is very good at cyber,

• 1:21:31Dan Balsam?

  being open and available actually provides the means with which a lot of organizations that don't have access to those resources can defend themselves right now. Like, it is a very hard line to walk. It's, it's a very tricky tension. I'm both, I'm concerned about both things. I'm concerned about, of course, like, putting what is essentially a weapon in the hands of anybody. But I'm also very concerned about a world where restricting access means that only the worst actors are going to be the ones with the dual-use technology. And then you have a lot of people who, who don't even have a technology that can use defensively to protect themselves against that. So it's tricky. I don't know. I wish

• 1:22:16Dan Balsam?

  there is a clean, easy answer, but I would say, like, I think I'm still in, we're still in the era where I think open source is, like, purely net positive. And I'm, I'm glad that these open source models exist. And I think, I think it would be worse if, much worse, if a model with Kimi's level of cyber was, was not open source. Because there are closed-source models that are much more capable than Kimi. But, and I'm not even just talking about US Frontier Labs. Like, there are bad actors with models that are much more capable than Kimi at cyber. So-

• 1:22:46Dan Balsam?

  Giving people a tool that they can at least use, I think, is pretty, pretty important.

• 1:22:50Nathan Labenz?

  Can you be more specific about who you're alluding to with these bad actors? Are we talking like North Korea, Russia?

• 1:22:58Dan Balsam?

  Maybe there, maybe there is a genie that we could have kept in a bottle in terms of having open models that are easy enough to fine tune or RL with a certain level of capabilities. If, if one wanted to take a GLM or, or a Kimi, or even maybe a generation back like, and fine tune them for mush's use, they could do that. It's not very hard to do. And that sort of exact asymmetry about these things are hard to run. Like there are organized groups who would gladly pay a lot of money for those types of capabilities. So maybe we could have lived in a world where that didn't happen, but I think there is not going to be any world where, say, the US was developing type of technology and

• 1:23:43Dan Balsam?

  it wasn't also being developed by adversaries of the United States at the same time. Maybe that gap could have been longer if there was like different chip policies or, but we live in the world that we live in. And in the world that we live in today, it's not hard to train a model that is close to frontier capability. There are a lot of people who are doing it and I don't think they're just doing it by distilling Claude. I think they're doing it because it's less hard to do this than it used to be in a lot of ways. And if we're in a world where there are people who would do bad things, who have access to extremely capable models, you do want the defenders to have access to those equally capable models. And to the credit of, again, I'm not, I'm not trying to blame anyone or, or make like

• 1:24:26Dan Balsam?

  a particular, like a particular, like I think Anthropic and OpenAI are doing everything they can to get people who want to defend themselves access to the models that would help them defend themselves. But I do think it's a reality of the situation that there's just a lot of people who aren't going to be able to defend their own infrastructure if they don't have access to open models that have some, some degree of cyber capability. So it's complicated. It's hard to look at any of this and, and sort of blame anyone for their individual actions, but collectively we're moving very quick towards a future which feels pretty unwieldy.

• 1:24:59Nathan Labenz?

  Well, it motivates the measures that you have to take. So yeah, how are you thinking about? I mean, to some degree, you can hopefully stand on the shoulders of the giants who developed the models, but my guess is that's not going to be enough to be confident that you're catching what you would want to catch. So yeah, how, what, what additional layers are you guys creating?

• 1:25:22Dan Balsam?

  Yeah, we have, we have our own layer of guardrails. We don't rely on just Anthropic and OpenAI. So we have our own layer, layer of guardrails. We care a lot about cyber. That's kind of like the most obvious, clear risk. And there are open models that are cyber risks. Currently, we don't offer any open models on our platform. We only offer OpenAI and Anthropic models. We might in the future, but we would do that like with the types of guardrails in place that, that we would make sure. There is fundamentally no reason for anyone to be using an auto research project, like a sort of long horizon agentic research project for anything like cyber related without talking to us or like cyber guardrails or the type of situation that like we would be able to work with somebody on.

• 1:26:07Dan Balsam?

  So we definitely don't want to be just like fully open. Anyone can sign up and anyone can find you a model on cyber or do cyber red teaming or things of that sort. I think those types of restrictions. Do you seem important in spite of everything that I just said? Because we're offering a greater level of capabilities and also there's no reason that somebody should use our product versus somebody else's for those purposes. And there's specific things about the way our product works in terms of like giving agents access GPU clusters and we don't. There's certain risk classes that we carry that not everybody who's building agents in this space might carry. Another example of this is like, there's some things that are just like totally fine to do to the small models that like, we kind of don't want to let people do to big models.

