Training an AI Scientist with Feedback from Reality, w- Liam Fedus & Ekin Dogus Cubuk (from a16z)

Today’s special crosspost features a16z General Partner Anjney Midha with Liam Fedus and Ekin Dogus Cubuk of Periodic Labs. They discuss experiments in training frontier AI models on real-world physics lab experiments—rather than digital reward functions—and how coupling language models with automated materials synthesis could accelerate the discovery of room-temperature superconductors and transform R&D in advanced manufacturing. Shownotes brought to you by Notion AI Meeting Notes - try one month for free at: ⁠https://notion.com/lp/nathan Core innovation: Periodic Labs is creating a "physically grounded reward function" by integrating experiments directly with AI systems. Physical lab integration: Unlike traditional AI labs, Periodic includes actual physical experimentation facilities where AI can run tests in the real world. Novel approach to training: Periodic takes pre-trained language models and further trains them on physics and chemistry knowledge to improve performance in scientific domains. Read the full transcript here: https://storage.aipodcast.ing/cache/combinedtranscript-20251002T100327544Z.html Sponsors: Linear: Linear is the system for modern product development. Nearly every AI company you've heard of is using Linear to build products. Get 6 months of Linear Business for free at: https://linear.app/tcr AGNTCY: AGNTCY is dropping code, specs, and services. Visit AGNTCY.org. Visit Outshift Internet of Agents Claude: Claude is the AI collaborator that understands your entire workflow and thinks with you to tackle complex problems like coding and business strategy. Sign up and get 50% off your first 3 months of Claude Pro at https://claude.ai/tcr Shopify: Shopify powers millions of businesses worldwide, handling 10% of U. S. e-commerce. With hundreds of templates, AI tools for product descriptions, and seamless marketing campaign creation, it's like having a design studio and marketing team in one. Start your $1/month trial today at https://shopify.com/cognitive PRODUCED BY: https://aipodcast.ing

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 pleased to share a special crosspost from the A16Z podcast, featuring A16Z general partner Anjney Midda, who also recently joined me on The Cognitive Revolution to discuss sovereign AI. Today, in conversation with Liam Fettis, former VP of Post Training Research and co-creator of ChatGPT at OpenAI, and Ekin Dogus-Chubuk, former head of Material Science and Chemistry Research at Google DeepMind. Together, they've co-founded Periodic Labs and just announced a $300 million seed investment led by Andreessen Horowitz. Before diving in, a quick note: while Turpentine was recently acquired by A16Z, my editorial independence remains unchanged.

• 0:44Nathan Labenz?

  And I'm sharing this episode simply because I think it does offer a really valuable perspective on the future of AI-powered science. Regular listeners will no doubt notice some overlap between this conversation and our recent episode with Radical AI. Both companies believe that there simply isn't enough high-quality experimental data in the existing scientific literature to train foundation models for physics and chemistry. And both have raised serious capital to build automated physical laboratories that are meant to connect AI-generated hypotheses directly to real-world experiments, using feedback from physical reality as the reinforcement learning signal, with the goal of teaching AI models a form of scientific intuition and thereby accelerating scientific progress itself.

• 1:28Nathan Labenz?

  Of course, there are still many possible ways to focus such an ambitious project. And while Radical AI has recently announced a contract with the US Air Force to develop high-entropy alloys for use in hypersonic aviation, Periodic Labs has set the goal of discovering a high-temperature superconductor as their North Star, with the expectation that to get there, they'll need to achieve countless sub-goals along the way, including autonomous synthesis and autonomous characterization.

• 1:55Nathan Labenz?

  Importantly, while the science and macro strategies are similar, the conversations are actually quite different. Whereas I tend to explore the technical details in arguably tedious depth, Anjane focuses much more on the human and organizational dimensions of building an AI for science company. As you'll hear, because no human comes close to holding all of the scientific knowledge and intuition that Periodic Labs hopes to train into their AI systems, they prioritize people with intense curiosity and mission alignment, and they don't require advanced degrees. They take pride in their no stupid questions culture, and they host weekly teaching sessions in which ML researchers, physicists, and chemists can all learn from one another. And recognizing that even $300

• 2:37Nathan Labenz?

  million won't be enough to achieve their ultimate goals, and that even a wildly successful company is only one part of the broader scientific ecosystem, they have thoughtful plans to commercialize their progress in the form of an intelligence layer for advanced manufacturing companies, while also starting a grant program, even at this early stage, meant to elicit key contributions from academia. Overall, I love the vision and ambition on display here, and I admire the conviction with which A16Z and others are backing it. And while this doesn't come up in the episode, I've long believed that long-term AI safety might best be achieved by creating domain-specific super intelligences, which would mean that the AIs that advance fundamental science don't need to have advanced theory of mind

• 3:20Nathan Labenz?

  or persuasion skills. And in any case, as much fun as I'm having playing around with Sora 2, it does seem quite clear that a future of truly radical abundance requires AI systems that go beyond the digital world and iterate directly against nature's own ground truth. With that, I hope you enjoy this conversation about building an AI research company meant to develop systems that autonomously explore and deeply understand the physical world. From the A16Z Podcast with hosts Anjane Mida, and Liam Fedus, and Ekin Doges-Chubuk, founders of Periodic Labs.

• 3:55Speaker 01

  Ultimately, science is driven against experiment in the real world. And so that's what we're doing with Periodic Labs. We're taking these precursor technologies and we're saying, okay, if you care about advancing science, we need to have experiment in the loop.

• 4:09Speaker 02

  The applications of building an AI physicist, for lack of a better word, that can design the real world are so broad. You can apply them to advanced manufacturing. You can apply them to material science, to chemistry, any process where there's R and D with the physical world required. It seems like we'll benefit from breakthroughs that Periodic is working on.

• 4:28Speaker 03

  For example, if we could find a 200 Kelvin superconductor, even before we make any product with it, to be able to see such quantum effects at such high temperatures, I think would be such an update to people's view of how they see the universe.

• 4:44Speaker 02

  So Liam, you were the co-creator of ChatGPT. Doz, you were running some of the physics teams at DeepMind.

