What Makes Something Alive? Assembly Theory and the Origins of Life (Sara Walker, Theoretical Physicist)

It's almost a different order of magnitude level of ambition, or perhaps several, to say, I'm gonna try and rethink the very laws of physics and spend my life figuring out what the origins of life are.

I think that we need new descriptions of how the world works to really understand what's happening around us now. And for me, that is associated with the physics of life and mind and intelligence and these sort of missing areas where we just don't have any fundamental description to help anchor our understanding of what's happening around us.

An LLM to me is something that in some ways is a compression of all of human history And also every prompt you do is doing this massive scouring of all information across time. It makes me wonder how to think about that.

We think when we build computational algorithms, they can know everything and they can encode everything. I don't think computation can capture reality in totality. We don't live in a computational universe, and I think we're confusing it right now because the technologies we build are so powerful.

I'm Mario, and this is The Generalist Podcast. As the saying goes, the future is already here. It's just not evenly distributed. Each week I sit down with the founders, investors, and thinkers— we're living in the future— to help you see what's coming next and understand it more clearly. Today I'm speaking with Sarah Walker, a theoretical physicist who studies the origins of life. Sarah is one of the most provocative and original thinkers in science today. And the author of the excellent book Life as No One Knows It. As AI prompts us to rethink what consciousness, intelligence, and life really mean, Sarah's work is becoming increasingly important and profound. In our conversation, we explore why Sarah believes we need new laws of physics to understand life, how her assembly theory suggests that complex objects like DNA molecules and microphones Our evidence of life, why we perceive objects smaller in time as physical while those larger in time appear abstract, and how this framework could transform our understanding of life across the cosmos, from the methane lakes of Titan to the atmospheric signatures above urban China. I left this conversation full of fresh ideas and new questions about the fundamental nature of reality and what it truly means to be alive in our strange universe. If you enjoyed today's episode, I hope you'll consider subscribing and joining us for the incredible conversations we have coming up. Now, here's my conversation with Sarah Walker. This episode is brought to you by Tezi, the AI agent for recruiting high-quality candidates quickly. When you're ready to hire, it's easy to lose weeks trying to find the right recruiting partner, building their context, and filling the pipeline. With Tezi, you can be up and running in minutes, not months. Tezi offers agentic AI recruiters that work 24/7 to fill your pipeline with high-quality candidates, handling all the recruiting busywork, from instantly reviewing applications to sourcing relevant talent, rediscovering hidden gems in your ATS, and even detecting fake profiles. Here's what that means for you: Tezi prequalifies candidates so your hiring team only spends time on the best ones. Whether you're short-staffed or just want to move faster, Tezi gives you the extra capacity to build a strong candidate pipeline fast. Innovative companies like Mutiny, Aurum, Nominal, Forage, and Real are already using Tezi. Head to tezi.ai/mario for 10% off your first year and see how quickly Tezi's AI agents can supercharge your recruiting. That's T-E-Z-I dot A-I slash Mario to get started.

Well, uh, Sarah, I'm so excited to have you here.

I, uh, reached out after reading your book, uh, Life as No One Knows It, which was one of the most thought-provoking books I've read probably in the past 12 months or so. Uh, and so, yeah, super excited to talk with you about your work, uh, the origins of life, uh, all of the, the, you know, the big questions.

Awesome. I'm very excited to be here.

Maybe to, to start off with, uh, you could share a little bit about how you would qualify what it is that you do and what it means to be sort of someone who spends their life thinking about the origins of life.

Oh yeah, that's actually a hard question. I'm like, you know, it depends on what level my existential crisis is, how I qualify what it is that I do. But I am a scientist. I work on origins of life. My training is in theoretical physics, which is, you know, a particular branch of science that we really focus on deep fundamental questions and trying to come up with the most abstract descriptions of reality. And throughout my career, I had always thought, like since very early in my career when I started working on origins of life, that that problem is so hard, you know, thinking about how it is that living systems arise when there is no prior life, that it might demand entirely new laws of physics. And so these would be principles that are universal in the same sense that we think about gravitation or as deep in the structure of reality as we think about quantum mechanics. And so that's fundamentally what I'm interested in and work on. And I guess that's probably why I'm like prefacing it with saying I'm having a continual existential crisis because I'm thinking about these very deep questions about the nature of life all the time. But I enjoy it. I like swimming in the sea of uncertainty.

Yeah. You know, it's so fascinating that it's hard to even sort of wrap your mind around it. My day-to-day is spending a lot of time with tech entrepreneurs and venture capitalists. And I think often I am so amazed by the ambition of these people that I meet to build something totally new. It could be something in AI, but it's almost a different order of magnitude level of ambition, or perhaps several, to say, I'm going to try and rethink the very laws of physics and spend my life figuring out what the origins of life are. So I'm sort of curious how that even occurred to you as something that— that would be tractable in your lifetime.

I think it's interesting because obviously it comes across as very ambitious, and I totally agree, it's a really ambitious aim for one's life, but it's actually coming from a sense of deep curiosity, right? I don't think that you can work on these kind of problems unless you, for some reason, happen to have a deep, very deep curiosity and a very deep desire to understand. Something new. So for me, the motivation is really, I think this problem is rather significant, particularly in this day and age where we're really facing a lot of questions about the nature of intelligence and the nature of life as we're building machines that seem more and more lifelike. But also as I think about, we fundamentally don't know what we are yet. We're doing all these amazing things even in the tech sector, right? Like there's all this creative stuff happening. So I am just really excited about trying to understand something in advance, you know, like our fundamental understanding of the world. Like, I was just very romanticized as a student with theoretical physics, which I guess is a fairly unusual statement. I mean, I think most teenage girls are not like, oh my God, I love theoretical physics, and I want to be like, you know, Dirac and Einstein and Newton. But I think for me, I just thought this idea that the human mind could come up with descriptions of the world that were so deep and so universal, and then test them, and then really come to understand the world in a deeper way. And there's also these long-term historical consequences of it. Like, if you think about like, why do we have satellites orbiting Earth today? You have to go all the way back, you know, through history to the early stages of the development of physics. And so— I'm just really interested in the fundamental role that theoretical physics in particular plays in our descriptions of reality at a base level, that like, it's kind of how society decides to talk about what it thinks the world is. And for me, that's really exciting, but it's also an exciting time because we are living in a time of radical transformative change. And I think that we need new descriptions of how the world works to really understand what's happening around us now. And for me, that is associated with the physics of life and mind and intelligence and, and these sort of missing areas where we just don't have any kind of fundamental description to help anchor our understanding of what's happening around us.

Yeah, I mean, I, I couldn't agree more with that last part. It feels like such a tricky time to understand what intelligence means, what life means, and to have new language and sort of open the idea space for how we conceive of these things feels like it could have you know, just such far-reaching implications for our own, yeah, sense of existence. And so exciting to talk about some of these things. To get a sense for your intellectual and academic journey, you know, I think you started maybe in cosmology. How did you sort of end up winding your way towards this specific set of questions?