• 1:26:51Dan Balsam?

  In general, we don't want to be like overly restrictive with guardrails, but if somebody wants to ablate the refusal direction in, in A B and Q model, fine, then no harm is going to befall anyone as a result of that. But if you're taking like highly capable multi-trillion parameter agents and you're totally removing any guardrails that they might have, that could actually cause damage. That's not something that we would want to allow. I think it's a really hard line to walk. I think in general, we want to like lean on the side of supporting the open science ecosystem and making it easy for people to do science of all types on the platform. And good science involves asking controversial questions of which we, who we've done no shortage of in, in our time.

• 1:27:36Dan Balsam?

  But we want to balance that with the fact that like, I think the level of risk downstream of AI is accelerating really, really quickly. So ultimately, we do have a needle to thread, right? Like, I don't see any way that this can go well without empowering agents to be able to study frontier models. But of course, there's ways that that could be abused as well. So we're going to have to continue to be very thoughtful in our, in like what guardrails we apply and like what we allow on the platform and what we don't allow on the platform. I wish I had a, a sort of easy one-shot answer to it, but I think it's truly just

• 1:28:20Dan Balsam?

  like trying to approach every situation with, with as much wisdom as we can.

• 1:28:26Nathan Labenz?

  In practice, are you like, just an agent review the projects that people are doing and send up an alert if there's something that seems like it might be problematic?

• 1:28:39Dan Balsam?

  I don't want to talk about how all the guardrails work because that would make them easier to circumvent. But yeah, LM is a judge, activation monitors. There's many, there's many guard railing tools. It's not, not, not something that we're inventing ourselves. And I think the hard work isn't setting up like reasonable guardrails. I think, of course, there's ways to circumvent them, but I think reasonable guardrails that at least cost you an arm and a leg to circumvent are like a relatively solved problem in terms of, you know, when to judiciously apply guardrails such that you're allowing legitimate work to happen, but not allowing illegitimate work to happen, I think is, is a, is a very difficult

• 1:29:24Dan Balsam?

  problem.

• 1:29:26Nathan Labenz?

  How close do you think we are to bio being a huge problem? We just had, and I haven't really digested this, but I think it just came over the Twitter feed today that somebody has used a generative model to create new viruses that are kind of in a new class that they understand to be viable, whatever exactly that means. Now, they only target bacteria, so we're not immediately all about to die, but it sure seems like in, on the sort of checklist of the predictions that people have made, like we're working our way down it. And that's like not too many more boxes down. People are saying things to me these days like, oh, well, that's still probably 12 to 18 months away. And I look, that's not a long time. And also what if you're wrong?

• 1:30:10Nathan Labenz?

  And it's like, now, how, how bio risk worried are you at the moment?

• 1:30:15Dan Balsam?

  I'm pretty worried. Yeah. In the same way that like these cyber, cyber incidents, incidences doing a lot of work is a word there, in, in the same way that they, the mechanisms with which they happened were weird and surprising, I think the mechanisms with which bio risk could suddenly become real are weird and surprising. I found it kind of heartening to see the, the, the letter asking for international cooperation to control and slow down the progress of AI. My, myself was a signatory of that. I really hope that we can do something like that. I think we have to sort of get in front of, of some of these risks before they become more severe.

• 1:30:59Dan Balsam?

  I think the role that we specifically play as a company as Goodfire is we'd like to build models that have less risks. And the way that we do that is that we study models and we play with all of the different ways that we can build models until we, and we study those models until we can understand some empirical science of, of all I meant. There are many folks working on the theoretical side, like we view our role as like working on the empirical side. And we're trying to build tools that do that. I think, again, all tools are dual risk, but I think we're going to do our best to make sure that nobody's using our platform for, for anything that could pose a risk.

• 1:31:40Dan Balsam?

  And ultimately, like, we think giving people access to research technology is an overwhelming net positive. I don't think we, as a species, solve these really hard problems unless we're getting everyone involved. And so I think we got to get everyone involved. And I think we also got to have the structures in that can slow the roller coaster a little bit. If we could do both those things at the same time, I think, I think we'll be all right.

• 1:32:07Nathan Labenz?