• 4:51Speaker 02

  Let's talk about how you guys met and what was the moment where you realized that you guys had to leave both of those labs to start Periodic?

• 5:01Speaker 01

  I believe we met eight years ago at Google Brain flipping over a large tire.

• 5:06Speaker 02

  Yep.

• 5:08Speaker 01

  At the Google.

• 5:09Speaker 02

  You gotta give us more on that story.

• 5:13Speaker 01

  So Google Rails was one of the gems at the Google facilities. And I think that's where Doz and I met. And there was just this massive tire that like a single person basically can't flip, single, like by themselves. And so Doz was trying to flip it and he like pulled me over. He's like, I think the two of us could do it, so.

• 5:31Speaker 02

  And why were you trying to flip this tire?

• 5:35Speaker 01

  Why not?

• 5:38Speaker 03

  But I tried doing it, I couldn't do it. And then I was like, who's the strongest person I can find? And I was either Barrett or Liam. And I thought Liam, and it worked, we just flipped it.

• 5:47Speaker 01

  Right.

• 5:48Speaker 02

  And was that the moment where you guys both realized you had physics backgrounds? How did that happen? What, how did you go from flipping tires to flipping experiments?

• 6:00Speaker 03

  Yeah, I mean, so I don't know if Liam remembers this, but we would catch up, you know, over the years. And we would often end up talking about quantum mechanics or like superconductivity. This was like very common, but I never thought that we'd end up working on physics together. So Liam was working on L L M's and they were going really well. And I was not using L L M's, but I was noticing that L L M's are becoming more and more impactful in my work. So one way it was becoming impactful is when I was trying to remember some things about chemistry, physics. I could just talk to the chatbot and I actually learned a lot of stuff I forgot. Another way was, of course, coding, like we were writing simulations and the L M was so helpful in writing these simulations for us. So then the question was like, can we use L L M's kind of more as a first class citizen in the physics research?

• 6:44Speaker 01

  Yeah, I think kind of leading up to this decision to leave. Doj and I were just, you know, connecting and talking about these different tech trees. We're looking at the improvements on language models, on reasoning. We're seeing what high compute reinforcement learning can do. And on the material science side, we're seeing scaling laws within physics, within chemistry. Both with respect to simulations, with respect to experiment, and it's like the same kind of principles at play in M L. And I think to both of us and to a lot of people in the field, the goal of this technology is accelerate science, accelerate physical R and D. It's, you know, chatbots was like a great milestone along the way, but we really want to see technology out in the world. And we felt like this was just the right place to begin.

• 7:30Speaker 01

  Physics is very verifiable, it's a great reward function, fairly fast iteration loop. You have simulators for large classes of physical systems. And we felt like in order to create this A I scientist, this is like the beginning of of this path. So built that conviction and decided to found periodic.

• 7:49Speaker 03

  Well, let's take a second to talk about what periodic is and what does it do? So periodic labs is a frontier A I research lab. This trying to use L L M's to advance physics and chemistry. We feel like having experiment in the loop tightly coupled with simulations and L M's is extremely important. So we're building up a lab that will generate high throughput, high quality data. And we will use L L M's and simulations in conjunction with the experiments to try to iterate. Science by its nature is an iterative.

• 8:26Speaker 03

  Direction and we feel like LLMs using all these tools that are available to humans can do a great job in accelerating physical R and D.

• 8:34Speaker 01

  I'd say the objective is let's replace the reward function from math graders and code graders that we're using today. So like math graders, you know, to give an example, you have a prompt, what is two plus two? You know, the ground truth is four. You can put a lot of optimization pressure against problems like that that are programmatically checkable. And what we're doing, and by having the lab, is we create a physically grounded reward function. That becomes the basis on which we're optimizing against. And so if a simulator has some deficiencies or some issues, we always error correct because for us, the ground truth is the experiment, like the RL environment, nature like is our RL environment

• 9:19Speaker 01

  in our setting.

• 9:21Speaker 02

  Let's just take a second for folks who might not be familiar to explain what you guys mean by a lab that will verify RL in the real world. Can you talk a little bit about how experiments work? How are AI models trained today? And how is it, how are those different from how they're going to be trained and developed and post-trained and deployed at periodic? And it might be helpful to talk about how you created ChatGPT.

• 9:48Speaker 01

  So ChatGPT originally, the technology evolved very rapidly over the, over the last few years. When we were first creating it, it was a very standard RLHF pipeline. So you have a pre-trained model and it's sort of like this like raw substrate. And what you're trying to do is take this auto-completion model and turn it into something useful. The way we did it at that point was we would have supervised data. So given some input, we would say this is a desired output. So if we're trying to get it to act as an assistant, you know, we create some tuples like that. Then you run reinforcement learning, but now you're learning against a reward function that's trained against human preferences.

• 10:29Speaker 01

  So humans will say, well, given this input, I would prefer completion A to completion B. And you do that over and over again, and you can create a reward function that can then be optimized against. That is sort of the basis of how we created ChatGPT.

• 10:45Speaker 01

  But then there's a huge gap between the original model and what we have today. And I think part of that is reasoning, but also part of that is just much better, more precise reward functions. So the reward functions that we were using originally couldn't determine whether you were mathematically correct or not. So early versions of ChatGPT were mathematically not particularly strong. And it's sort of results from the reward function. What did you optimize against? You, the reward function basically encoded be a friendly assistant, try to help people get to their thing. But it had no sense of is this mathematically correct or not? Is this code valid or not? And we made huge advances over the correctness of our reward functions.

• 11:30Speaker 01

  But this is all digital. We're creating tasks based on the internet, textbooks, papers. And this is great. This lays a foundation, but ultimately science is driven against experiment in the real world. And so that's what we're doing with periodic labs. We're taking these precursor technologies and we're saying, okay, if you care about advancing science, we need to have experiment in the loop and that becomes our reward function for our agents. So as Doja was saying, our agents are doing the same type of things you'd use for coding or to help answer a query. But now instead of just giving tools like here's Python, here's a browser.

