EVERY SPOKEN WORD
40 min read · 7,793 words- 0:10 – 2:58
Periodic Labs: founders’ backgrounds and the biggest 11-month reality check
- SPSpeaker
This one's gonna be good. Um, I'm so excited to welcome today Liam Ferriss and Dorje Chubak. Thanks for coming, guys.
- SPSpeaker
Thanks.
- SPSpeaker
Uh, Liam and Dorje are the co-founders of a company called Periodic Labs. How many people here are aware of Periodic Labs? The kids are paying attention. [laughs]
- SPSpeaker
Not bad.
- SPSpeaker
Let's just start with between when-- So, you know, you were part of the-- You helped set up the post-training team at, at, uh, OpenAI, and you were there for the ChatGPT moment, uh, which, you know, arguably, you, you and the team there kind of helped shape that continuous post-training loop. And then Dorje, on your side, you had worked on Genome over at DeepMind. So these two parallel trees were, like, progressing, and then Periodic came together a year ago. We're almost at one year in now, right? No. It's, uh, it's May now. Okay, so we're, like, 11 months in.
- SPSpeaker
Yeah.
- SPSpeaker
Yeah.
- SPSpeaker
We'll call it, like, June, we kinda started.
- SPSpeaker
Okay. So if you had to assess what updates, what priors have you had to update most aggressively over those, the last 11 months relative to the priors you had before starting Periodic?
- SPSpeaker
I have one.
- SPSpeaker
Go ahead.
- SPSpeaker
Um, so when we were founding Periodic, we knew that building up these labs would take some time, especially, like, a high-throughput autonomous lab. So very much in the spirit of AI and machine learning, we're following just scaling these things up. Uh, we think new opportunities in science will come at scale. But we had this kinda cartoon version of Periodic where it's like, okay, well, the first year, obviously, it's gonna take some time to build up the physical infrastructure, so our first year is gonna be in silico. We're gonna be doing design of new materials just strictly computationally. Then year one, we walk up to the high-throughput lab, flip it on, and start doing science, and that's not at all how it went. So instead, it's like creating these small labs, semi-manual, semi-autonomous, uh, allowed us to direct the research program, understand, like, what kind of equipment we wanna scale up to the big lab and just, you know, to your point earlier, close the feedback loop as quickly as possible. Uh, so that was, that was a huge update.
- SPSpeaker
Any-- You, Dorje, do you have any priors you had to update?
- SPSpeaker
I think we were really excited about kinda having these LLMs not just do coding, but impact real atoms. So we were expecting that we'd see crazy results, but I think the results are crazier than we expected.
- SPSpeaker
Such as?
- SPSpeaker
Like there's something amazing happens when you have automated intelligence affecting not just Python, but the actual atoms that surround you. Um, so we've been able to, um, synthesize materials much better than before. We've been able to make computational progress much better than before. Uh, and it's just starting, so it's been really exciting.
- 2:58 – 5:46
The end-to-end “AI ↔ robots ↔ materials” closed loop in Menlo Park
- SPSpeaker
J-- Act- I, I realize it might be helpful to-- I know it's a cartoon version that you described we had, uh, a year ago, but it might be helpful to simplify just so everybody understands what the pipeline is. So let me, let me try explaining, and you guys can red team it. Um, there's a 30,000-square-foot facility in Menlo Park.
- SPSpeaker
I think 40.
- SPSpeaker
40,000? Okay. We've got a 40,000-square-foot facility in Menlo Park. Half the team is former machine learning, uh, folks from OpenAI, DeepMind, and so on, and the other half, uh, is, on the floor is, um, physicists, chemists, and so on from Stanford, MIT, Caltech, and so on. And then just, you know, across the-- So, so that, that's the-- If you had to think about the layout of the lab, half of the square, you know, footage is, is people sitting from two different disciplines, and then the other half is a automated facility where there's robots. Uh, there's, there's AI, um, there's an AI system that you guys have been training, which is the end result of the pre-training, mid-training, post-training pipeline. Those AIs then predict new materials, ideally searching for high-temperature superconductor candidates. Then a robot or set of robots synthesizes those materials into certain forms, powder or whatever it might be. Then there are machines that verify whether those materials have the properties that the AI said it would, and then you pipe that feedback, that verification back into the training loop over and over again. Is that roughly correct?
- SPSpeaker
That's right. And maybe one thing that's not as intuitive is the AI's job isn't to just make these, like, very, uh, grand predictions about what's a good superconductor. But there are lots of small things that you have to do to make progress in science. Like very minor things, like how do you mix powders correctly, or like how do you make sure this impurity isn't there? Or like we often have sample mix-ups. How do you detect a sample mix-up and correct it before it affects results? So AI is actually extremely helpful on those mundane things as well. And to be honest, scientific progress is like a bunch of mundane things attached to each other. So, um, it's been really good for that as well, yeah.