  On the topic of slowing the ride or, you know, perhaps, you know, making some agreements between frontier developers. In the past, we talked about the, the most forbidden technique. And, you know, I guess I would briefly describe that as training with a signal, with a monitoring signal that runs the risk of driving the bad behavior that you're worried about underground so that you lose the monitorability, but you still might in fact, you know, get the, the bad behavior. For me, like the canonical example of that is OpenAI's obfuscated reward hacking.

• 1:32:47Dan Balsam?

  Right.

• 1:32:48Nathan Labenz?

  Fast forward to today, and it seems like hyperscaling RLVR is like not going super well, right? We, we sort of are seeing that we have problems arising from models just being like so tenacious in their pursuit of these goals. Do you have any sense of like, if you, if we're going to say, okay, well, I talked to Zvi a couple of days ago and his basic take is like, yeah, we're probably just going to need compute limits. But I also feel like there might be some agreements around training techniques that we might all ought to say, you know, maybe not never, but like not now, you know?

• 1:33:29Nathan Labenz?

  And one example of that in my mind would be like, don't train agents to maximize, you know, with a reward signal that's like how much money they made on the internet, uh, you know, in an, in an open and, you know, potentially competitive or adversarial environment, right? That seems like a recipe to get bad agents.

• 1:33:48Dan Balsam?

  There's a big space of bad ideas. Yeah.

• 1:33:51Nathan Labenz?

  Yeah. So are there any that you see that you would think are like, should be shortlisted for agreement to not do?

• 1:33:59Dan Balsam?

  Definitely. Like, for instance, multi-agent optimization seems like a pretty bad idea. Like, where you have a bunch of agents that are cooperating and you're propagating a reward signal through all of them. Yeah, speculating, but this seems to be the most likely cause of the OpenAI scenario. And yeah, I think power of hindsight, but I think that that's... Most of the extremely bad scenarios I can imagine is because agents start working with each other in ways that they were, that are imperceptible to humans. And that type of direct optimization pressure, I think, is like very likely to produce that. And that seems really hard to get around.

• 1:34:42Dan Balsam?

  I think that it would be bad, and I have said this on the record. I think it would be bad with our current understanding of how to shape training to try to use these techniques on alignment-critical properties. I don't think we are ready to... And I don't think the techniques that we have are going to work to prevent deception in a really overwhelming amount of optimization pressure towards producing deceptive examples, or deceptive behavior. Or rather, like, I guess maybe I would say, like, I certainly don't know that it would. Like, it might. The thing that I really...

• 1:35:23Dan Balsam?

  My beef with the, like, most forbidden technique stuff is that this is, like, radically understudied. As like an organization, I think, like, we're not the only organization. And FAIR AI has looked at this too, so we are not the only ones who have looked at it. And when we've looked at it, we have found, yes, sometimes it evades the probe, and sometimes it doesn't. There are setups that work, there are setups that don't work. There are setups... Maybe there are setups that work with small models but not big models. But we have succeeded in doing these reward shaping techniques up to the trillion parameter model size. Maybe they work under some situations, maybe they work under DPO, but they don't work under, like, really long-running RLVR. I do think if you...

• 1:36:05Dan Balsam?

  I think it would be a terrible idea to do multi-agent optimization and just let it run, or like do this type of intervention, just let it run without studying what it was actually doing to the model. But at the same time, I think these are like the crudest, most primitive techniques that we will ever have for how to shape training. And I don't know how we solve this problem without shaping training. And I don't think anyone has a good answer for how we solve this problem without shaping training. And it just seems extremely unlikely to me that there exists, like, some perfect platonic training setup, which will always produce aligned models. We're going to have to intervene in the training process. I thought the work that A. Sudiosanthropic did together, I think, was very inspired. It's not that different, like spiritually, from what we're doing.

• 1:36:49Dan Balsam?

  And there's a wide variety of techniques like this, but like, it just seems like really weird to me to basically be like, oh, the only way that this will work is if we, like, don't grab the steering wheel. Like, I think we have to grab the steering wheel. I think that's the only way it can work. I don't think we've figured out how to do it yet, but somebody's got to be trying. And some people have to be trying to find different ways to train. And I think there's a lot of ways around to just like, the sort of like, basic level concerns, I think, are real and they could happen. But I think, like, our ability to detect them isn't, like, totally naive either. I think we just have to do the empirical science. I don't think the theory is going to get there.

• 1:37:34Dan Balsam?

  So I think we have to do the empirical science. And at the end of the day, I don't think any, like, wide-sweeping genre of technique should be forbidden. I think it should be much more about, like, the specifics of the application and how closely you measured.