• 12:12Speaker 01

  Now we have tools like quantum mechanics, so simulate different systems. But ultimately we're going to a lab and then that becomes like the basis of what is the system optimizing against? So that's sort of just like the natural like end state of these systems.

• 12:27Speaker 02

  People in AI often say lab. Often what they're referring to is quite different from what you guys mean by lab. Doja, what's the difference?

• 12:35Speaker 03

  That's right. So as you mentioned, so far the LMS have gotten really good at logic and math. There is like verifiable rewards. What is like the next frontier in terms of, you know, inquiry after logic and math? I'd say it's physics. And then when you say physics, there are different energy scales. So there's astrophysics, studying galaxies, there's fusion, nuclear physics, but then there's. There's the energy scale of physics that's more relevant to our life. And that's the quantum mechanics, like Schrodinger's equation. This is where, you know, biology happens, chemistry around us happens, materials happen. So we felt like our first lab should be basically probing that quantum mechanical energy scale. Um, and for us, that would be physics at the level of cell state physics, material science and chemistry.

• 13:17Speaker 03

  Uh, the kind of one of the more fundamental ways of making things around us is powder synthesis. So you take powders of existing materials, you mix them and you heat them up to certain temperature and it becomes a new material. Uh, so that's one of our labs. We're gonna have a powder synthesis lab. And turns out this is one of those methods where robots can do it, like very cheap, simple methods. Um, I don't know if you saw this coffee making robot in the SF airport. You know, a robot that's basically at that level can mix powders and put it in a furnace. Um, and there's a very rich field, so you can actually, using their method, discover new superconductors, magnets, you know, all kinds of materials that are very important for technologies around us. But at the core of it is just quantum mechanics.

• 13:59Speaker 03

  Uh, and we feel like teaching these L L M's to be foundation models, but for quantum mechanics, will be the next frontier for L L M's.

• 14:07Speaker 02

  Why haven't the models that are currently out in the world and deployed able to do this?

• 14:14Speaker 03

  Great question. I think, like, as you mentioned earlier, science is by its nature iterative, right? Like even the smartest humans tried many times before they discovered the things they discovered. And I think maybe this is one of the confusing points about L L M's. L L M's can be very smart. But if they're not iterating on science, they won't discover science. You know, to be honest, humans won't either. Like you put a human in a room without any chance to iterate on something, they won't discover anything important. So we feel like the important thing to teach these L L M's is the method of scientific inquiry. So you do simulations, you do theoretical calculations, you do experiments, you get results and the results are probably incorrect or not what you want at first, but you iterate on it.

• 14:57Speaker 03

  And we feel like that hasn't been done yet. So this is what we want to do. But we feel like you have to do it with the real physics, not just the simulation. So this is why we have our own lab where the L L M will have the opportunity to iterate on its understanding of quantum mechanics.

• 15:12Speaker 01

  Fundamentally, machine learning models are good at what you train them to do. And that's sort of like the nature of it. And so if a model is acting badly, you're like, well, did you train it to do that task? Kind of building on Dosh's point, there's sort of like an epistemic uncertainty, this like reducible uncertainty that you aren't really building or collapsing unless you're actually running an experiment. So for instance, one of the engineers on our team was looking at a reported property of some like some physical property in the literature, and it spanned many orders of magnitude. So if I train a system on that, these systems aren't magic, the best it can do is replicate that distribution, but it's really no closer to a deeper understanding of the universe,

• 15:56Speaker 01

  physics, chemistry. Then another point is it's very uncommon to publish negative results. All of the results are basically positive and a valid negative result is very valuable. A negative result could be discarded because it's like, well, that was sloppy science. But there are valid negative results, and that's a learning signal. And this is something that our lab will produce as well. So I think these three things, there's just like noisy data, no negative results, and you need the ability to act in order to actually do science, which is an iterative endeavor. Those are like the core theses of why we need a lab.

• 16:31Speaker 02

  And what might be the core way to measure if periodic progress against that goal in your guys' minds?

• 16:39Speaker 03

  One simple one is, let's say, high temperature superconductivity. What is the highest temperature superconductor we synthesized? Today, the best number for ambient pressure is 135 Kelvin or so. So we'll know very easily if you're doing well, if we can go beyond that number. So that's pretty fundamental. On the more applied side, you know, there's a processing of materials and its effect on the materials' properties. So we can just measure these properties directly. Let's say it's the ductility, it's the toughness, strength of the material. And as we measure it, the LLM will get a very clear signal. It's hard to hack, you know, unlike these other LLM training techniques. It's like, really what you see in real life is the signal that's going to the LLM?

• 17:20Speaker 01

  Yeah, effectively. It's like, do you? Can you design the world around you? So you're like, I need something with this property. Can the system discover and produce that? Both from like a fundamental scientific discovery perspective, but also in industry. So like someone's working in space or defense or semiconductors and like, yeah, we're having these issues. We're trying to achieve this property of this material or this layer. Can the system accelerate the development of those technologies? So it's very grounded. That's how we'll know it's working.

• 17:54Speaker 02

  It feels like the applications of solving, building an AI physicist, for lack of a better word, that can design the real world are so broad. You can apply them to advanced manufacturing. You can apply them to material science, to chemistry, to all, anything that, any process where there's R and D with the physical world required. It seems like we'll benefit from breakthroughs that periodic is working on. Why hasn't it been done before? And what is it about this moment in history that makes it the right time to attack this problem?

• 18:27Speaker 01

  Maybe one comment is difficult.

• 18:30Speaker 02

  What makes it so difficult?

• 18:32Speaker 01

  I mean, I think part of it is the team. So in our view, this has been enabled by frontier technology in the last couple of years. And so Doja and I have been so focused on basically putting together like this N of one team. Like these group of physicists, chemists, simulation experts, and some of the best machine learning researchers in the world have never been part of one concerted effort. And we feel in order to actually achieve this, you need all these expertise. You need these pillars to do this.

• 19:04Speaker 04

  Hey, we'll continue our interview in a moment after a word from our sponsors.

• 19:09Speaker 02

  So when you guys went about designing the team, you know, after you left OpenAI and DeepMind, what was the primary heuristic that you that you used to guide yourself in figuring out who we wanted on the team?