- SPSpeaker
Yeah, I mean, I think the task is much easier than coming up with some candidates, doing this full loop, and you're like, "Okay, do I have a new high-temp superconductor or not?" There's a lot of, uh, local error correction. So you can say, "How good is my system at characterizing some material?" So materials don't come out of the oven with, like, labels on them. You have to figure out what you actually made. So you shoot, like, one process is you shoot X-rays at it, you look at the diffraction pattern. And making sense of these patterns, uh, can be difficult, so that's a system that, like, we focus on building. Um, but then when you kinda get all these little pieces together, then that constitutes the end-to-end loop.
- 5:46 – 8:22
Why an “eval-first” mindset is harder in science (and where semiconductors fit)
- SPSpeaker
Um, I'm gonna connect the dots between this lecture and the previous one. We had Rob Ryachi from DeepMind, um, and R-Rob's been part of the pre-training team on NanoBanana and their world models and so on. And one of the things, if you guys remember, Rob talked about was, uh, wh-when advising them how to, uh, pi-
- SPSpeaker
Execute on their projects. The final project for the class is the one-person Frontier Lab. He said, you know, the, the key is to start, when even if you're trying to build a very general capability, in, in your guys' case, it's a, um, general reasoning engine about the physical world, like physics and chemistry. Uh, his a- advice was start with an eval that's specific, get to the state-of-the-art on that domain or that eval, and then go from there. In your guys' case, that eval has ended up being, or, or that domain has ended up being semiconductors, uh, as a result of superconductors. Can you talk a little bit about why semiconductors is the right place to measure evals, or am I even describing it correctly?
- SPSpeaker
Yeah, I mean, I think, um, maybe I'll, I'll, I'll kind of separate the question. So one is, do you start your ML campaign with an eval? And one thing that we've seen is science has, like, a lot of messiness. It's decision-making under uncertainty, and we've seen huge progress for reasoning models when there's this great verifiability, this objectiveness to it. So a task like what is two plus two, you know the ground truth is four. Uh, the labs have become excellent at throwing a huge amount of computation against these things. Uh, but in science it's often kind of murkier. It's like, well, we have some evidence of, um, from, you know, this instrument. We have some other evidence from adjacent experiments. How do we make sense of that? How do we decide what to do next? Um, and I think w- we've learned at Periodic that it's actually necessary to spend a bit more time learning the shape of the problem, uh, before just like, here's my like eval and just like blinders on, hill climb that. Um, and then m- maybe to the point about semiconductors, um, Periodic has an extremely deep understanding of like atomistic physics and everyone knows semiconductors are shrinking and there's, uh, a really good opportunity to start modeling and sort of a- alleviating some of these big bottlenecks. And every AI company is scooping up as many chips as possible. There's just incredible demand. Every new generation of like logic and memory is running into materials engineering problems, and so we wanna alleviate that. So we wanna accelerate the progress even further. Uh, computation is physical and we wanna help accelerate it.
- 8:22 – 9:54
Superconductors vs semiconductors: choosing the atom–electron interaction frontier
- DCDorje Chubak
I mean, yeah, and from a physics side as well, so, um, what governs superconductivity and what governs semiconductor properties, interfaces are pretty similar. Uh, you know, we have this hypothesis that AGI won't be this magical thing that generalizes to everything. If that's true, then you wanna pick a direction that you think really matters and then push AI in that direction. So we felt like interaction between atoms and electrons is where we, where we wanna be, um, and it's super fun. And that works really well for superconductivity because there's one form of superconductivity that just, uh, phonons and electrons interacting and there's one form of superconductivity that we don't understand yet, but we feel like it's related to phonons, electrons and maybe magnetic, uh, properties.
- SPSpeaker
We should do like a show of hands of how many people understood that statement.
- DCDorje Chubak
Yeah. As I was saying that I was looking at, uh-
- SPSpeaker
At least one.
- DCDorje Chubak
Yeah. And then you could, yeah. Um, so, you know-
- SPSpeaker
As a quick show of hands, how many people are physics majors? Okay, so at least five people understood that.
- DCDorje Chubak
Yeah.
- SPSpeaker
Yeah. Okay.
- DCDorje Chubak
Like, like taking a step back, right? Like, for example, we're not doing polymers in the lab right now, so we don't expect that what we're doing necessarily extends to polymers at atomistic scale. But superconductors, semiconductors is like a very deep part of human technology, and there's a sense in which Moore's law is the most impressive thing humanity has done. Um, so we really wanna keep pushing on that and, um, we've learned so much. For example, we're pushing on superconductivity but we learned so much about synthesizability of materials, how to synthesize them better because again, like they are governed by similar things like thermodynamics, atomistic interactions.
- 9:54 – 14:14
Superconductivity 101 by analogy: resistance, heat loss, and why discovery matters
- SPSpeaker
I, I think given that feedback we're gonna have to spend like two minutes just giving people context on what is a superconductor-
- DCDorje Chubak
Right
- SPSpeaker
... why is it useful and what it's, what is its link to semiconductors. Can we just for, just for, uh, exhaustiveness start there?