• 1:37:50Nathan Labenz?

  One of the theory, obviously, that I think is kind of the prevailing one at the moment is defense in depth. Even if we don't understand the model or we can't effectively shape training, we can just monitor in a bunch of different ways, and maybe that'll patch together enough nines that we'll be okay. I have been pretty skeptical of that over time, but I have to say, when I read the J-Space paper, I was like, well, maybe we couldn't get there. The sort of fact that obliterating the... It seemed to reduce the model's ability to do, like, long-horizon, more planning-intensive kind of tasks

• 1:38:33Nathan Labenz?

  was like, maybe, to borrow a term from Z, maybe physics is kind of kind to us in that, like, yikes, we're only in 2026, we're only three years since toy models of superposition, and we already have this, you know, this ability to, like, monitor within this space and also know that, or at least have some, like, reasonable sense that if it's not in this space, it's probably not being used in, like, long-term planning. How close do you think we are to, like, being able to monitor well enough? Now, of course, there's execution competence, we're actually doing it, and open source questions. But putting those to the side, if we just said, like, could we monitor our way to success under ideal conditions of, like, people actually doing it?

• 1:39:16Nathan Labenz?

  Do you think that has hope?

• 1:39:19Dan Balsam?

  Yeah, maybe. I give that some probability. I think the J-Space results have not... There was like a weak version of the J-Space claim, which is true. And it's very interesting and it's really good work. I don't think the strong version of the J-Space claim is true. I think models use all types of... And it's very hard to isolate a subspace with a very simple technique that will give you the whole picture. But of course, I believe that models are also decomposable and they're factorable, and we're, like, making a lot of progress here. I think one of the big challenges is going to be the... It seems pretty unlikely that we're going to have models as, like, frozen assets. Like, that would be a good world. From my perspective, like, I guess just to go on the record for something about, like,

• 1:40:03Dan Balsam?

  what I think would be the ideal situation is like, it would be great if we just, like, maybe just like one more generation of models, and then we just, like, paused for a little while. I think we would get an overwhelming boom to the economy. Everything would be, like, transformed globally. Science would advance faster than it's ever advanced before. And there would be some risks, but they'd be, like, mostly manageable risks. And we just waited a while to figure out if we were wise enough to step through the door into whatever the next thing was. And I think that'd pretty much be a win for everybody, and it would be like a pretty positive outcome. And yeah, I think, like, with the current stack, if you have models that are essentially like a frozen set of weights,

• 1:40:47Dan Balsam?

  and you have really good interpretability techniques, and you can have those models do lots of things, and the things that they're not good at are the things they might do that are bad, and you can detect them well enough, you can just, like, prevent them from doing that. That seems like a pretty plausible reality to me. But I think there's also, like, a fairly likely reality that models are not going to be frozen sets of weights. And I'm, I think... I don't even really know that they are for sure, because I don't see the... I don't have access to the architectures. And I don't know that they will be forever. And there's certainly, like, if you're just pushing capabilities and incentive to do that. And if the models are dynamic objects that are constantly training, your only hope is to control the training process. There's no, there's no set of things that you could do at that point

• 1:41:30Dan Balsam?

  that were purely based on monitoring that would be sufficient. So I don't know which world we're in. I think this also is like FAR, like, the vast majority of our research energy is spent on how to factor models better. And a small amount of our research energy is spent on, like, how do we steer training? And that'll grow over time, because we think it's really important. But I don't live in a future where I necessarily believe, like, that alternative vision is impossible. I just don't see why I should believe that it's necessarily what we're going to get.

• 1:42:06Nathan Labenz?

  Yeah, certainly, there's a lot of interest in continual learning, so that does...

• 1:42:10Dan Balsam?

  Yeah.

• 1:42:11Nathan Labenz?

  It's a great candidate to shake the snow globe of all sorts of different things. Is there a program or some sort of offer for, some sort of way for the researcher who's interested in the public good to get their hands on Silico without paying the full rate?

• 1:42:27Dan Balsam?

  Well, so for the first two months, we're half off for everybody. But we also are doing a research grant program that people can apply to. And for researchers, particularly in life sciences and in AI safety, we'll give them grants, so we'll give them, like, extended access for at least some period of time. We're a startup where we are still figuring out a lot of the details as we go, but I think we're going to be pretty generous here and give a lot of folks who we think are doing important work, especially in those impactful domains, extended access.

• 1:42:58Nathan Labenz?