• 19:19Speaker 03

  So in terms of expertise, we wanted to have a LLM expertise cover the experimental expertise and simulation. And for each of these, we wanted to have basically world-class talent. And of course, for each team, there's actually a lot of sub-teams, like it's like a fractal, right? Expertise is very fractal-like. So for the experimental side, we want to cover cell state chemistry, cell state physics, automation, and kind of the more facilities, like the more operational aspects of experiments. On the simulation side, there's the more kind of theoretical physics parts, there's the more kind of coding aspects of simulations. And the LLM side, of course, there's mid-training, RL, infra.

• 20:00Speaker 03

  And yeah, for each of these, we tried to get basically the best people who have innovated in these like sub-pillars.

• 20:08Speaker 01

  The technology that we think is necessary to do it has really just emerged in the last couple of years. And this data isn't like on a Reddit forum or something. Like you need to actually go produce experimental data, simulation data. It's siloed across all of these advanced industries. And many of them, while there's a desire, they may not have knowledge of, you know, some of the most recent techniques that's been driving this recent wave in AI.

• 20:36Speaker 02

  There was a moment in time when models like, or papers like the GPT-3 paper, for example, that, you know, said language models are few-shot learners and proposed the idea of scaling laws. And then there was a follow-up paper, if you guys remember from OpenAI, that was called, I think, Scaling Laws for Generative Modeling. That just showed that as long as you just kept throwing, you scaled up the amount of compute and data in the right combination, you could very predictably improve the performance of these models. And the theory was that if you just kept doing that, you know, at infinitum, there would be a bunch of emergent capabilities. These models would be able to reason about all kinds of problems out of domain, out of distribution.

• 21:21Speaker 02

  Wouldn't that argue, how would you, how would you square the circle with that school of thought that, that you know, naively, the current pre-training and post-training sort of pipelines at most of frontier labs won't just eventually crack physics as well? Why is this idea of physical verification so necessary? And is that school of, is that school of sort of reasoning wrong?

• 21:47Speaker 01

  Yeah. Excellent question. Scaling laws empirically seem to continue to hold, so that's not in question. But I think there's a question of what is this Y-axis? And that test distribution is very different from like what we're talking about. That test distribution, let's say you're pre-training on the internet, might be, you know, a representative sat from the internet, and you'll have these sort of predictable scaling properties. But that's not going to capture that you, you have a very different set of scaling properties with respect to different distributions. So try to make this a little bit more concrete.

• 22:28Speaker 01

  Let's say hypothetically, we're training a coding model, and we have unit tests to provide some reward signal. So the model writes some PR, we check that the unit tests go from failing to passing, and we say, this was successful, we're going to reinforce these things. You might say, you start optimizing this, and now the system is becoming ever more capable of writing code for its own development. And you have this acceleration, you have this kind of takeoff scenario. Code is one of the most promising areas for this, because there's abundant of data online, you have this feedback loop where the system itself can begin to improve itself, and it's,

• 23:11Speaker 01

  it's a very promising technique, and we're all seeing the benefits of advanced coding models, and it's accelerating quickly. However, that model is not going to then cure cancer. The knowledge simply doesn't exist. It doesn't, you need to optimize against the distribution you care about. So that model, while it's going to be a very valuable tool as a software engineer, it may help a cancer researcher do their analysis. It simply doesn't have the data, the knowledge, or the expertise iterating against that environment. And I think that's just sort of like the fundamental belief we have.

• 23:48Speaker 03

  Yeah, I mean, so actually Liam and I worked on this a bit when we were looking at the scaling laws for vision models. And, you know, this also came up a lot in the clip paper from OpenAI. Like, the in-domain generalization and the out-of-domain generalization are monotonically correlated, but it's not linear necessarily. And so what that means is you can keep improving your model and it'll improve as the power law in domain. And for out-of-domain tasks, by which I mean, as Liam said, the things that you're trying to do is a bit different than what's in your training set. It will also improve as a power law, but the slope of that power law may not be good enough. So that you might need to, you know, spend centuries before you get to that one. We saw this in the NORM paper, for example. We published a paper where we saw that

• 24:31Speaker 03

  as you increase the size of your training set, the IID performance, the in-domain performance improves as a power law. Out-of-domain performance also improves as a power law, but depending on what the out-of-domain is, like how far you are from training distribution, that power law might have such a small slope that is basically useless. So this is one of the reasons we feel like the best way to make progress is to make your target as close to your in-domain training set as possible. And the best way of doing this is to basically iterate on changing your training set to be more like what you want to do. So this is one answer. The other one is actually maybe even simpler. The experimental data we want actually doesn't exist. So, for example, if you look at, like you wanna say,

• 25:14Speaker 03

  learn on the experimental data in literature for synthesis. Turns out the formation entropy labels, which is like the energy it takes to basically assemble the atoms in the shape you want, is so high that if you train a machine learning model on it, it's not predictive enough to predict the next one. So and one of the reasons for this is Liam mentioned, people don't usually publish negative results and negative results are usually very context dependent. So what's a negative result for someone might be positive if they do things differently. So yeah, so not only is there this domain shift problem where what you're trying to do might be different than your training set. So the power law won't have the largeness slope you want. But the other problem is for some of these things you want to do, there's no

• 25:59Speaker 03

  data for it. For example, for superconductivity, there is a lot of data sets you can look at, but the noise floor on them is so high that training on them usually doesn't help.

• 26:07Speaker 01

  Doge, me, the entire team are deep believers in scaling up and scaling laws. But it's just do a beeline for the thing you care about. And in our case, we care about advancing science, advancing physical R and D. That's that's sort of like the thesis.

• 26:24Speaker 02

  Is there a tension between being super bitter lesson pill and just throwing more compute at the problem and the, I guess, domain specific pipelines that the lab you guys just described will have to focus on? In the case of periodic, I think you mentioned the first beelines you guys are making are towards superconductivity and magnetism, right? What is it about those domains that make them good candidates for the first for the first few pipelines that periodic is working on? And why are they just are they it stops along the way to an AI physicist that generalizes across all kinds of domains? Or is there a danger of them being essentially off ramps that don't result in sort of a

• 27:05Speaker 02

  the AI sort of scientific superintelligence that that is the North Star for what you guys are doing?