- DCDorje Chubak
So as we know, physics is endless, it's boundless and there are different energy scales. So you know you can go really deep and start studying string theory or, uh, quarks. Um, but those studies don't affect our day-to-day life very much. The energy scale that affects our life is usually, you know, a few eVs and this governs chemistry, biology, but also batteries, semiconductors. In that energy scale, the thing that really matters is, uh, quantum mechanics. So it's not even necessarily quantum field theory but you might benefit from that, but it's just quantum mechanics. And in that regime you just have, uh, atoms and electrons. It's ions and electrons and they're interacting with each other. So this is like if you've taken a introductory solid state physics class, this is what you would learn. Uh, that's the level at which we're studying. We're not doing any quarks, we're not doing any strings.
- SPSpeaker
Well, for a non-physicist, let's, even though it, this, actually, you know what, this is systems class. We have mastery over different, um, frameworks. For a second, let's reason by analogy.
- DCDorje Chubak
Yeah.
- SPSpeaker
How about we say a, a superconductor is a pipe through which things need to flow. Can you build on that analogy for a sec?
- DCDorje Chubak
Yeah. So, um, one thing that we've learned about materials in the last century or two is that, uh, when you pass current through a material you have resistance. Um, depending on what the material is, that resistance might be driven by electrons colliding with ions and every time they're scattering you're basically losing, uh, some, uh, to energy to heat.
- SPSpeaker
So on, on average today in materials like on a, in semiconduct- on a chip, what percentage of energy is lost?
- DCDorje Chubak
Uh, yeah, so like for a copper wire it will all be basically heat to resistance, right? Um, so like a major amount. So if you're putting a lot of energy into these chips in data centers you're also suffering from, uh, heat loss. You also have to then dissipate that heat. Um-
- SPSpeaker
So would it be fair to say roughly on average at least 50% of all energy that goes in a data center is wasted? Because of materials bottlenecks
- DCDorje Chubak
So, I mean, we've been told that Moore's law might be, like, a combination of just materials improvements, uh-
- SPSpeaker
Right
- DCDorje Chubak
... stitched on top of each other. Um, what's, what was amazing is back in 1910, around, um, there was the first time we could, uh, liquefy helium, and we realized that there are certain materials that don't have any resistivity.
- SPSpeaker
Hmm.
- DCDorje Chubak
So, uh, current can just flow through it without any, uh, dissipation, and that's an incredible observation, one of the most amazing things we learned about, um, physics. But we still haven't really, like, taken that technology and applied it to our life enough. Uh, we're still using very old superconductors and MR machines. We're using liquid helium. And if you think about the exciting future we can imagine, like fusion energy, uh, quantum computers, uh, maglevs, uh, lossless transmission, these all require superconductors to some extent.
- SPSpeaker
Well, what I've learned w- uh, interacting with Dorje is if I say something, like at least 50%, he doesn't answer, that means that's a no. But, but, but, but the point is not-
- DCDorje Chubak
I don't wanna give you-
- SPSpeaker
Let's say-
- DCDorje Chubak
... the correct numbers.
- SPSpeaker
Exactly. Let's say-
- DCDorje Chubak
DX knows. [laughs]
- SPSpeaker
Double-digit percentages-
- DCDorje Chubak
Hmm
- SPSpeaker
... are lost in a, in a chip because of energy transmission loss. Is, would that be accurate?
- DCDorje Chubak
Yeah, seems reasonable.
- SPSpeaker
Okay, let's start there. Um, and then the idea is by using an AI system to discover new materials that reduces that loss, we can unlock extraordinary gains in efficiency, in energy transmission. Roughly, that would be the eval-
- DCDorje Chubak
Yeah
- SPSpeaker
... right?
- DCDorje Chubak
That'd be one. You can also do communication with these, um, and then you can do exciting physics like quantum computing, uh, crazy magnetic fields for fusion.
- 14:14 – 15:01
Meet ‘Onnes’: naming the agent and the industrial-scale science philosophy
- SPSpeaker
Are, are, are we ready to talk about the name of the agent that you guys have been working on?
- SPSpeaker
Yeah.
- SPSpeaker
What, no, you should go ahead and talk about why you named it that. I think it's a interesting metaphor for what, what's going on.
- DCDorje Chubak
One, one of our incredible researchers, Mansingh, has named it Onnes, and there are two reasons. Onnes was the one who, uh, whose lab liquefied helium and ex- discovered superconductivity, but also because Onnes was one of the first people who said scientific research should be done at industrial scale. So he was kind of strange for his time, like back in 1908, where he would hire professional engineers, professional technicians and run kind of like an industrial scale lab to do scientific research, and that really, you know, is inspiring to us because we feel like science is very important, and it should be done very seriously at industrial scale with a big sense of urgency.
- SPSpeaker
O-Onnes as in O-N-N-E-S.
- DCDorje Chubak
Yeah.