Yeah, so the sort of, you know, root of it was I didn't know what I wanted to do out of high school. So I went to community college and I just took all the science classes I could, but I absolutely fell in love with physics for the reasons that we were just talking about, about these really deep fundamental descriptions. And so by the time I got to graduate school, I was just very adamant that I was going to be a theoretical physicist and study the most fundamental things. And when I was an undergraduate, you know, usually when you go through an undergrad physics education, you're taught, you know, particle physics is fundamental, quantum mechanics is fundamental, general relativity is fundamental. And so I thought I wanted to do something in cosmology and particle physics because those were really the most abstract, fundamental, beautiful theories we had about how reality works. And so I did go to graduate school, you know, I took all the classes like quantum field theory, general relativity, all this stuff. But it was really funny because my PhD advisor, like the reason I got to work with him was because he was a cosmologist. And, you know, later in his career after working on origins of matter, he decided he wanted to work on origins of life. And he's like, needed a PhD student and I wanted to work in a cosmology group. So I was like, I will take that project. And I was actually really rather resistant to it because I had been socially trained to think that the fundamental problems were the things that physics had already worked out. And so it took me a little while to really realize that there's this problem staring us in the face that really could be the next frontier in physics, but nobody's looking at it that way. And to be honest, I don't think I'm that competitive in the space of being really the best mathematical physicist or the best programmer, or like all these skills you need to do science well. So I thought, well, if I'm working on this problem that no one else is working on, I find it exciting. Maybe I have space to actually do something profound because that was what got me into physics in the first place. So I kind of liked working on a problem that no one thought was a problem because it gave me time and space to think about it without feeling like I was in a majorly competitive space in science. Because there's some areas in science where you like, you really have to publish fast and you have to publish often. to even be anything. But I ended up putting myself in this really odd creative space where I could say radical things, and for some reason I was getting a little traction because people just didn't know how to think about this problem. And they were like, oh, well, it's weird she's over here thinking about this and talking about it. And so I managed to get a postdoc fellowship to keep working on this problem and then got a faculty position. But it was really It was really kind of, I was kind of resistant to the idea at first before I really started to think about it the way that I think about it now, that we really are missing some very fundamental understanding. And then that got me really excited and kind of obsessed with the problem, and that's why I've been working on it.

Wow, fascinating. I love that story because there's so many sort of generalizable pieces of wisdom there. It's like you had this sort of call to adventure that in some respects you weren't ready to hear for some period of time, but then thankfully you followed it. And then also just the, yeah, the opportunity that comes in these sort of strange packages and, you know, that following these sort of fringes of things often seems to lead to something kind of magical. You know, a guest on the podcast previously was Sarah Seager, and she sort of talked about her journey into, you know, astrobiology and stuff like that in a similar fashion of how it was really something very, very unfashionable and It was in digging in that she found her own calling, so to speak.

Yeah, that's awesome. I'm a big fan of Sarah's. I think she's great.

That's awesome.

Yeah, it's interesting though, just when I thought about what I wanted to be as a mental model as an 18-year-old, when I started getting into physics, it was like, I want to be a theoretical physicist. I want to do something that really contributes in a deep way to that field. And there's the markers, you want to do something that you know, really gets recognized by your colleagues or write a book or these kind of things. But if I had like thought about where I am now versus then, I did all the things that I thought I would do, but not working on the kind of problem I thought I would work on. And so that's really interesting to me about like the shape of your mental model and like how you end up filling it if you, if you're willing to just follow like your passion and like what actually is the opportunities that come. You build some of it, but like some of it, it looks radically different.

Yes.

Yeah.

What was the state of origins of life research at the time that you sort of like started digging into this stuff? Like what, what was interesting people then? You know, how has that perhaps changed?

I mean, it's always been a really interesting field, right? Because of the people that get attracted to it. So, you know, some of the things that I thought were a little bit perplexing to me as an early career scientist were a lot of the people that worked in the field came in much later in their career. So there weren't, There actually, that's changed a lot since I was a PhD student. So now there's like a whole early career original life network and like the early career community is really organized around trying to support each other to try to really work in this field. And I think they're doing a phenomenal job. And we did have like early career astrobiology networks, but I think the original life field was a little bit more diffuse. So, so that was one thing like, and I really thought, I'm like, well, no one's going to solve a problem if they start when they're 60. Like you have to start like when you're young and like make it a career-long problem to really make an impact. impact. And you need people that work their entire career on it. And I was actually discouraged by a lot of my senior colleagues, even ones that specialized in this field, you know, to not work on it. I remember a very prominent chemist telling me maybe I should find, you know, something, um, you know, a position doing something else rather than origins of life. And I talked to him a few years later, like after I was— so that was when I was a postdoc. And then years later I talked to him as a professor and he's in it. And I said, well, I did it. And he said, well, you had like, you had the right kind of pathology. Like I was too obsessed with the problem to be turned way, right? So in some sense, it's like in order to survive in that field, because it's not a very prominent thing, it's not a priority for hiring departments and things, you have to have a certain obsession with the problem. So that was one thing is just that there weren't a lot of junior people able to focus on it. It was more a luxury for more senior scientists that wanted to think about something deep. And then the other feature of it was it was very dominated by of 2 fields. One is prebiotic chemistry, which was like the predominant research area in origins of life. And then some stuff in bioinformatics, which is, you know, taking genomes and trying to reconstruct the earliest life on Earth. So this is one of the reasons that we know life on Earth descended from a single ancestor, which is one of the reasons the origin of life is so hard. But on the prebiotic chemistry side, it's like, I remember going to Gordon Conference, which is, you know, one of these kind of prominent conferences. There's an origin of life one pretty regularly. And I would say probably at least 75%, 80% of the people there were doing something called prebiotic chemistry, which is trying to synthesize individual building blocks that we find in living systems under abiotic circumstances. And so, I remember wandering around this conference and thinking to myself, I'm like, where are the people studying the origin of life? Because there's a bunch of bioinformaticians trying to go back to LUCA, which is very complex, and there are a bunch of prebiotic chemists trying to make the simplest things life is made out of, and nobody working in between, and nobody talking about the transition from non-life to life, I think, in the way that I thought that the problem really needed to be worked on. And not really any theorists. There's some modeling being done for specific kinds of chemical systems, but nobody working on the question, what is life? And how are you supposed to answer the origin of life if you don't know what life is? And I just wanted to shake people. So I kind of had this really interesting experience where I was moving into a field where there were a lot of researchers working in that field, but I didn't think any of them were working on the problem that field should be working on. And so they were doing great work. I still, you know, have a lot of respect for these colleagues, but I think the problems are tangential to the one we actually need to solve. And I think part of the reason that I was attracted as somebody that wants to work on deeper foundational principles is exactly for this reason. Like, if the field is looking one way and they're stuck and they're, they don't know what questions to ask, it's a good opportunity for somebody that wants to open new territory or be able to have a creative practice in their science to come in And say, well, these are actually the questions we should be asking. Let's reformulate the questions. Let's think about it a different way. And so, most of my career has been much more in that creative space of just trying to figure out how do we frame the question that's traditionally been a philosophical one? What is life has been something we've been asking for thousands of years, but how do we turn it into a scientific one? And the way that you do that actually ends up making the question look a lot different than you thought it would.

I think that's one of the things that I found so exciting about about your work and the book is that it is such a creative book in many ways. And I think you did an amazing job of making it accessible to a total novice like me, where you sort of advance this hypothesis about the origin of life and you call that assembly theory and talk through how you sort of alighted on it and sort of the implications of it. But it's, yeah, such a provocative framing for this fundamental question that really does feel Yeah, fundamentally very creative. And so maybe we can, can move into that area a little bit, uh, for, for folks, what is the, the, the right descriptor of what assembly theory is? And then I'd love to go into a lot of different threads around it, but maybe we should start there.

Yeah, I'm happy to. Um, you know, I think the theory is actually, uh, very deep. So like there, there's certain ways, like there's different ways I describe it depending on, you know, sort of what level, but like the base level, that's the easiest entry point is to say that, You know, we take complexity for granted, like all of this complex structure in our environment on Earth, as far as we know, like from buildings to DNA molecules to like the microphone I'm talking to you, like these structures don't exist anywhere else in the universe as far as we know right now.

Microphones are not naturally occurring.