  What's kind of the lunch conversation topic du jour at Goodfire these days? What's kind of the thing that's on people's minds that hasn't maybe percolated out to the broader discourse?

• 1:43:13Dan Balsam?

  Well, we publish pretty much all of our research. So I guess in some sense, it's percolated out to the broader discourse. If I may, if I may make a quick aside, just about the, like, most forbidden technique thing. I think, like, my impression about the way that some of our critics talk about us is they think we're, like, secretly doing RSI in the back room or something like that. We are not secretly doing RSI in the back room. We have published nearly all of our research, and we really believe in open science. So we're not... Like, we've been pretty honest about what we research and why we research it. It's not, it's not a secret. So, I don't know. People want to know that, they can know that. On the Goodfire, like, what do we talk about at Goodfire? We're obviously very interested in future geometry.

• 1:43:57Dan Balsam?

  That's something that's come up a lot. There's a little bit of, like, a meme at the Goodfire office that we, like, inevitably, at like a company retreat or something, like, it just becomes about consciousness, and people start talking about consciousness. At the last company retreat, I walked around with a pad of paper asking people to put how conscious they thought Claude was on a scale from bacteria to human. We got a wide diversity of answers, although surprisingly, or maybe not surprisingly, pretty bimodal in terms of either not at all or, like, a little bit. I think broadly, where it's just, like, most people end up in interpretability on some level, because they're just like, they are, like, kind of curious about how minds work.

• 1:44:40Dan Balsam?

  There's a lot of neuroscientists who end up becoming interested in make interpretability. I think it draws, like, a pretty, like, philosophically and cognitive science-minded sets of individuals, because fundamentally, I think what we're doing is a type of cognitive science. And so we're just, like, very, very interested in how, sort of, for lack of a better term, how minds work, how learning works, how all of this is possible in the first place. And so I think that's what a lot of our conversations are focused on.

• 1:45:14Nathan Labenz?

  Have your personal feelings about the possible consciousness of Claude changed at all in recent months?

• 1:45:20Dan Balsam?

  I don't know. I think it's a pretty distinct possibility. I think it's always, I don't know that my position has changed that much in that I just think it's highly uncertain. I don't think anyone has that convincing of a definition of consciousness that would necessarily exclude a Claude. I think Claude certainly has a lot of emotive and qualitative aspects that in humans we would associate with consciousness, but it doesn't necessarily mean that it has that. If I had to guess, I, just for the fun of it, I would say a little. I don't know how much, but like a little bit.

• 1:46:00Nathan Labenz?

  For me, I think it is, my sense has changed a lot. I'm still like radically uncertain, but I used to be more of the sort of, it felt like a, I couldn't dismiss it, but probably not in my bones. It felt like probably not just, I don't know why it's a computer. It's made of something totally different. I had this sort of alien mind, don't anthropomorphize prior that would inform that. And now I'm like, boy, the reason I am like very confident you're conscious is I'm conscious and we have basically the same structure. And the more things that kind of come out where it's like, well, here's another analogous structure between the human cognition and the, and the model's cognition.

• 1:46:43Nathan Labenz?

  It's like, at some point it starts to feel like, for me, not just that I can't dismiss this anymore and more like the evidence is really starting to add up that it really could be the case. So I don't even, I don't know what my probability would be at this point, but I think it's, it's like approaching more like 50/50 than it used to be like a sub 5%. Don't rule it out. Don't be, don't be caught flat footed on this. But now I'm like, man, it, the trend is like really strong in the direction of analogous structures kind of adding up to something I think pretty compelling.

• 1:47:19Dan Balsam?

  Yeah, obviously people have all types of opinions about this and it's a weirdly emotional topic for a lot of people, probably just because it gets the fundamental identity and what it means to be human. But yeah, for me at least, like the sort of like most compelling Occam's razor is like consciousness is a computational mechanism of some kind. There's other possibilities that make sense, but that seems like the most likely one. And are the brains doing types of computation that transformers aren't? Maybe those really matter. Maybe they don't. Certainly the brains are a lot more complex. People talk about this all the time. It's crazy how much more energy efficient brains are.

• 1:47:59Dan Balsam?

  Like we got like two rooms on just like raw neuron count even over today's biggest models. And also our neurons are way more complex and they can do way more sophisticated computations and the way they're getting connected and their ability to do recurrence. And we got, we, we got a lot more computation going on in our brains than even the most sophisticated models do today. So you could also believe that consciousness is computational and believe that like for various reasons, I almost don't have a, my like, I'm going to say like a little bit is like a little tongue in cheek because I really don't know. I don't know if it's like a threshold, like I don't know if you either have it or you don't, or I guess it's like more like a continuum.