- 15:01 – 16:50
Onnes vs ChatGPT-style systems: grounding intelligence in experimental reality
- SPSpeaker
Which a few people have told me they thought meant honest, but, uh, like sounded like on-honest, but, um, how should-- Like could you contrast, Liam, the Onnes system relative to what was the system, the system you were working on at, uh, OpenAI when you co-created ChatGPT?
- SPSpeaker
A much deeper understanding of just the physical world around us. Um-
- SPSpeaker
A-a-as measured by?
- SPSpeaker
Eval's capability to, uh, engineer new systems, predict, uh, that it's gonna be stable, um, be able to predict what is it that you made. Um, so like maybe an example is in order to discover a new material, you need to be able to synthesize it, you need to be able to make it, and the prediction of that can be really challenging, so that's a, a huge focus for ourselves. Um, and I think there's just like this basic premise that machine learning models are good on the data you train them on, on the tasks you train them to do. Uh, but absent this data, there's some generalization, but it's not infinite. Uh, we don't think that you can just start thinking your way to a new room temp superconductor by reading a textbook, shutting it, and thinking super hard. Uh, you have to make contact with reality. You have to carry out experiments. Um, and so basically our systems have, through this new data, through our computational predictions, through the data produced from our labs, just have a much deeper understanding of how to do these types of things. Um, so we're not really spending much time thinking about how do we improve coding, like that's continuing to improve, uh, through the existing efforts, but really focusing our efforts towards, you know, the construction of the physical world.
- SPSpeaker
Okay. Shall we transition to questions?
- SPSpeaker
Sure. That sounds great.
- 16:50 – 19:53
Modeling and optimization for science: why active learning beats Bayesian optimization (in practice)
- SPSpeaker
The question is: How do you guys think about combining different types of models to push the frontier of scientific discovery?
- SPSpeaker
Or like optimization approaches?
- SPSpeaker
Yeah.
- SPSpeaker
Yeah.
- DCDorje Chubak
Yeah. So great question.
- SPSpeaker
Thank you for keeping me honest.
- SPSpeaker
[laughs]
- DCDorje Chubak
So there's a trend where a lot of physicists move into other fields. One of the fields they've been moving into a lot is machine learning, and, uh, the, a common thing is the physicists will first fall in love with Bayesian optimization because there's more theory in it, and there's a sense of, um, uncertainty. Um, but, like one thing to keep in mind is if your machine learning model has generalization problems, which every model will have, right? Every model will only be trained on a subset of the universe. Um, its uncertainty about uncertainty will be even bigger. Like generalization issues for point prediction is smaller than generalization issues for uncertainty. So what we see practically, I think, is Bayesian optimization is probably not that useful. Uh, on the other hand, active learning is the opposite, I think. So if you're working on an academic dataset, like you take one academic dataset, you do uniform split, train and test, uh, active learning won't help. So you know, when we were doing deep learning research back at Google, there were a lot of people trying active learning to improve ImageNet accuracy, and it would never work. Um, so then that gives people the false impression that active learning doesn't work. But in reality, I think it's the opposite. In real life, like something like imagine you're training Waymo- You have to do active learning because Waymo is good at some things to a point and then not other things, and you have to now kind of push into the other things. So this is very relevant to us because when we do science, our models understand some of the science really well and some of the science not at all, and active learning means we can push into the part we don't know yet slowly and then expand generalization bit by bit. So we do active learning all the time. Like, every day we're running new experiments, and usually those samples are, could be considered active learning because they're, uh, predicted by our existing model for the future. Um, yeah, that's kind of how I view maybe the different approaches.
- SPSpeaker
Hey, you wanna add?
- SPSpeaker
Yeah, I think, like, on the LLM side, we use standard optimization techniques. Um, there's a lot of carryover in that, and I think we're finding those to be effective. But I think big plus one to this notion of, like, active learning, and effectively what you're doing every day is pushing the frontier of, like, what we understand about these systems. And it kind of goes just back to my comment earlier, where it's insufficient just to read a textbook, close it, and think really hard. You have to carry these things out. And as part of doing our discovery process, we intend to produce things, and we routinely make the thing that we intended to make. But sometimes we just find anomalies, and we have to track those down, um, and so that's a, a really fun part of the scientific process, and we're like, "Hey, this peak is just not accounted for," um, and that material doesn't exist, so it's like, of course you... And then that provides data for the next version of the system, so it's very core.
- 19:53 – 21:44
AI scientist vs AI agent, and what counts as ‘new material discovery’
- SPSpeaker
A question is, "What's the difference between an AI scientist and an AI agent?"
- SPSpeaker
So it's a model that is like a large language model. It's, uh, thinking. It's orchestrating tool calls. The tools might be, um, standard tools, or the tools might be other neural nets that we train, so the, a neural net's invoking other neural nets. Uh, but you can just kind of map it to the, the same thing.
- SPSpeaker
Another question is, "What is the definition of new material discovery at Periodic?"