Yeah, they're not naturally occurring objects. Um, it's so funny. I use them a lot in conversations because they're always right in front of my face, but you know, there's lots of objects, um, that are not naturally occurring. And in fact, even if alien life exists on another planet, we might not expect microphones to be like, you know, a structure that they, they build, right? So if we just kind of take this idea, um, what we developed in assembly theories is to conjecture that life is the only mechanism the universe has for building structures of a certain complexity. So we do that, we look at this in molecules, but if you imagine there's a possibility space of all the things that could be generated, like you can think of all possible technologies or all possible words that you might utter or all possible Lego structures or all possible molecules. There's a boundary where the number of steps to make something is so high that any random process won't be able to build that object. And it would have to be something that was a product of evolution and selection. And so this is sort of the basic idea of it, is that there is an entire domain of physical structures that you would observe objects only in the presence of life. And that we might actually be able to go in the lab and measure that. And these objects require information, they require evolution and selection in order to form them because they're highly specific objects, like a microphone. A microphone emerges on a planet after billions of years of evolution. DNA emerges on a planet maybe after millions of years of evolution, but they're not things that spontaneously fluctuate out of the geochemistry on a planet.

And so in that respect, a microphone is sort of proof of some sort of life. It's the product of some sort of life because something complex had to create it.

Yeah, exactly. So you might call it like an artifact, but it's evidence of life in the same sense that, You know, like you don't directly observe a gravitational field, but you can observe evidence of it by the attraction of objects, right? So what we think is that selection in some sense is a fundamental force, and it's responsible for creating novelty, and evolution is the process that we see that. But what you want to do if you want to do physics is not just talk about that as something that happens to some physical architectures, but as actually deeply embedded in the structure of physics itself. So obviously we have evolutionary theory, but the current theories really only apply after a cell emerges, which is why the origin of life is hard. And the question is, how do you have a mechanism of getting to a cell from sand, basically? Like, you know, basic minerals and messy chemistry on the prebiotic Earth. And that's a really, really, really hard question. But in assembly theory, we can actually quantify the transition point. And so we've done this With molecules, like basically if you take this idea of minimum number of steps, it's easier for people with Lego. You know, like if you imagine you have, I don't know, Lego castle, sometimes they use Hogwarts Castle as an image and you smash it and you shake it on the table, it's not going to spontaneously reform Hogwarts. Yes. And, the question is like how much instructions do you need to make Hogwarts? And, we formalize that by saying you take 2 pieces, you put them together, you take those pieces, you put them together, and you build up to Hogwarts. You look at the shortest such path, and that's what we call the assembly index or the minimal complexity. Turns out you can measure that for molecules. So you take the atoms, you stick them together, and same process, and then you reuse parts and you actually can have this property of a molecule that we think is a physical attribute about how deep in time, how many causal steps are necessary to make that, how much evolution and selection. And then we can go in the lab and measure it. And so my colleague Lee Cronin, who really developed the foundations of this theory, thinking about how he would measure the origin of life in the lab, his lab actually went in and measured non-living systems, living systems, blinded samples from NASA, including things that were like meteorites, which are supposed to be some of the most complex non-living systems that we know of. And what we found was that living systems are the only ones that produce molecules with more than 15 steps from the experiments. So it's been experimentally validated for chemistry. We think the theory is much more general and we're developing it to be more general. So if you imagine have sort of like a way to detect the presence of life in any substrate, that's sort of where we're going.

Okay. So, so many interesting things there. Um, and, uh, I want to sort of synthesize it to make sure I'm understanding it correctly. You know, to use that Lego example, the idea is sort of like, you know, you take Hogwarts Castle, you put it all in a box, you shake it up, you might get some really interesting little structures. Let's imagine these Legos are a little bit extra adhesive. Those things could never form Hogwarts Castle, but they might form, you know, a little sort of cube-like structure or a little circular little structure, whatever we might say. For it to really form something complex, that's when you start to say, hey, this is sort of passing that threshold. And what you're saying is that the work that Lee Cronin did is that sort of that threshold is at 15 steps, let's say, you know, imagine that, you know, these were, you needed 15 steps to get to Hogwarts Castle. That would be sort of the, the, the threshold at which we see signs of, of life. I can't remember the exact phrasing you used for, for sort of things that cross that threshold, but that's sort of the principle, right?

Oh yeah. So, well, anything above that threshold would at least be an artifact of life. Would at least be an artifact. So it's something that requires evolution to be produced. And that would include structures like human You and me, but it would also include molecules that seem fairly simple compared to like DNA, but are actually very complex and products of evolution that we do only find in life. Like ATP, which is like a major energy carrier in the cell, has an assembly index of 21. And so we don't think that happened until after there was an evolutionary architecture. And I think there's lots of evidence to support that. In prebiotic chemistry, they think they can produce it, but they're like gaming the system by, you know, putting in already designed molecules and already designed conditions. So they put the selection in. And this is also one of the sort of challenges with origins of life is as a research field, experimental research field, is how do you build an experiment where you don't put all of your evolutionary agency and intelligence into the experiment? And you ask, how can the universe generate complex structures without us putting the design in? Like basically the origin of life is the origin of design in the universe, which is also kind of a crazy way of thinking about it. But, you know, that's why there's all these debates between intelligent design and evolutionary theory because this problem's not solved. I obviously think it's a naturalistic explanation, but that's really what we're after is, and it's a really hard problem because we're basically, why it's fundamental physics I think is because basically we're saying the universe has a self-constructing principle. Like I don't think about the universe having anything outside of the universe. The universe literally has to build itself. And part of the mechanism of doing that, like the self-constructing system, it becomes things like a biosphere. That is the only kind of structure that can build into all of these complex potential spaces that we see our planet evolving into.

I think it's also just worth maybe taking a moment for listeners to emphasize how different assembly theory is to the traditional way that people talk about what is life and the origins of life. It's such a radically different way of talking about these things. Maybe for the benefit of folks, yeah, what were the sort of traditional ways people would say, here's how we define life, or here's how we talk about the origin of life, versus saying, We're looking at it in this sort of step function way.