• 27:12Speaker 03

  Yeah, like I feel like, for example, the high temperature superconductivity goal is actually a goal that has so many subgoals in it. It's a bit like when DeepMind and OpenAI started and said, we're going to AGI. But what that meant was they had to do so many things before they got to these cool results. Like for us, if we want to get a high temperature superconductivity, we probably need to get good at autonomous synthesis, autonomous characterization. We need to get good at characterizing different aspects of the material, using the LLM to run the simulations correctly. So it's a North Star and there's so many goals in the on the way that would be very, I think, impactful for the community. That's one reason. Another reason is I feel like high temperature superconductivity is such a Fundamental interesting question.

• 27:55Speaker 03

  For example, if we could find a 200 Kelvin superconductor even before we make any product with it. That in itself says so much about the universe that we didn't know yet. You know, to be able to see such quantum effects at such high temperatures. I think would be such an update to people's view of how they see the universe. So we feel like it'll be really impactful for humanity even before we make a product out of it. I think that's one of the reasons. A technical reason also is superconductivity is a phase transition. So it's pretty robust to some of these details that we cannot simulate yet. So for example, when you make the material, the superconducting temperature usually is more dominated by its kind of crystal fundamental property than the defects or microstructure.

• 28:37Speaker 03

  Whereas there are certain other materials properties where even if the crystal has the property you want, there's so many other factors that you cannot simulate that would prevent you from seeing that property. So superconductivity has this like nice philosophical upside to it, has this technical upside to it. And it's like really rallies both the physicists, like there are people who studied physics for 40 years and really excited about superconductivity. And there are people who've never studied physics, but very excited about superconductivity. It's like quite rare to find a topic that unites the whole team.

• 29:09Speaker 01

  Yeah, I mean, it's like Durs said, it's like we, in order to do this, there's so many foundational pieces to solve. And our tactic is in order to actually get to this goal of AI scientist, you need to make contact, do the full loop somewhere. If you say you're doing this in just like very vague terms, you sort of just end up back on archived papers and textbooks. And so it's really important for us to do the loop, but then create this repeatable process. Like, how do you go from subdomain to subdomain? And there's really interesting questions about how well do the ML systems generalize between these things? What is the generalization of a system between like superconductivity data to magnetism data, for instance?

• 29:53Speaker 01

  And maybe that looks very different than its ability to generalize to fluid mechanics. And I think there's like fundamental arguments to make there. But the goal is create this repeatable system, prove it, and then just go through the different domains that way.

• 30:09Speaker 04

  Hey, we'll continue our interview in a moment after a word from our sponsors.

• 30:13Speaker 02

  So I can see the argument for why cracking room temperature superconductivity from an experimental basis is extraordinarily valuable for humanity. But you guys are building a startup. And to use an analogy for why you need to have a clear medium term path, or shorter medium term path along the way to a North Star that is both commercially viable and net positive to society. What we've seen, for example, with other frontier labs that are working on automating white collar work or software knowledge work is that, you know, there's this North Star of an AI researcher. But that along the way, there were a bunch of sub-goals and so on.

• 30:58Speaker 02

  But a concrete kind of application that opened up a ton of commercial value and benefits for users on the way to that AI researcher was the idea of AI programming, right? Software engineering has become probably the first major domain that's that's caused people to really update their priors about how useful AI models are beyond kind of consumer applications. And in terms of productivity, their impact's been extraordinary just in a few short months. So if the traditional frontier labs' North Star was an AI researcher and the path along the way to get there was programming, AI programming, what is that for periodic?

• 31:37Speaker 01

  Basically, co-pilots for engineers, researchers in advanced industries. So maybe perhaps just being in Silicon Valley, we really think about like computer oriented work. Everything is digital, everything is bits. But there's so many industries, like we were kind of talking about a few, like space, defense, semiconductors, where they're dealing with iteration of materials of physics, and that's part of their workflow. Like, how are they designing these new technologies, these new devices? And in the absence of data, in the absence of good systems, they don't really have particularly good tools. That is our opportunity. And these are massive R and D budgets.

• 32:20Speaker 01

  So, yeah, while high temp superconductivity is a great North Star.

• 32:27Speaker 01

  We very much understand that technology and capital are intertwined. We're going to be able to massively accelerate science if this is a wildly successful commercial entity. And to do so, we want to accelerate advanced manufacturing in all these different industries, become like an intelligence layer in these, for all these teams, to accelerate their workflow and start reducing their iteration time, get them to better solutions more quickly, accelerate their researchers and their engineers.

• 32:54Speaker 02

  Let's click a little bit deeper on that in practice, sort of a day in the life of a periodic team member, where, let's say, half the team, is this roughly right, about half the team are ML scientists with machine learning backgrounds and the remaining half are physical scientists with physics or chemistry backgrounds. How do you start by uniting the cultures, right? How do you take somebody whose primary career so far in work has been experiments in a lab, in wet labs, doing physics and chemistry, and give them an intuition for ML and vice versa? Because, you know, you guys are both physicists who then had the career trajectory where you also had the chance to be at Frontier AI Labs and were part of training systems

• 33:38Speaker 02

  that are now considered sort of landmark, hallmark machine learning systems like ChatGPT, like Gnome. But for others who might be coming from one domain, how do you get the team to build an intuition for the other?