- DCDorje Chubak
That's a great question. Uh, so there are different dimensions to it, right? One dimension could be the structure, so, uh, we, like, if you look at ICST today, there are about 250,000 crystals, inorganic, um, and they have a certain subset of prototypes. And, you know, very rarely scientists find new structures in the sense that it's a crystal structure that has never been seen before. Uh, so that could be a discovery. Another one could be a material with properties that didn't exist before. So for example, today the ambient pressure, superconductivity, state-of-the-art is something like 133 Kelvin. Um, if you find something at 160 Kelvin, that could also be, um, discovery. We also really care about discovery of synthesis recipes, so even if a material has been known before, if we can make it more efficiently or make it with slightly optimized properties, we're really also happy. Um, it's just about, you know, pushing the boundary of what humans can do for organizing the atoms around us. Like, you know, like, since, since you're students here, maybe I can, uh, chat with you about this a bit. Like, material science is often not very exciting, but it's very hard to understand because it's basically physics of atoms that we interact with. Um, so I think it's very exciting if you wanna, if you wanna try it. Um, and materials discovery basically in my mind is our control over how we set up the atoms around us, um, which is obviously very exciting, right?
- 21:44 – 25:54
Why conviction formed quickly: unmet hype, poor benchmarks, and massive materials upside
- SPSpeaker
Your question was what led us to have conviction and-
- SPSpeaker
Yeah, like, what, yeah, why do you have conviction in Periodic?
- SPSpeaker
Yeah.
- SPSpeaker
I, I think it's-
- SPSpeaker
How about, how about that? Can we abstract to that? Is that the question?
- SPSpeaker
Yeah.
- SPSpeaker
These guys, [laughs] I mean, you guys can tell, like, look, I, I, I think you've had a chance to see over the course of the class, right, um, just a really wide area of leaders, right? Whether it's people at the, whereas Jensen at the chip level or, um, Scott Nolan at the energy stack, part of the stack, and so on and so forth. But, um, I just think, you know, for a long time people have talked about AI actually being useful in, in the physical world, and, uh, I, I met with a lot of teams over the last few years, especially when I was at a16z, who were making attempts at that, um, but the, just, the progress just wasn't showing up as much as the marketing was making it seem like. And so I, I was pretty frustrated with that, and then, uh, I had a chance to, actually, after last year's class, um, started spending some time at the applied physics group on campus, um, the connect, uh, working on evaluating frontier models on their ability to reason about condensed matter of physics data, and, you know, thank you ZX for the opportunity for that. Um, and we came up with, uh, a draft p- paper, um, for submission to NeurIPS that kind of b- like benchmarked a bunch of last year's generation frontier models, and they were just terrible. Um, I think then Jason Kwon at OpenAI, who I sent the paper to, Jason is the chief strategy officer at OpenAI. I said, "Hey, would, you know, these results are kind of terrible. Um, uh, I just wanna make sure we're not making any mistakes in the methodology. Like, who would be a good, who, who could give us peer review on this paper?" 'Cause that's the scientific tradition, right? You, like, measure stuff. If it looks off, get some peer review. Then we got introduced that way, and I think separately, 'cause, 'cause Jason was like, "You should definitely talk to Liam and Dorje. They've been th- thinking about the same thing." And I think you had, you had just left OpenAI. Dorje, you were wind, like, rolling off DeepMind, and there was just a meeting of the minds, and it just became clear to me that, um, it, it wasn't, it, I don't think it was very, it was one of those things where the opportunity was quite clear. The need for the world to have, make more progress on this domain was, like, the, uh, the, the, the value creation is extraordinary, right? Like, we, humanity is generally The history of humanity can be separated into material eras, right? Stone Age, Bronze Age, Copper Age, Silicon Age. Like, how do we get to flying cars? I mean, really, as a civilization, one of the bottlenecks is materials. Um, and I think, uh, we had kinda converged on a similar theory of that opportunity, and everybody knew it, it was super valuable, but somebody had to just go do it, and the guys knew how to do it, given their experience at DeepMind and OpenAI. And so I don't know, I think, I think because, like, from, from me- When we first met to, like, term sheet and starting work together was four days, five days.
- SPSpeaker
It was fast, yeah.
- SPSpeaker
Uh.
- SPSpeaker
Yeah.
- SPSpeaker
I don't know. What, how, well, uh, yeah, that's, that, that was my answer, but you guys should, you know, talk about the... Feel free to answer.
- SPSpeaker
Yeah, I mean, well, I mean, I can't answer for why you're excited about it, but, I mean, I think one reason why I'm excited, and I think, you know, both of us, is just the ability to engineer the world is just so profound. Um, so much of the world has kind of been accidentally discovered or just, you know, trial and error, and we see just how quickly the digital world is changing. But if we can make these things smarter about our reality, um, we think that's kind of, like, what pulls forward this, like, sci-fi future. And yeah, I think we're, we're very optimistic on the course of the technology, but also kind of cognizant that you need to make this, this leap into the physical world. It's not gonna come just from, like, you know, again, reading a textbook. You need to make that connection. Um, and so I think the implications of this system is just really profound across so many things. That's, that's what gets us excited.