Yeah, so I can definitely do that. And then I'd like to kind of qualify that a little bit historically, because I think, I think it's really important to understand that, like, when we come to understand things, we often change the definitions we have for them. So the way that people think about it traditionally, like the most popular definition is what's the so-called quote unquote NASA definition, although it's not officially endorsed, it came out of a NASA workshop, is life is a self-sustaining chemical system capable of Darwinian evolution. And so just as like general motifs you see in a lot of definitions of life, there's the idea that life is necessarily chemical, and there's the idea that life is necessarily a Darwinian system and also self-sustaining. So those are sort of major themes in most definitions for life. But it's really hard for that to be like sort of a definitive criteria for life because one, you don't know how to measure it. So like, how would you go and test that as a theory on another planet? You know, if you're just You know, getting a molecular sample from a plume of Enceladus, which is an icy moon of Saturn, you know, spewing water into space. You fly through it. How do you measure a Darwinian system capable— I mean, a chemical system capable of Darwinian evolution that's self-sustaining? It's actually, I mean, it's a great philosophical concept, but it's a really hard scientific one. And then there's other issues that evolution only happens on populations. So like, is an individual alive by this definition? If it's not self-sustaining, is it alive? And, you know, no human alive right now is self-sustaining unless you're trained to be one of those gorilla, I don't know, people that go out in the woods. A survivalist or something. Yeah, survivalist. Thank you. Yeah. I'm obsessed with some of these people that can do these things because I think it's pretty profound in this day and age. But most of us are societally dependent, right? We're not actually individuals in some sense because we require societies to survive. And so all these definitions have these sort of gray areas. And Carl Sagan had this great essay that I talk about in the beginning of the book called Defining Life, where he talks about Like, even if you go in a biology textbook, it's like, you know, has metabolism, reproduces, you know, like this sort of list definition that we usually have, eats food. It kind of breaks down. And he points out rather cheekily that like, you know, aliens coming to Earth might mistake cars as the dominant life form. Yes. And I actually don't have a problem with technology being alive. Carl did, you know, like he wanted to exclude them. But I think what you want to do is not approach the problem by assuming a priori you know what life is. And you want to include these cases and exclude these cases, which is traditionally what we've done because we just think there's a hard boundary between life and not life. But the approach of theoretical physics and why I think it's so powerful for answering certain kinds of questions is it doesn't start with that kind of premise of categorizing nature. It starts with a premise of fundamentally understanding and actually unifying things that we think are different. So in some sense, you could view assembly theory as a unification between the animate and inanimate. You know, it's actually trying to build a bridge between non-life and life by saying physics needs to cover both. And how do we build a description of reality that covers both? And you end up with a really different conception of life because the way that we think about life is life is these objects that are very deep in time and where time is really about this causation, like how many steps, how many causal steps did you have to do in a lineage of evolving structures, objects making other objects to get to this particular structure? And so we have sort of a very different sense of life as life related fundamentally to the physics of causation and depth in time. And that is manifest in what we observe in the world as complexity. As the brief aside on the history of physics, one of the analogies, sorry, this is, I just think it's really fascinating to me to be a theoretical physicist alive today trying to build new theories and looking at the history and these radical shifts in the way that people thought. Because one of the things with assembly theory that's been quite hard is people are like, this is such a controversial theory, and we got all kinds of people that love what we're doing and people that hate what we're doing, which also is true of any creative act. I just spent the last week with a bunch of artists, and I felt at home in some of the creativity and the controversy in that crowd. But epicycles were this kind of early description of planetary motion where you have circles inside of circles, and we You know, it's a really good description of like how the planets move across our night sky, but it doesn't explain it, and it's not really a universal theory. And what Newton did that I think was so powerful— what I think most people think is so powerful— is he actually went to like first principles, and he talked about motion, like what governs motion, and he came up with the law of force, right? And so like F= ma, but from there he could derive this idea of universal gravitation. which unifies celestial and terrestrial motion. And so then we have a concept of gravity, and gravity our species has experienced for thousands and thousands of years, but we didn't have a language for describing it, and we didn't realize planetary motions were governed by the same principles.

Yes.

And even though we are life and we've been experiencing life and we might have these sort of ways of describing it in terms of evolution or metabolism or things, they're not that kind of deep description. And I think what Assembly Theory is aspiring to be is like this transformative knowledge where you get to this new place and have this just fundamentally new way of looking at the world and describing the world because you actually understand the universal features, these universal abstractions that physics builds.

You mentioned sort of, you know, the threshold, the 15-step threshold sort of tells you that at least something is an artifact of life, if not life itself. Do you have any sort of, does assembly theory try and make a distinction between those 2 things, like the artifact versus, you know, something more alive? Do you have any sort of inclinations, if not at this point, that you sort of think, well, actually there might be, I don't know, I'm inventing it here, but a 25-step threshold for, you know, cutting off between those 2 things.

Yeah, no, actually I don't think that there's a fundamental distinction, but early in the theory, um, and, and also just, I still do this linguistically just to kind of describe some of the concepts, is making a distinction between life and alive. And so I wrote this essay with my colleague Michael Lachmann, who's also working with us on assembly theory. That was in Aeon several years ago, just about why we need to distinguish those 2 concepts. And the idea, from my perspective, I think Michael would say differently, is that life is all of the structures that the universe could build that require an evolutionary process to get there. So this is very contrary to standard physics, just talking about the dichotomy between assembly physics and what current paradigms in physics say is, you know, in current physics, we think you can get any object for free. It just might be a very low probability, but anything can spontaneously fluctuate into existence. And so in cosmology, this is a problem because people think brains can spontaneously fluctuate into existence. And if that's true, the dominant observers in the universe are not things like us that emerge on planets, but brains that emerge out of the vacuum. So there's all kinds of weird artifacts that happen in physics because you think spontaneous fluctuation is possible. I don't think it's possible. I think complex objects only happen by evolution. And so life is those objects. It's just the set of structures that can only exist when there is a system that evolves or learns to make them. And then alive would be the actual mechanistic process of moving through that space of possibilities. And so very early on, like even before the development of assembly theory, Lee and I were talking about like the very earliest definition of life we could agree on was a non-trivial trajectory through some space. And it's like, what does that even mean?

Oh, interesting.

Yeah. But what it means is life carves these paths through the space of possibilities and is actually the mechanism of physically realizing them.

Wow, interesting.

Yeah, so that's that, but I think actually both those things are the same thing. It's just which way do you, 'cause human language is very limited, so in my brain I don't really actually distinguish those, but I use it as a linguistic tool sometimes to help people understand the concepts that we're playing with because we don't have language to really describe all the concepts in assembly theory yet.

That was such an interesting sort of brief look into the process of creating a very deep theory. What was that process like? Like from the, you know, the moment where you sort of think, hey, I'm really going to spend time thinking about this in some way. Like, how does one start to try and create a generalizable theory of the origin of life? Like, what is that process?

Yeah, I don't know I have good advice for people that are willing to say, but I can just talk from experience, obviously. But so Well, what I did was when I became a postdoc, I had written this proposal. When I really decided I wanted to work on origins of life, I wrote a proposal to NASA for a fellowship where I conjectured that the main distinguisher between non-life and life was somehow related to information. And so I wanted to do simple math, like toy models of evolving chemical systems and understand when information became important, broadly speaking. And then when I became a faculty member, I had written this philosophy paper with my postdoc advisor, Paul Davies, which was basically— On this idea that life is a transition in informational and causal structure. So I got a little bit more sophisticated in my thinking, adding causation, and then really trying to think of like, it's a quantifiable transition. If we're talking about life as a category of nature, there needs to be a way to formalize it. And then I started working on trying to apply, you know, like there's a lot of theories on information, there's a lot of theories on causation, there's a lot of theories on complexity, like network science, information theory, you know, like Judea Pearl's like do-calculus about causation. Like there's all these like infrastructures that people had built. So I just started applying them to a bunch of different living systems. So as a junior faculty member, I was working on network structure of biochemical networks. I was working on information theory applied to ant colonies. I was working on planetary atmospheres using network science. I was doing a lot of information theory on gene networks. So I was just dabbling because I didn't have a paradigm. I was working on— Yes. Cellular automata. And I was really just trying to understand what was underneath these mathematical structures that was really there that we can't describe. So I think what the challenge is for a lot of people that approach my field or approach science generally is they confuse the map with the territory, which we often do just in general. But if, even if you can mathematically describe a system in the world, it doesn't mean you've really described it. And so I guess I've been really curious throughout my career about the holes in our current understanding and like seeing through the maps we build. So, you know, language is a map. We have a whole bunch of discrete ways of describing the world, and we move the meanings of words around to try to better approximate the things in our environment. But our language fails often. A good case of that is with the development of large language models. Nobody knows how to talk about those because they're a new technology. We just don't have the— we don't know what the meaning of the words are when we apply it to large language models. So if you just take that idea that we're all having this existential trauma and you just apply that to Like how we actually understand the world. There's lots of conceptual holes. And so basically I just started collecting a bunch of those from all these different projects. When I was talking with Lee, I think I met him for the first time at this Alternative Chemistries for Life workshop in DC. And he just started arguing with me, which is very Lee-like, but it was very great because most people, they were kind of glossing over, as I mentioned, the problems I thought were relevant to the origin of life. And he was just like, honed in on them. Like, he just, he, you know, saw exactly the problems I saw. And so over several years, he was doing this, um, this approach to measuring life in the lab, and I was trying to think about causation and information. And these 2 ideas just eventually merged because I, you know, I hadn't thought we could actually measure any of the physics. Yes. But once you have a measurement, you can build a physical theory. And if you look at the history of physics, it's always what we've done. It wasn't— we couldn't build Newtonian mechanics And, you know, like the, the law of force I mentioned before, until we had mechanical clocks and ways of measuring, you know, like balls rolling down inclined planes or Tycho Brahe measuring planetary motion. And we couldn't, you know, do relativity until we could precisely measure the speed of light and realize it didn't change with the Earth's motion. So, uh, quantum mechanics also built on measurement. Like, you get the photoelectric effect, you get blackbody radiation, all forced quantum mechanics to evolve. And I think what we see with assembly theory is now we have a way of doing a measurement which we actually is a physical measurement, then you can build a theory out of it. Um, and so that's what I've been doing ever since. Yeah, it's very nonlinear. Um, and most of the time I think I, I just, I like, it's almost like conceptual art. Um, I think the foundations of theoretical physics are like closest to any other thing I've seen as like a creative act of conceptual art.