• 33:51Speaker 03

  Yeah, so this is a great question, and we feel like it's actually crucial for us to make sure these teams work very closely with each other. So one of the things we're seeing is the physics and the chemists need to figure out how to teach the LLM how to reason about these things. Because I think the Frontier AI Labs have figured out how to train them on math and logic, but not yet on physics, chemistry. So one thing we're seeing that's been really, I think, productive is the physicists and chemists are thinking about what are the steps we should include in the mid-training, in the RL training, that will teach the LLM how to reason correctly about quantum mechanics, how to reason correctly about these physical systems. Another one, of course, is the LLM researchers are learning quite a bit about the physics,

• 34:35Speaker 03

  the simulation tools, the Goals, so they've been working together really well. We have weekly teaching sessions where the LLM researchers teach, you know, how the RL loops work, how the data cleaning works, and then the physicists and chemists are teaching about different aspects of the science. The history of science, that's also very important. So we feel like this has been going really well. And, you know, one way of looking at this is the things we have to teach the LLM to be able to discover, say, a superconductor include being able to read the literature really well, like read all the papers, the textbooks, find the relevant parts, and then being able to run simulations, theoretical calculations, and then take action, run experiments.

• 35:17Speaker 03

  You know, we feel like this is quite similar to the physical R and D researchers in these companies. They have to read the literature, read maybe internal documents or external documents, and then run simulations, run theoretical calculations, and then actually attempt the thing experimentally, learn from that. So we feel like all the progress we're making towards our internal superconductivity or physics goals actually is making our LLMs much better at serving our customers who are doing very similar workflows.

• 35:47Speaker 01

  Yeah, I think just culture, no stupid questions. You can ask just like the dumbest like physics question, the dumbest ML question. And I mean, there's a few faculty as part of our company and they're actually excellent teachers. So I mean, these like learning sessions have been really fantastic. And another thing I noticed is computer scientists often think in terms of like APIs. So scientists will say something and they're always trying to map it. You're like, okay, well, what's the input, what's the output, what's the target, how do I map that back, what are, and it's always just like this translation. And I think we also have built up, as part of the team, there's people like on these different edges.

• 36:27Speaker 01

  So like if you have a simplex of like, you know, pure ML, LLM, pure experimentalist, pure simulation, there's people who like to live in this inside as well. And so they have been like excellent bridges for translating between these different groups of people. So it's like active learning to like learn the other spaces, creating APIs, and then these kind of bridge connector peoples. I think Doge being an excellent example of that.

• 36:54Speaker 02

  Is it a requirement for somebody who wants to join P R like to have to have an advanced degree in physics or chemistry?

• 37:02Speaker 03

  Absolutely not. You know, one of the jokes we're making is who was the M B A player who was saying that I'm much closer to LeBron James than you are to me? We were saying the opposite of that to candidates because the amount that even our best physicists doesn't know about physics is much bigger than the amount that they know about physics. So for this new candidate, even if they have no background in physics, how much they have to learn about what we're trying to do is actually not that different than how much the best physicists has to learn because there's so much chemistry to learn, so much material science to learn. And I think this is one of the interesting aspects of science today, you know, in the past in eighteen hundreds, there were these physicists that could do so many different things at the frontier.

• 37:41Speaker 03

  Today, we've reached a point where our intellectual knowledge is so large that a leading thinker can usually only advance in one very specific field. And maybe this is actually holding us back because say to discover an amazing superconductor, as we keep going back to this example, you have to know so much about chemistry, physics, synthesis, characterization, and unfortunately, I don't think any human knows enough about all of these. So we have to collaborate. So I think our team is kind of like a small example of this where we have, as Liam said, like a lot of different points in that simplex. And for any person, they have so much to learn, but that's true for basically every other scientist.

• 38:24Speaker 03

  So for example, I supposedly come from the physics side of it, but I've been learning so much more physics because we now have people from different areas of chemistry in the team, different areas of physics. And I think it's true for L L M researchers as well. I mean, they come in, there are aspects of L L M that they probably didn't know until they started working with other researchers in our team. So I think it's a great, and it's like a small example of what we're trying to do with the L L M because we're trying to teach this L L M all these different things that we're learning as researchers. It's like a really fun experience, I think, yeah.

• 38:54Speaker 02

  And what are you finding makes a great researcher at P R like that's different from what might make a great researcher at OpenAI or Anthropic or DeepMind?

• 39:04Speaker 03

  I would say there's

• 39:06Speaker 01

  Very high overlap, but probably one of the biggest determinants is, do you care about this mission? Is accelerating science, is to you, is that like the big goal? And I think looking at the team right now, it's just incredibly mission-driven set of folks who are like, yeah, this is the North Star, let's do that. If someone really wants to improve some mega corps products, yeah, you'd probably be better off at that mega corp in iterating and improving their products. But if you care about scientific discovery, I think Periodic Labs is the best place to do that.

• 39:39Speaker 02

  How big is the team today?

• 39:40Speaker 01

  We're roughly 30, I believe.

• 39:42Speaker 02

  Yeah. And as you think about taking a lot of the research that's going on at the company and deploying that out in the real world, the kinds of customers that we've talked about, space, defense, advanced manufacturing, these are mission-critical industries that are known for being essential to whatever part of the economy they're part of, but often they're not the most, they're not the fastest to adopt new technology. How do you think about deploying the kinds of frontier agents that we've talked about that are great at science, great at physics, in companies or organizations that might

• 40:27Speaker 02

  not be anywhere close to as sophisticated as you are in AI or ML? Is there, do you have a working thesis for how to make sure that the arc of progress is not bottlenecked on deployment? It sounds like you have a fairly good thesis on how to unblock the arc of scientific progress on the research side, but when it comes to deployment, what might be a working theory that you guys are optimistic about that would help get the systems that Periodic is building out into the real world?

• 40:54Speaker 01

  Well...

• 40:57Speaker 01

  Maybe one thing that we've noticed in our conversations with all these companies is they all are looking for their AI strategy. They understand that like the technology is shifting really quickly and they're looking at how they're doing their work and it's not changing as quickly as they think it should be. Some industries also are losing like kind of key expertise in different fields and they're like losing these like senior engineers, senior researchers and they're like, okay, how do we like preserve that?

• 41:25Speaker 01

  But one thesis is understand, it's you know, kind of thinking about these like APIs and thinking about what are the evaluations, what are the biggest bottlenecks for these companies, looking at some of the problems they face and we can map that to our systems and we say, well, we think we can dramatically accelerate this. And so it's not coming in and saying, hey, we're going to transform your fab line on day one, we're going to transform how you're doing everything, forget everything. It's like, no, we're going to solve a really critical problem, well scoped, very clear evaluations, we kind of co-draft that with them and just show them like how powerful this technology can be when you optimize against the thing you care about.