- 25:54 – 27:23
Student anxiety about AGI: what remains undone and how to pick impactful projects
- DCDorje Chubak
Actually, can I ask you a question?
- SPSpeaker
Yes.
- DCDorje Chubak
Are the students feeling anxiety over, like, their education and career given what's happening with LLMs and AGI?
- SPSpeaker
Yeah.
- SPSpeaker
Raise hands.
- DCDorje Chubak
Okay.
- SPSpeaker
Wow.
- DCDorje Chubak
The reason I ask this, um, I recently gave a physics colloquium next door, and I feel like the biggest concern I heard from physicists were, like, "What are we doing here if AGI is gonna do all physics?" Like, in my mind, this is kind of confusing to me because the LLMs are out there, but there are only a few things they really revolutionized. Um, like in my field, I would say the thing they revolutionized is force fields. You know, we approximate quantum mechanics with machine learning these days, and graph neural networks have done a great job. Way-Waymo is incredible. You know, Tesla, they can do self-driving. Uh, translation, coding seems really good. But there's so much more to improve and, like, you know, i-if you're interested in something, just use LLMs as an excuse and go improve it. Um-
- SPSpeaker
So if you were in the class today and you were working on your senior project, what would you work on? Or your final project, sorry.
- DCDorje Chubak
Ah. So when I was finishing my undergrad, I worked on using machine learning to design analog circuits, so I would do that. Um, yeah, whatever, you know, the student's really passionate about, interested in, it's probably not revolutionized by LLMs yet. Somebody has to do it. The LLMs won't do it by themselves. So, um, I think it's a very exciting time. You know, like, golden eras in human history is usually short, but it's really good to be in the golden era because that's when most impact happens, so this is, this is, you know, your, your chance to be part
- 27:23 – 29:03
Simulation vs reality: DFT strengths/limits, catalysis difficulty, and ML force fields
- DCDorje Chubak
of it. How do you go about solving problems and predicting atomistic properties? Do you use, uh, tools like density functional theory? And how do you, for example, model something like catalysis? Um, this is a technical question, so I can answer it-
- SPSpeaker
Please go ahead
- DCDorje Chubak
... briefly.
- SPSpeaker
Yeah.
- DCDorje Chubak
Um, so you know, I should say I feel like, um, simulation tools are good for some things and not good for other things. Um, if you look at the theory of density functional theory, it's clearly good for ground state properties. It's literally the Kohn-Hohenberg theorem. So we find it to be really good at predicting, um, formation enthalpy of ground state, but we don't think it's good for predicting band gaps or excited states. Um, catalysis is really hard because catalysis requires you to know the atomistic structure, and we don't actually know what's going on. It's very messy. There are steps, like atoms form steps. There are defects. Sometimes individual defect makes a difference that you can't model, so we haven't been modeling catalysis as much. Maybe that's a more empirical direction. Um, we have been benefiting... So as I mentioned earlier, machine learning really revolutionized force fields. So I don't know if you know this, but I think one of Einstein's first papers was, could be considered a force field paper. He was, I think, studying empirical methods to approximate the interaction between, like, a surface and a, um, molecule kind of thing. Um, and force fields have developed, developed tremendously, lots of big improvements from Stanford. Uh, but when machine learning came about, it really changed how we do force fields. And in our company, we have really good machine learning expertise, not just in LLMs but also in graph neural networks, so we have come up with new architectures, new capabilities. But it's still, I mean, it still cannot do everything, uh, and can't do catalysis, so yeah.
- 29:03 – 31:35
The chicken-and-egg of discovery: no ‘complete’ dataset, so iterate with sample-efficient RL
- SPSpeaker
You wanna start?
- DCDorje Chubak
Sure. I mean, yes, there's definitely a chicken and egg problem, but taking a step back, that's not just a problem for our company or for machine learning, right? The chicken and egg problem of scientific discovery is, applies to everyone. I mean, you might even think if there's an alien civilization somewhere doing science, they'll also have the same chicken and egg problem. The chicken and egg problem is the science we know we understand well. The science we want to discover we don't yet understand well, so you always have to iterate towards it. That's why the earlier question about active learning was a good one because the only way you can iterate towards it is basically do something like active learning. Um, there are some datasets that's like ImageNet that's helpful to us, for example, force field datasets. So you might have seen, um, Meta open source this OOMA dataset that's like 100 million, uh, density functional theory calculations. So it's a bit like ImageNet where you can train on it, and the model you train is useful for other tasks. Um, but in general, I feel like there'll never be this comprehensive dataset for science because if there was, that would stop being science. That would be like textbook. That would be education, and then you'd immediately train on all that data and try to discover new science. And this is partly why Liam and I are so excited about this company because there's no end to it. You know, if you automate accountants, at some point, there's only so much accounting to do. But if you can automate science, there's no end to it. You can keep discovering more and more interesting science. You can keep developing better and better technology. So it's nice to be in a field where there's no end.