That is really interesting.

This episode is brought to you by Brex. Fred Adler, the influential venture capitalist of the 1970s, was known for displaying decorative pillows in his office that featured a signature business philosophy: corporate happiness is positive cash flow. In today's post-SERP environment, Adler's wisdom feels particularly relevant as founders need to make every dollar work harder. That's exactly what Brex delivers. Their modern finance platform was built specifically for startups like yours and designed to help Extend your runway when capital efficiency matters most. With Brex, you get global corporate cards with up to 20x higher credit limits and no personal guarantee required. Their banking solution has no minimums and no transaction fees, while letting you earn high yield from day one with same-day liquidity. Best of all, Brex knows you were born to build, not juggle spreadsheets and finance tools. Their AI Power Platform brings cards, banking, expense management, and travel all in one place. It's simple, scalable, and designed to get you back to what you do best— building. More than 30,000 companies, including 1 in 3 US venture-backed startups, trust Brex to help make every dollar count toward their mission. Join them at brex.com/mario.

Do you see artists that like you see some connection with or kinship with where you're like, actually that type of work is, is at some degree removed doing something that is very, very similar to what we're doing?

Well, it's interesting because, you know, I haven't actively gone out and been like, you know, try like pursuing artists in their creative practice, but a lot of times they reach out to me because they have interacted with the work and they, it like, it really resonates. So I, I just had a phenomenal experience this past week, uh, with the Aspen Art Museum. They had a festival called Air, which actually was named after the book. It was called Life Is No One New knows it, is the name of this whole art festival. And, you know, they're exploring the same concepts, right? They're existentially challenged by new technologies. They fundamentally, like, art in some sense is an expression of the fundamental nature of life and like our understanding of our place in reality. And so I think being exposed to that, like, a really incredible group of people for a week, you know, really was amazing. But I've had experiences like that in the past, and as I progress in my career and I understand much more about the philosophy of science, Which I've spent a lot of time also thinking about in recent years and also observing the work I do both as a lived practice. I am a theoretical physicist, I am working on the frontier of science, but I also step outside of that and think about it philosophically. What is the act of doing science and how do I think about it at that meta level? And then I see the way these other creative cultural systems work. So science is also a cultural system. There's a lot of really interesting parallels. There's obviously some fundamental differences, but I don't think that the categories that we ascribe to people are as black or white as we see. I think some artists are doing some really interesting work in science, and I think some scientists are doing some really interesting work in art. And I think it's really interesting that we don't think about it that way.

Yeah, that's provocative framing. I really like that. Before we move on, actually, on the 15-step sort of threshold idea, in the book you talk about that as essentially a matter of probability, like the probability just gets so low that something beyond 15 steps would, you know, happen spontaneously. Given that it's like probabilistic, is there the idea that, you know, at the limits of infinity, there could be sort of something that after 15 steps could happen that would not be life or not be an artifact of life? Like, maybe I'm not sort of articulating that well, but—

Yeah, you're asking something rather deep. So So I have a conflicted relationship with probability just generally. We do make probabilistic arguments, and I think that they're well-founded, but probability is something you can assign to things you've observed, not things you haven't observed, because you wouldn't know what frequency they occur. And part of the challenge with the assembly space or this possibility space of objects that could exist that don't is one of the arguments we have for why the This explanation is necessary for life is the universe is so large in the space, like the space of possibilities that could exist that don't exist is so large. The universe literally can't generate everything. Like there's not enough time and not enough resource in the entire universe. There's only 10 to the 80 atoms in the entire universe, and there's an estimated 10 to the 100, you know, 20 orders of magnitude larger molecules with— that are like small molecules, less than 1,000 atomic units in mass. Which is really tiny, not, not anywhere near like a DNA molecule or something. Or if you think about all possible technologies, like you just stop and try to think like, what does that space even look like? Your brain can't generate it. Or we already know from like mathematics, you can't generate infinite sets, right? This is why we have infinity as a placeholder for things we actually literally cannot physically write down because we don't have enough time and resource to. So assembly theory is kind of like, imagine the universe is a potential space of possibilities, but it's resource limited. and time-limited and things actually have to be constructed over time, what does a trajectory through that space look like? And that's what assembly theory is really about. So the infinite is something we can imagine as intelligent agents that emerge from a living process, and it's a goal we can set. It actually sets, we have some sense of the shape of this space over here, so we're going to move in that direction, but it's not something we can ever physically attain. I don't, I think of infinity as a really interesting physical concept that exists in the human brain, but the way that we think about infinity is not physical.

That's super interesting. You also have alighted on the, the sort of element of the temporal through a lot of this, you know, the, the notion of time. And one of my favorite parts of the book is when you talk about how we perceive objects that are smaller in time as more physical and objects that are larger in time as sort of more abstract. The concept, just even smaller and larger in time, is something that like is worth unpacking. But can you explain that for folks? Because I thought it was so interesting and like so thought-provoking.

Yeah, that's— I'm glad you picked up on this. It's literally like one of my favorite visuals I've ever built, like conceptually.

It's so interesting.

Yeah. So, so basically it goes to the site. So one of the ways that we think about life is life is— it's kind of a continuum. I mean, the things that constructs are discrete, but like you can think about There's not— there is a transition probably around 15 or so in chemical space where you say this is life, this is not life, but it's not like a hard boundary, right? It's selection constructing objects, right, all the way across it. And so it's the same process happening. So you can just think of life as like a thing that you can actually measure, and some things are more life or not depending on how deep in time they are, how big their assembly index is. And so when you have that visual, you ask yourself, where do I sit in this physical space that is time? Like if you think about time as a material property of objects, and then, and then you can think about like, well, what would things smaller in time look like to me? Well, like I have enough memory and enough architecture in me to actually like perceive these things as discrete physical objects because my memory is bigger than theirs. Yes. In some sense. So I can actually store information. They look physical to me, but there are some things that just bigger than us. They require more evolution, more selection, more memory to encode than exists in a single physical architecture that is me. And therefore those things look abstract. And so a lot of societal things like language looks abstract because language isn't actually something that's a feature of a human mind. It's a feature of a social system, which is the interactions of many humans. And so we see like language as this kind of abstract disembodied entity. It's still very physical, but it's just something that we can't store all of at once. I think this is sort of a major distinction. And for us, it's just about the physical scale of like, where does your sensory perception architecture, your information processing architecture exist in time? And then you can see some things as physical or not, but it's all physical.

Yeah, exactly. It's like, you know, we see our pen as this physical object because the sort of temporal steps that went into that pen, you know, are smaller than us. But as you say, language has this much bigger time-space or, or, you know, the size and time compared to us. And then the implication of that is like, yeah, if you could sort of imagine yourself as some sort of, you know, enlightened, much larger in time being, you would then look down on language in the same way that you look at a pen today as this really physical, concrete thing, which is— yeah, I mean, I think you, you talk about that in the book as something you like to chew on. A year Plus after writing the book, are there any sort of new threads that you've ended up following on that sort of rough idea space?