• 42:06Speaker 01

  So, you know, nothing particularly like surprising here, but you know, sort of like a land and expand type method as you might expect, but really looking for who are the biggest promoters within that company and what are the biggest problems, make sure you're solving a very real thing for them and intersect that with where is our technical capability at the highest.

• 42:29Speaker 02

  You know, you were on a call this morning with one of the customers in your pipeline. We don't need to name who, but what was some of the things you heard as their most urgent problems that they'd like for periodic to solve?

• 42:40Speaker 03

  So one of them was simulations, you know, they spend a lot of time training people on some of these simulations they need to use, it's critical for their development.

• 42:50Speaker 03

  Being able to automate those simulations, I think, would be quite enabling. The design process.

• 42:57Speaker 03

  And then, kind of like, some of the small things, like, matching the formats, being able to feed, you know, the simulation results into the design pipeline. All of these seem quite important. And then being able to treat the data together in the same place. What else?

• 43:11Speaker 01

  Well, I think there's a really fundamental question. So a lot of these companies will rely on retrieval. So that's sort of like a super lightweight thing. Someone shows up with a neural net and they're like, great, we'll just retrieve over all of your data and then that's your solution. However, as we've seen with things like ChatGPT and other things, it's when you pre-train on the data, when you actually encode the knowledge into the weights, it's not just a retrieval system. You have a richer, deeper understanding of the material. And I think this is a big fundamental challenge. So, for instance, for this customer, they can give privileges to their employees and have retrieval as acting on behalf, like, the system acts as the user.

• 43:55Speaker 01

  And so you can match those same kind of like privileges for access. But if you start doing pre-training or mid-training on different parts, it's like, well, if you pre-train on every piece of data, that might only be accessible to, say, like, the CEO of that company. So then you have to figure out, how do you sort of bucket that knowledge and create different types of systems? But I think right now, like, we're, after talking with the user, they don't seem to have a great solution for sort of distilling all of the knowledge

• 44:28Speaker 01

  in, into like a single model, or into a set of models. So like going, you know, going beyond retrieval to, you know, proper training. And then I think also, the supervised training they're doing is really akin to like the early days of ChatGPT, where it's like, input, output, you have, you know, a few examples and kind of transforming this new way of thinking. It's like, no, high compute reinforcement learning is really effective. This is how you should think about the strategies it's using. This is how you create effective tool-using towards those problems. And this is how you optimize it effectively.

• 45:03Speaker 03

  Could you describe for folks who may not?

• 45:06Speaker 02

  Be familiar with it, what do you mean by mid-training? Because people are familiar with pre-training, they're familiar with post-training, but in the periodic context, what does mid-training mean?

• 45:13Speaker 01

  Yeah, sorry for the lingo. So I think this term came up years ago where it's like, well, we had pre-training, we had post-training, but sometimes you need to put in a little bit more knowledge. So before search worked really well, there was an issue of freshness. So we had pre-trained models and they have a knowledge cutoff. So there's like a scrape of the internet at that point, but users want more real-time knowledge. So it's like, how do you get that in there? And enter mid-train. Mid-train is basically you're taking new data, new knowledge that's not in the model, and you continue pre-train. And this differs from standard post-training, where post-training typically is more reinforcement learning, supervised learning.

• 45:54Speaker 01

  And the mechanism is basically, or the goal of it is just put a lot of knowledge into the model that doesn't exist before. So that's mid-training in a nutshell.

• 46:04Speaker 02

  And in the periodic context, does that mean essentially going and injecting a ton of custom sort of data from an experimental implementation in a particular customer or particular industry? What is the, what are the sort of the lines, the atomic unit that you guys think will, of mid-training? That will improve the capabilities of the models on problems that they're just terrible at today?

• 46:33Speaker 01

  I mean, it's just, it's all the knowledge. So it's like, you can have very low-level descriptions of physical objects, like crystal structures, for instance. You can also have higher-level semantic descriptions of like, well, this is how I made material XYZ. And trying to get all this data into the model is really valuable. So it's like simulation data, experimental data, none of this exists. And basically putting that knowledge into the model and making sure that these distributions are connected in some way. And what I mean by that is if you just sort of mix together distribution A, B, and C, there's no guarantee of generalization. What you want to hope to see from these systems is

• 47:18Speaker 01

  the inclusion of this other dataset is improving performance on the other datasets. And so these are sort of just like machine learning techniques or machine learning problems to solve. But basically, just make it an expert in physics and chemistry and where it was deficient before.

• 47:34Speaker 02

  You guys both know that I spent some time running evals on a bunch of these models at the Stanford Physics Lab earlier this year, and the results were that the models are terrible at scientific analysis.

• 47:47Speaker 01

  Because they weren't trained to do so.

• 47:48Speaker 02

  Because they weren't trained to do so. But on the other hand, you know, many of the existing research teams working on the general models are investing in trying to make these better. Is there something about the way you're building periodic that gets the draft off of all of that progress in the base models? Or do you have to start everything from scratch and therefore not be able to be composable with advancements happening in the mainline models today?

• 48:10Speaker 03

  Yeah, I mean, we benefit from all of different advances. So one of them is the LLMs are getting better, and we definitely benefit from that because we take a pre-trained model and then mid-train it, you know, high computer. Another one is the physical simulation tools are getting better. They're open sourcing new ways of simulating, new ways of using machine learning to predict properties. So we get to basically utilize all of those. And it seems like machine learning has made such an impact in the physics and chemistry fields that we expect these improvements to continue.

• 48:44Speaker 01

  I think another thing is when we think about tools for agents, we think of like, here's a browser, here's a Python. But increasingly, people think about tools as other neural nets, as other agents. And so if you look at a lot of like physics code, it's not particularly deep. It's not, this isn't competition programming. This is like, kind of like hacky scripts. But you can rely on some of the best systems for, you know, wherever they spike on. So neural net as a tool to these agents is something that immediately accelerates our work. So you don't have to like replicate every, everything.