- SPSpeaker
Yeah, and then maybe some more comments too. Um, I think the sample efficiency of these algorithms is, is quite high. Uh, so we're able to really quickly get going in some chemical spaces or some search spaces with a relatively limited amount of data. And from an ML perspective, one of the core things we think a lot about at Periodic is how do we push the frontier of sample efficiency, especially in, like, reinforcement learning? Um, so again, kind of going back to how a lot of frontier labs, uh, create models, you typically have, uh, the ability to do, uh, for your policy many, many rollouts. So you're asking a math question, you could do a thousand rollouts, a million rollouts. You can kinda scale that really quickly. Whereas in the physical world, it's not plausible to just expand that, um, arbitrarily high. Uh, we're scaling it up a lot at Periodic, but we can't just, like, increase it by a factor of ten or a hundred on a whim. Uh, so anyways, we're, we're focused a lot on sample efficiency here and how to make better use of the data, and a big part of that is the com- model-based reinforcement learning. And so I think we've been happy with the sample efficiency so far.
- 31:35 – 42:04
Open research problems and practical barriers: synthesis, characterization automation, and partial observability
- DCDorje Chubak
The question was, what are some open problems in AI for science that, uh, people in academic labs should focus on? And the other question was, going back to your 2022 paper, uh, you had this active learning pipeline for discovering zero-Kelvin stable materials. Are you still using these algorithms? Good. Um, so I mean, one thing we really care about is being able to intentionally synthesize materials. You know, like, so if you, if you go into a lab, I think we're really good at making materials if they've been made before. We're also pretty good at making materials if something similar has been made before. But if I bring you a completely new material, it's actually pretty hard to figure out how to make it, and it's a lot of trial and error. It's a lot of, like, just brute force trying things. Uh, if, if I were in a lab today, I would really try to figure out how to make the synthesis approach more intentional. And of course, this is very different for different fields. We're focusing on one kind of inorganic crystal space, but, um, like, I'm sure this is also a case in different fields. I heard in biology, some things are easy to synthesize, some things aren't. Uh, I'm sure that's so. But the reason I think synthesis kind of sounds boring, but it's actually extremely important, is that's how we build the things around us. Like, once we can get good at making things intentionally, now we can construct an amazing technology. Um, in terms of whether we still use those tools, I mean, we still use active learning. We still definitely use density functional theory. Um, one point of that paper was if you can make your zero-Kelvin convex hull bigger, it will make, it will give better synthesizability predictions, and that's definitely correct. So you can actually observe that the more densely sampled your convex hull is... It's, sorry, very technical again. The, the better your predictions will be. So we actually use the biggest possible convex hull, uh, including our work, other work that's publicly available.
- SPSpeaker
Yeah, and I'd say maybe answering a, a different, uh, abstraction, uh, I think making sense of noisy data, sort of, sort of like decision-making under uncertainty, uh, consistency of results. So, um, you know, the internet's, uh, full of results, some correct, some not. Uh, and so I think, like, a system that can be more effective, uh, at kind of parsing these things and saying at le- at least, like, flagging, like, okay, these pieces of evidence are consistent. This is inconsistent. Um, I think those are useful technologies. Sample efficiency, again, um, another really key aspect. So I think these are, like, some things that would apply for many different scientific domains.
- DCDorje Chubak
Yeah, something else that, like, Liam's answer reminded me we really care about and I think others can also improve on, is automating characterization analysis. So, you know, in every scientific lab, there's a lot of different characterization instruments, and they usually, uh, give you the measurements in a way that's not immediately interpretable by a human, and you have to spend a lot of time analyzing those results. But we've seen that you can actually make tremendous progress if you give an LLM the kind of tools that human scientists use and then automatically analyze those results.
- SPSpeaker
Yeah. I mean, maybe even, like, a frontier problem of how do you come up with a good hypothesis? Um, so it's, it's quite easy to grade whether or not you got that math answer correct or whether your code compiles. Uh, but correct hypoth- or, like, a, you know, interesting hypothesis, um, I think is really challenging and, um, that's, I think, really a frontier for this field. Um, another aspect too is we often get the sense that the knowledge is in the model. It's like it's in the weight somewhere, but it doesn't really, um, surface it at the right times. Um, so we would hope for some intelligent reasoning, some use of that knowledge on a scientific problem, and it won't do it necessarily. Uh, but if you ask it directly about all those different skills or, um, pieces of knowledge, it does have it. You can reliably surface it. And so these models aren't really combining the information in, uh, as intelligent a way as a human had all of that in their brain. So I think there's a lot of deep work to be done there. You can basically timestamp the universe. So you're like, okay, like, here's the, the set of, um, known, uh, like this is the known data we have experimentally, computationally at some timestamp T, and pr- uh, produce a world modeling task where you're trying to then predict outcomes of future things. Uh, so again, AI scientists, AI agent, all the same here. That LLM, that agent can then use some of these computational tools in the service of predicting these things. Um, so that's one example where the agent is invoking these things in order to predict, uh, the future things. So, like, the world model Is not just in, uh, weights, it's not just, like, inference time, but it's using these tools in order to do that. The question is, uh, you guys are tackling a lot of difficult problems. How do you stay motivated when things aren't working? Um, I think it's, I mean, I think there's, like, a huge amount of internal motivation, so if we had tied all of our motivation to some, like, extrinsic goal of, like, you know, the room temp superconductor, then we would've, like, lost, like, motivation because we haven't discovered the room temp superconductor yet. Um, we don't know if it's physically possible. We, we hope to be, but, um, if it, if it's physically possible, we're gonna aim for it. Um, but I think there's very clear progression that you can track. So, um, you can say, "Are we making progress, uh, technically on the different pieces we wanna see, and is that consistently getting better?" And, you know, like, again, this whole end-to-end discovery loop is incredibly difficult. We haven't seen, uh, any very compelling, uh, instance of it yet. Um, we hope to soon. But there's very clear attribution to all the little pieces, and are we making progress on those little things? And that gets me, Dorje, and the team very excited.