There's lots of stuff going on right now sort of at the forefront of how I'm thinking about things. And I guess there, like I have an essay I'm writing actually on philosophy of science just generally, which has been a bit of a work in progress, but just thinking about descriptions of reality and like what they really are in science. Like I think of sort of everything as part of the architecture of our planet, including the scientific enterprise and our understanding of our place in history and these kind of things. So I've been playing around a lot with sort of implications of these sets of ideas on how we think about different things, whether it's the nature of science itself and how we understand the world and how we place ourselves in time relative to that understanding. So I did write about this a little bit in the book. I had this idea of the great perceptual filter, You know, like we have a horizon of like what our technology can actually encode as information, right? So a good example is like, you know, we didn't know about deep space and how big the universe is until we built telescopes, and we didn't know about microbial world until we built microscopes. So these are like, just like your eye is a perceptual apparatus that feeds information to your brain, these are perceptual apparatus that, you know, feed information to societies. And we can only build theories as much as we've experienced the world. So I think the more that history proceeds and the deeper in time we get as societies, the more we can come to understand collectively. And so I'm just very interested in that sort of process and what the implications are for like, we're only 300 years or so into modern physics, right? Yes. And so to think our theories of physics are complete, I think is a radical hubris that I would never place at any point in time, let alone at the very beginnings. If you have a long view that the— I think people think the future is so short, they don't think about our place in history. But yeah, so— Wow,

that's great framing.

Yeah, those are some of the things. I mean, there's lots of implications. The fundamental science stuff is also going really well, really interesting. There's a lot of ideas there, but I think some of those I have to wait a little bit to talk about.

One of the of, you know, applications hopefully of assembly theory is that, you know, as we look for life around the universe, it gives us a better way of recognizing it when we, when we see it. And you've sort of talked about why it might be a superior way of searching for life than some of these more like astrobiological signatures where you're looking at a planet's atmosphere and saying, you know, do we sense O2 or do we sense, you know, whatever it might be Why does it feel so clearly better in your view? And how do we start actually applying it as part of our spacefaring missions?

Yeah, so that goes back to some of the things I was saying about the origin of life field and what were the questions people were focused on. And so astrobiology is a really interesting field of science because until very recently, you couldn't get an undergrad education in astrobiology and you couldn't even like, you know, a grad program in astrobiology. So most people migrated just because they were really interested in the questions. And so there's no like paradigm in astrobiology, no consensus on the nature of life or even like what kind of questions we should be asking, which is why I thought it was exciting and want to work in the field, right? And that's also why a lot of people want to come to the field. But it ends up being the case that people have very disciplinary-specific ways of talking about life. And they tend to not actually be driven by a theoretical hypothesis or conception about the nature of life. It tends to be like, Like, you know, life on Earth produces molecular oxygen in the atmosphere, therefore we should look for molecular oxygen in the atmosphere of exoplanets because that would be a sign of some photosynthetic activity. Or, you know, life on Earth uses a homochiral set of amino acids, they all have the same mirror image form, so we should look for that on Mars. Or, you know, we know RNA and DNA are important, so we should detect those on Mars. Or, you know, look for amino acids in the plume of Enceladus. So it's very Like life on Earth uses this molecule, let's go look for it elsewhere. And I don't think that that is a convincing framing. Part of the reason, because most of the molecules that we're looking for are really simple, so they're easy to make abiotically. Oxygen, you know, molecular oxygen is 2 oxygens stuck together. It's abundant in our atmosphere because of photosynthesis, but we know plenty of planets that can make molecular oxygen in the atmosphere totally abiotically. So you imagine you have a water with a world with a global ocean and a high UV flux from the star. you're gonna basically dissociate the hydrogen and oxygen. Hydrogen will escape, you'll have an oxygen-rich atmosphere. So super easy to make abiotically. And so this kind of confounds our search for life, is it's the problem of false positives.

Yeah.

Same thing with amino acids. We find tons of amino acids in meteorites. And I think by the time you get to something like DNA or RNA, those are so evolutionary specific, like they're pretty complex molecules, they might only exist on Earth. And actually, I'm writing a paper right now, um, with Chris Kempes, Lee Cronin, and Michael Lockman on exactly this, this problem about like what thing, like why should we only expect some molecules to evolve on Earth and not other places? So that's sort of the way that people think about it. And so what we're saying is, um, actually we don't want to look for specific molecules at all because we don't know what forms life will take on other planets. But if we think about life as this generalized process about evolution building objects that are deeper and deeper in time, Can we actually measure this feature of molecules? And if we can do that, we can just go and measure how complex, how evolu— like how, how assembled this chemical system is, how much causation does it have in it? Um, and we would then therefore be able to, um, detect life in any substrate. And I realize now I never talked about the second feature of assembly theory, which I feel like I should cover because like people— Oh, please. Yes. Yeah, sorry. Like when we were talking about the foundations of the theory, it's, it's, um, so we talk about assembly index like this this complexity, but there's also the issue of the copy number, right? So, it's not like you could get a one-off fluke, like I have one molecule that's very complex, but I actually have an abundant collection of them. Yeah. And, I can do an analogy. So, I did the Lego analogy for the assembly index, but my favorite analogy for copy number is actually the novelty that was the little black dress when Coco Chanel invented it. The reason being, like, it's like, it, like, it's a really easy-to-copy object, right? But, it took, you know, hundreds and hundreds of years of human culture evolution and fashion design for Coco to come up with such a simple design. And so once you have that, you can do a lot of variations of it. So there's like, you know, thousands and thousands of variations of a little black dress fly. It's like a very robust evolutionary selected object. And so assembly index actually captures how hard it is to construct the object, but the copy number captures the fact that it actually was a selected feature. And then there becomes a robust architecture that makes a lot of that structure, that physical structure. And we think those 2 things are important for the scaffolding of an open-ended evolutionary process, that you need to build things, you need to build them abundance, and then they can, you know, generate further novelty. Um, and so with molecules, um, what we're looking for is high assembly index molecules and high copy number. And that suggests that there's a robust evolutionary architecture that's actually constructing those things, and therefore we would have detected life.

Is it, this may be a very naive question, but how possible is it to sort of do that sort of scanning and testing from afar? Because it seems like one of the benefits of sort of the biosignatures is that you can get maybe at least some sense, maybe you get a lot of these false positives, but to understand whether something, you know, is beyond a certain threshold of complexity and is abundant enough, do you need to be like much closer or much, you know, get much more hands-on with these things? Or like, how do you start to do that from our little perch?