• 49:24Speaker 02

  There's a historical pattern that a lot of the fundamental research in the physical sciences that we're talking about here, physics, chemistry, biology, has historically been done at university labs.

• 49:40Speaker 01

  Is there a role at all that the university ecosystem you think will play in Periodix's future, or do you think these are just completely divergent paths?

• 49:46Speaker 03

  Absolutely. I mean, so much of the simulation tooling we use have been developed in academia. Many of it is in Europe, for example, a lot of the novel synthesis methods. So we definitely benefit from a lot of these different, very deep, technical progress. Like, for example, a lot of the physical simulation tools are these, you know, complicated for train code that in our team, for example, we don't really like know how to develop very efficiently. But we feel like there's definitely a very deep connection between academia and industry labs. So, for example, recently, a lot of the large-scale simulations have been done in industry labs like Microsoft Deep, Mind and Meta. But a lot of those

• 50:29Speaker 03

  tools have been actually developed in academia and then passed on. So there's actually really nice synergy there.

• 50:35Speaker 01

  I think it adds a few other things, too. So, like you found when you were evaluating models on their ability to do scientific analysis, they were deficient. This was probably, I mean, not a direct goal for those teams training those models. So I think academia and these collaborations say, well, help us inform, what are the important tasks? Like, how do you do this analysis? What skills do we want to put in the model? A skill could be a full analysis, or a skill could be like a smaller primitive as part of a larger analysis. But also, secondarily, it's how do you think? So, one of the physicists was looking at the reasoning strategies of one of our models. He's like, "It's all wrong.

• 51:19Speaker 01

  It's all wrong." And we're like, "What do you mean?" He's like, "No, this should be thinking higher level. It should be thinking in terms of symmetries. This is the book that encodes the thinking strategies. That will be more effective." And of course, your reinforcement learning environment needs to reward those types of strategies. But given some of the most premier scientists are using these strategies, they're likely effective. And these are types of things where it's like an industry-academic partnership can just be so powerful, because industry just simply is blind to these types of analyses, these tools, as well as just this way of thinking.

• 51:53Speaker 03

  Yeah. And there's a way of connecting that to the tooling question as well, because, you know, language is very important. But then, in the human brain, we also see a lot of visual processing, like geometric. So it's plausible that while these LLMs will keep Getting better and better, they'll actually benefit from having a geometric reasoning that's separate. So today, we can do that with equivariant graph neural networks. We can do it with diffusion models that are kind of geometric tools by construction. And the LLM can call them so then it can have both the language aspect, which is very good for, say, synthesis recipe, but also the geometric aspects, which is very good for representing atoms, just design geometries in general.

• 52:33Speaker 02

  So how are you thinking about deepening periodic ties with academic labs?

• 52:37Speaker 03

  Yeah, this is very important for us. So we have two major initiatives in this direction. One of them is we're starting an advisory board. This will be kind of expertise spanning from superconductivity to solid-state chemistry to physics. And we want to make sure, you know, we're in touch with this kind of long-term research directions. A lot of important government funding goes to these groups, and we want to have a tight coupling between what's important for them and us. So this, you know, includes superconductivity expertise, such as Zix Chen from Stanford on the experimental side, and Steve Kjellsson from the theory side. We also have synthesis expertise on the advisory board from Mercury, Kanatsidis from Northwestern University, and Chris Walverton on the high-throughput DFT side. And our second

• 53:22Speaker 03

  initiative is going to be through a grant program. You know, we really want to enable some of this amazing work going on in academia, and some of their work isn't a good fit for industry. You know, it's best done in academia. So we want to kind of accept grant proposals, and we want to enable and support the kind of work that's going to help community, especially in relation to LLMs, agents in synthesis, materials discovery, physics modeling. So, you know, maybe after this show, you can include the link.

• 53:53Speaker 02

  Yeah, we'll include them in the show notes. So for people who might be interested in joining periodic, what are you guys looking for?

• 53:59Speaker 01

  First off, someone deeply curious, someone who really wants to understand the machine learning, the science at a deeper level, who wants to make contact with reality, who wants to advance science. Like, this has to be a driving thing. But also pragmatic. What we're trying to do is incredibly challenging, and someone who has, like, very careful process, and they get to their solution-oriented, they get to goals quickly.

• 54:28Speaker 01

  And really, someone world-class along some dimension. We're looking across all these different pillars, so machine learning, experimentalist, simulation, and people who can bring some sort of innovation on what, how do you create a creative ML system? How do you bring new types of tools or new types of thinking to some of these state-of-the-art models? Someone who can advance simulations and make it more robust and more reliable with experiment.

• 54:59Speaker 03

  Yeah, and maybe one more thing I'd add is Liam and I have been really looking for a sense of urgency in candidates, because we want these technologies not in 10 years. You know, we don't want these LMS to start improving science in 10 years, but we want them ASAP. So if the candidate feels like a sense of urgency for improving these physical systems, discovering these amazing materials, innovating on superconductivity, they would be a good fit.

• 55:22Speaker 01

  Yeah, if you match all these, please reach out.

• 55:25Speaker 02

  All right, it sounds like we got to amp up the speed, the scale of stuff happening at Periodic, and we'll put the courier links in the show notes. Thanks for coming, guys.

• 55:55Nathan Labenz?

  If you're finding value in the show, we'd appreciate it if you'd take a moment to share 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 where experts talk technology, business, economics, geopolitics, culture, and more, which is now a part of A16Z. 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.

• 56:40Nathan Labenz?

  And finally, I encourage you to take a moment to check out our new and improved show notes, which were created automatically by Notion's AI Meeting Notes. AI Meeting Notes captures every detail and breaks down complex concepts, so no idea gets lost. And because AI Meeting Notes lives right in Notion, everything you capture, whether that's meetings, podcasts, interviews, or conversations, lives exactly where you plan, build, and get things done. No switching, no slowdown. Check out Notion's AI Meeting Notes if you want perfect notes that write themselves. And head to the link in our show notes to try Notion's AI Meeting Notes free for 30 days.