- DCDorje Chubak
And also, you know, like, I feel like we have this responsibility to take this LLM technology, the AI technology, and lead it to something positive. And, um, like, uh, I think part of what motivates us is this is one path of making these technologies more positive and useful for humanity. There are many other paths, but, uh, yeah, there's a lot to do, and it's very motivating to try to work on, like, a very revolutionary technology and try to steer it towards something positive. I can't think of, like, something more motivating than that, so.
- SPSpeaker
The question is around, uh, barriers to, um, doing work in the physical world. Um, and are those barriers expected to decrease over time? And yeah, so I think the obvious thing is, like, you have to build the thing. That's difficult. Um, I would say that our expectation would be the barriers should lower, um, as AI systems and as robotic systems get better. We hope that the creation of new autonomous labs or building new infrastructure gets easier and easier, so hopefully the labs we're putting together now are among some of the hardest today. Uh, yeah, so that's how I'd answer it. I think also you can pick your domain wisely, so if you're saying like, "Okay, actually, I'm not gonna deal with material science. I wanna go one energy scale deeper that, uh, Dorje was talking about. I'm gonna revolutionize high-energy physics." Um, in that case, you would be, uh, beholden to start constructing particle accelerators, and that's a much more difficult, uh, experimental loop. Uh, these are tens of billions of dollars, takes thousands of people. So whereas there's, like, other areas in the physical domain that, uh, could be much faster to iterate on, lower capital requirements, as a, kind of, like, a really good test bed for kind of making the connection of AI into the physical world. Uh, yeah.
- DCDorje Chubak
Also, I feel like the, the, the bar isn't too high. You know, so what is the state-of-the-art in frontier labs right now, right? Like math, combinatorics. I think maybe they'll go into, like, very theoretical physics. So even if you do slightly more physical things like, I don't know, grab a camera and, like, play with its output or, like, point a telescope in this, like, there's, the, the bar is so low in terms of how physical and relevant you have to make AI to make progress, you probably can just be slightly above it. And then there's, like, what we're doing, which is, like, a big investment into material science, and then there is what Liam is saying with, like, particle accelerators or something. But yeah, I think the, the bar is pretty low right now. But it may not stay like that. In a, in a few years, I think a lot of young, excited people will realize LLMs can do way more than coding, and then, uh, it'll be very fun, yeah.
- SPSpeaker
Question is around other kind of areas of where there's a high degree of verifiability that might be really exciting for these types of systems as well as, uh, areas that Periodic might be interested in in the future. Um, I think predicting asset prices is interesting, so [laughs] finance, um, I think that's, you know, highly verifiable. Uh, can you predict the future state of, you know, this stock or this bond, uh, this yield? Uh, so I think there's interesting things to be done there. Um, from a technical perspective, uh, there's, like, latency considerations, um, and that is something that Periodic is not interested in doing. [laughs]
- DCDorje Chubak
The asset prices or the latency?
- SPSpeaker
No, the latency, absolutely.
- DCDorje Chubak
Yeah, okay, good.
- SPSpeaker
Uh, fi- yeah, uh, pivot into finance, I don't think so. [laughs]
- DCDorje Chubak
And one, one more thing is, um, you know, currently these verifiable rewards aren't just verifiable, but they also have full context present to the model, like if you're doing coding, right, the whole code base, the compiler. Um, but in real life, most things aren't like that. Like, when we do experiments, one thing we notice is not all the context can be recorded. Um, this is, my suspicion is this is correct and true for almost everything in the world, where you can never, uh, record all the aspects of what happened. And then the question is, how can you improve these models in that uncertainty? Um, that's very exciting, and it doesn't just apply to physics, right? It applies to everything, including asset prices, yeah. Thanks, guys. [audience applauding] See you.
Episode duration: 42:08
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Transcript of episode 8cAQdELWYuo