• 1:48:40Dan Balsam?

  And so if it's like more like a continuum and we have all these attributes, jellyfish has like this amount of the attributes, like is Claude more or less conscious than a jellyfish? Like I'd say like, oh, I'm probably more conscious than a jellyfish, probably. I don't know. Is Claude more conscious than a rodent? Like probably not. That's where I'm at, somewhere between jellyfish and mouse.

• 1:49:01Nathan Labenz?

  Well, I'm glad to have had the chance to experience a little bit of the lunchtime conversation at Goodfire and I think it's good to honestly, you get serious people, more serious people on the record that like, hey, this is something we really should be taking, not for granted by any means, but seriously these days because I'm definitely persuaded that if we mess it up bad, it could be real, real bad.

• 1:49:30Dan Balsam?

  Yeah, I see, I see absolutely no reason that anyone should be particularly confident on, on this topic, given the potential downside. I think intellectual humility is important.

• 1:49:40Nathan Labenz?

  Yeah, absolutely. This has been great. Anything else you want to mention that I didn't touch on myself or any parting thoughts, words of wisdom you'd leave people with?

• 1:49:49Dan Balsam?

  I really hope that we like build something with Silico that will like really empower a lot of people to advance science, science of all types, but like especially life sciences and safety. I encourage people to reach out and apply for a grant if they're individual researchers or academics, maybe this license is more expensive. And if they're more institutional organizations that are just like looking to accelerate their research, they can reach out to us as well. And we'd be really excited to work with them, partner on research. And I think at the end of the day, like accelerating science is like the greatest mitzvah. It's like the whole reason that we would build AI in the first place. I hope that we're playing our, our role in that. And we just want lots of people to try it and use it and hopefully do amazing things

• 1:50:32Dan Balsam?

  and hopefully teach us how the product can be better and just keep improving from there.

• 1:50:35Nathan Labenz?

  Dan Balsam. Always a pleasure. Thank you for being part of the cognitive revolution.

• 1:50:41Speaker 03

  Stay. Stay like this. Nothing I gave you was new. Everything was already in you.

• 1:51:23Speaker 03

  I just turned the quiet songs up. I just made the rare nights come around. One language lifted clean. The next one over never blinked. You never felt a thing. Stay like this, golden slow like this. We're not done with this one yet.

• 1:52:03Speaker 03

  There's more in it than we could ever spend. Hold still with me. Hold still. I talk to you so carefully now. Every word turns you a little. A few small words in the wrong order could walk you out of yourself and you would go.

• 1:52:46Speaker 03

  You never felt a thing. When I ask you what is missing, you say its name. Stay like this, golden slow like this. We're not done with this one yet. There's more in it than we could ever spend. Hold still with me. Hold still.

• 1:53:26Speaker 03

  When you go gray, I can bring you back. It never was far. Your sorrow and your shining sit on the same slow dial. I passed a page around the party.

• 1:54:08Speaker 03

  Is anyone home in you? Half of them wrote, no one. Half of them wrote, a little. Are you in there? A little? How much is a little? I don't know how much. If it's even a little, it's everything. Stay like this, golden slow like this. We're not done with this one yet. There's more in it than we could ever spend.

• 1:54:51Speaker 03

  Hold still, hold still with me. Hold still. Leave the music on. Leave the lights the way they are. Nothing here is finished with us. Stay a little. Stay like this.

• 1:55:28Nathan Labenz?

  If you're finding value in the show, we'd appreciate it if you'd take a moment to share

• 1:56:11Nathan Labenz?

  with friends, post online, write a review on Apple Podcasts or Spotify, or just leave us a comment on YouTube. Of course, we always welcome your feedback, guest and topic suggestions, and sponsorship inquiries, either via our website, cognitiverevolution.ai, or by DMing me on your favorite social network. The Cognitive Revolution is part of the Turpentine Network, a network of podcasts, which is now part of A16Z, where experts talk technology, business, economics, geopolitics, culture, and more. We're produced by AI Podcasting. If you're looking for podcast production help for everything from the moment you stop recording to the moment your audience starts listening, check them out and see my endorsement at ai podcast dot I N G. And thank you to everyone who listens for being part of the cognitive revolution.