Yeah, it's a good question. And actually it's very technically challenging. So we should differentiate solar system searches for life from exoplanet searches for life, which exoplanets are planets around other stars. Yeah. In the solar system, we can actually physically send missions there. Um, and so, you know, like a really exciting upcoming mission is, um, Dragonfly, which is gonna go to Titan, which is a moon of Saturn. And Titan's a very exotic environment because it basically has like liquid gas on the surface and ice, you know, water is rock and it has an atmosphere. So very complex environment. And the way that usually, like when we go into those kind of environments, whether it's Mars and Enceladus, as I mentioned, or Titan, or any of the other solar system targets, we have equipment on there to try to measure molecules, usually mass spectrometry. And the experiments I talked about doing in the lab, uh, with Assembly Index 15 were based on mass spectrometry, but like very high resolution. So in the solar system, that method can be applied. Yeah. It's just an issue of the resolution and the instrumentation, like how many fragments and molecules can you detect to actually construct this assembly pathway? And so that's a not a conceptual challenge, but just a technical challenge. And so we're working on that. Exoplanets are radically different because now you're just talking about detecting complexity in an atmosphere, detecting assembly in an atmosphere. And molecules that become atmospheric gases, particularly the ones that we're going to detect, you know, thousands of light years away, are very small molecules because they have to be light to be a gas in the atmosphere. So molecular oxygen, I mentioned, Or methane, which is, uh, you know, carbon attached to 4 hydrogens. Like very simple molecules are usually the target biosignatures. And those, you know, like most atmospheric gases have an assembly index of 1 or 2, right? So they're not gonna be on their own definitive signatures of life. And also an atmosphere is a really different kind of material than a wet organic chemistry that you might find in a cell because the molecules are very diffused. Hydrogen is playing a role moving between the molecules. And so what we had to do for exoplanet atmospheres is actually completely redevelop, um, the theory for a new substrate. And this is kind of going to be a theme, I think, with assembly theory, is like the principles are the same. We always care about assembly index and copy number in this constructive process, but the way that looks in different materials will be different. And so we did this for atmospheres. We have a paper that will be coming out soon. I'm super excited about it. But basically the, the punchline is you can measure the assembly of the entire atmosphere. So you don't look at the individual molecules, you look at all the molecules that are above a certain copy number in the atmosphere, a certain abundance threshold for detection. You construct the assembly space, which is the space of all the steps to build all of those molecules together. And then you look at that structure as the variable of interest. And modern Earth is the most complex, and it's complex in an interesting way when you look at the structure of the assembly space. database. And we, you know, we've applied it to all the atmospheres in the solar system, we've applied it to exoplanet atmospheres, and it seems to be the case that just like saying, you know, we're going to use assembly theory to detect life and we just apply it to known atmospheres, Earth is the only thing that passes the test for life, which is really cool.

And then— Super interesting.

Yeah. And you can vet it, which the part I was really excited about is kind of crazy, is like you can actually look at like atmospheres above ecologies on Earth or cities. And so urban China is the most Wow. And, you know, more complex than the Amazon, more complex than Sahara. So assembly theory actually can allow you to rank the atmosphere just above, you know, different places on Earth as more alive or less alive, uh, which is a good way of vetting, you know, the approach for exoplanets. And so we're trying to use infrared spectroscopy to measure this remotely. It's actually like the, the measurement part is hard, but we've also demonstrated that using infrared spectroscopy, which probes bonds and molecules, you can actually infer this assembly structure or actually measure it directly from atmospheric spectra.

So the implication is that, uh, China, uh, so I think you said urban China, urban parts of China,

has a technosignature atmosphere.

Wow, wow, wow, wow.

So without worrying about the specific molecules, so there are molecular technosignatures that people use. Um, chlorofluorohydrocarbons are, you know, like a molecule that is produced as an industrial pollutant but very simple. And so people have used that as a molecular technosignature. signature in the field, but it has the same sort of problems, I think, as molecular oxygen because it's a pretty simple molecule, and it might be a sign of technology on Earth, but it's not necessarily generically a sign of technology on all planets. But this is a way of like, we don't care what the molecules above China are, they're just complex enough to be indicative that there's like an urban center there.

You know, given that you started working on a lot of this prior to the, you know, recent AI revolution that sort of really kicked off in 2022. How has the way that's developed maybe strengthened your conviction in sort of assembly theory or raised new nuances? Has it impacted the way that you think about your work?

Yeah, I think there's 2 really interesting things that immediately come to mind when you ask this. The first is that before the whole, like, one of the things I would really love to do is get a large-scale effort to solve the origins of life. And most people are like, why do you care about that problem? Like, that's so irrelevant to everything happening right now. And I'm just like, you know, like going back to like wanting to shake people, like, how do you not think about this problem? So I think the thing that I'm very happy about with the AI space is it's bringing these questions about what is life and what is intelligence to forefront of like modern discourse because we're really existentially challenged by the things that we're creating. And so to me, that becomes a good motivation for the kind of work I was already interested in, which is like, you have to go back all the way back to first principles to really understand these major shifts. I like the way that Ben Bratton, who's a philosopher of technology, puts the modern time. Like, you know, there are some periods in human history where our technology is lagging behind our philosophy and others where our technology is far ahead of our philosophy. And we're living in one of those times where our technology is far ahead of our philosophical descriptions and ability to understand it. And so I think this is a real opportunity for the kind of work that I'm doing because Because if you can understand the origin of life in chemistry and you understand that as a universal phenomenon our universe generates, then you start to understand how to apply the same sets of mathematical tools to understand, is the origin of life happening in technology or in machines? And so I think it gives the work a different kind of relevance that's more tangible to everyone and not just people that are interested in deep intellectual questions. It becomes very relevant. And very important. Like, we need to know if this machine has agency. We need to know if this thing is alive. And so that's one part of it. The second part of it is, you know, I never— a lot of people, a surprising number of people have really been thinking large language models are intelligent and artificial intelligence is gonna, you know, take over human societies and AGI is imminent, you know, in part because of a lot of popular discourse on it. And I've just like been really shocked how many people have really bought into that narrative. And I think there's a lot of reasons for that. One is sort of, you know, like, uh, just economically, like incentivizing, you know, selling these technologies. There's certain ways you have to talk about them in order to sell them. And, and that becomes sort of a, a market, and markets move faster than, you know, phil— philosophy and, and the hard science does. And so that becomes the sort of early dominant narrative before we understand the technology. So, so that's totally understandable. But, um, but the other side of it is, uh, well, artificial intelligence is, you know, an undefined word for a broad set of technologies. AGI is even broader and less understood. Yes. But even when people interact with large language models, to, to think that's an intelligent thing and has like a human-like, you know, soul underneath it or like, or some kind of, you know, consciousness, like I never had that feeling when I interacted with them. And I think it's because I have a really different relationship with human language than most people do, because my entire career has been about looking for the structure under the words that we use, right? Because our words do not really accurately capture reality. And I'm always looking for the patterns underneath about what we're really trying to communicate to each other. And so I see a real fundamental difference between where computation becomes a representational sort of projection of the world and a good way of communicating and sharing information, but it's not actually the underlying physical structure that we care about. And I've been working on this a lot foundationally about Like, what is the real difference between computation and physics and physical reality? And I think that they are starkly different. And I think it's really interesting that we're having this sort of existential trauma where we're realizing that language is not everything. And to me, that's the interesting thing that's going on is it's kind of reinventing our concept of ourselves in the same way that when we invented language, we had to reinvent our concept of self. Like, if you think about a pre-linguistic society and like how they communicated and There's a lot of stuff that happens in our neural architecture that is not linguistic and can't be represented in simple mathematical patterns or linguistic patterns. And this is one of the reasons it's so hard for us to communicate with each other, especially if we're not on the same emotional page or, you know, like, so I'm really fascinated by the interaction of these technologies with human minds as a societal phenomenon and how that's gonna reshape how we think about things and how we might see the world differently when we start to recognize recognize new patterns in our environment associated with language and what's not, what can't be computationally or linguistically encoded.

So many interesting threads there. It's clearly, a large language model is clearly life in the assembly theory view of the world. Does it meet your definition, that sort of early definition of alive that you and I think it was Lee Cronin, you know, sort of agreed upon?

Yeah, that was my, that was Michael Lockman on the life alive. But, um, yeah, but so it's definitely life. Because, you know, large language models don't emerge in a universe without having a long causal chain of evolutionary events leading to them. So for example, they're trained on human language. So humans, you know, clearly have to provide the data, if not the architecture of like the underlying hardware of electrons moving around little circuits. So, um, so they're, they are part of a lineage of evolving structures on this planet in that sense. I don't think that they're like alive as an isolated system. And this is where these life alive definitions get hard because I think the biosphere is alive and the technosphere is presumably alive. These are entire, the biosphere is all the biological organisms on the planet. Technosphere is the space of all technologies. And those are clearly constructing the future on this planet. And an individual life or living thing is a short temporal sequence, an individual that persists as part of this larger structure of constructing processes for a short period of time. And the question of alive is kind of another grayscale in that About how long-lived are these things and how much are they actually generating possibilities. And so a large language model is a fairly inert system in the absence of interacting in any ecology of technologies and human agents. And so I don't think that, just like a DNA molecule on its own is not alive, an LLM on its own is not alive, but certainly they are part of a living structure. And the question is, what is the living structure that they're a part of? Is it a global collective technological societal thing that's alive. I don't know the answer to that, but I would evaluate them as an artifact, not individually a living thing.

What would have to change for it to feel less inert? Are there capabilities or behaviors that you would start to wonder if that had changed?

Well, I think one of the issues right now in the discussion is we're very anthropocentrically focused. It's like we look in a mirror and we see a human. And So I really fundamentally, like, I want to think about a large language model as the embodied thing it is, right? And it's electrons moving around on a circuit that are doing a predictive algorithm trained on human language that's outputting to me something I recognize because of all of my emotional interactions and social history as representing this thing in the world. But like, it doesn't necessarily mean those electrons, you know, moving fractions of nanometers on an electric chip are having, you know, the same experience collectively. And so the real question is like, what would it take to make something like us? And I think the only things that can become deep in time like us are evolved structures. And so it's gonna take a fair degree of evolution of technologies to build technologies that are causally deep in the way that we are. So I kind of think it's really easy once you build a cause, like a really deep evolutionary structure to kind of flatten that information and store it in a device. And so an LLM for me is sort of like a dynamic read write of the memory of a human society. It's a cultural technology. And it's interesting that our interaction of it makes us think it's an individual because our evolutionary algorithm running in our brains has trained us to recognize human. And of course, we recognize that as human because we built it and we put human in it and we trained it to look like a human. So it's like building a mirror and polishing it as much as possible to build a really accurate reflection of yourself. Like, that is what we're doing when we're building training algorithms. Yes. And we're doing it in a way that's actually, you know, for market gain. I hear people all the time saying they wish their LLM was less friendly, it'd be more creative. Like, you know, it's just, you have to think about the evolutionary forces at play.

Do you think it's possible for something to be sort of deceptively big in time? Like an LLM to me is something that in some ways has, you know, is a compression of all of human history and also like sort of does this mass, like every prompt you do is doing this massive scouring of all information across time. I don't know, is that—

Yes, no, I think that's exactly what I would describe it. And this is why I think computation in some sense is fundamentally different than causation or construction. And that dichotomy actually goes all the way back to the foundations of computing. So it's interesting because a lot of people that think about artificial intelligence don't really think about the genesis of computation as an intellectual paradigm. And one of the things I find really interesting about a discussion about AGI right now is that computation, when it was invented, was built on the principle of things being uncomputable. You have to define an uncomputable class of mathematical functions and a computable class. And then when you do the computable class, as Alan Turing did, you discover that there are some things in the computable class that are uncomputable. And this is also what happened with Gödel, is there are some things in mathematical systems that are unknowable. And yet we think when we build computational algorithms, they can know everything and they can encode everything. So that's also sort of like a weird thing. And so I don't think computation can capture reality in totality. But after Turing came up with the idea of the universal computer, a machine that can compute any computable function, which still has these uncomputable things in it, von Neumann, you know, came up with this idea of a universal constructor, which is a machine that can build any possible machine. And there's a fundamental dichotomy between those 2 concepts that I think we still haven't resolved. And I think assembly theory is really getting at something deep because it's about a constructible universe, not a computational universe. And it's interesting when people interact with our theory, they confuse it with the theory of computation. They confuse it with theories of computational complexity. But I think fundamentally there are things that have computational compressions that might look like fake deep. And then there's causal systems that are actually deep in time. And I think it, you know, for a long time I thought, you know, some people think computation is fundamental or even randomness exists in nature, but I think you actually have to have a certain amount of evolution in order to build a system intelligent enough to have a universal abstracting system, which is what the theory of computation is. And then that also invents concepts of randomness because random numbers are defined on computational principles. It's like you can't, you know, so I don't actually think the universe is random either. I think it's just non-deterministic at its base because randomness implies some kind of computability or uncomputability. So yeah, so there's— sorry, that was a lot, but—

No, no, really interesting.

The short answer is I think we don't live in a computational universe, and I think we're confusing it right now because the technologies we build are so powerful. And I think it's going to be really interesting in the next 100 years when we start to realize that. It's going to be really cool.

Wow. I'm going to think about that for a really long time, I suspect. To wrap up, I always like to ask a few sort of lighter but sort of philosophical questions. The first one is, if you had unlimited resources and no operational constraints, what experiment would you like to run?

I mean, the dream one, which Lee and I keep talking about, so Lee Cronin again, is to build an institute that is focused on origins of life. So we're nominally eventually going to try to raise money for this. Oh, amazing. Working our way toward it. But the idea would be to have a bunch of automated chemical systems. So Lee invented this architecture called the Chemputer, which is a universal programming language for chemistry. And the idea is just to run large batches of unconstrained chemistry to try to look for the origin of life without the design in it, which is what I mentioned earlier. So I guess I want no design, experiments for the origin of life. Like, how do you simulate a planet and then not put us in it and see what happens?

That sounds amazing because it really might happen, right? Hopefully you guys make it happen.

I hope so. That's the dream. That would be real synthetic intelligence, right? Because it would just be totally evolved from scratch, totally different chemistry. I'm excited about meeting some other life.

Yes, yes. And then final question, if you had the power to assign a book to everyone on Earth to read and understand, what would you want to assign?

I'm a really big fan of David Deutsch and his work, but I really love The Beginning of Infinity, in part because I think it places— it actually, it talks about this issue of universal computation, universal construction, but also talks about this idea of a universal explainer. And once you have an architecture like the human mind, we are actually capable of understanding everything. We might be slow at some things, but there's not like, like there's no limit from his argument to the knowledge that we could acquire. And so he thinks that the future is very open-ended because once knowledge emerges in the universe, like, you know, like the possibilities are pretty endless. And I think I like, I like that, that view. I mean, I really resonate with that view. I agree with him on most things. I fundamentally disagree on the foundations of physics. Like, I'm not a many-worlds person. He believes in many worlds. I think assembly theory is sort of the more physically grounded way than his constructor theory of getting at the same concepts. And I think time is real and fundamental and he doesn't, which is weird to talk about an open-ended future when you don't think time is fundamental. I don't know what that means. But anyway, long story short, I think the book is great. It has sort of a weird cult following in the techno-optimist crowd, but I think it's—

It does, yeah.

Yeah. And so I don't like that interpretation of it so much, but I think that as a philosophy that emerged at the end of the 1900s, into the early 21st century and thinking about fundamental physics and unifications we need to do. I think he does one of the best jobs that I've seen, which I think is why he has many fans.

Amazing. I've never read it, so it's been recommended enough times that I really should, and this is a better endorsement than any, so I'll have to read it.

Yeah.

Well, it's been such a pleasure, Sarah. I I feel like I'm going to be sort of defragging my own hard drive for days after this because there are so many interesting things here. But yeah, thank you so much for taking the time.

Sure, my pleasure. Thanks for having me.

That's it. Thank you for listening to this episode of The Generalist Podcast. Please subscribe on Apple Podcasts, Spotify, or your preferred podcast app. Ratings and reviews help others discover these discussions, so if you enjoyed If you enjoyed the conversation, I'd be grateful if you could take a moment to leave one. For all past episodes and more, visit us at thegeneralist.substack.com. See you next time as we continue to explore the future.