Sara Imari Walker "AI is Life" | Simulations, the Universe and the Origins of Life

You talk about like the fact there's there's there's a copy of you somewhere in the multiverse and I just don't believe that. I think like you only exist here. Uh like this is this is it. People think that they predict the future, but what they're doing is predicting recurring patterns from the past. The future will not be fully predictable ever. The fact that humans have the recorded history we have now is the only reason that we can even conceive how big the universe actually is. What is the universe actually doing as a physical system? It's building itself, right? Cuz like there's nothing outside of it to build it. Life is literally the physics of what gets to exist and why. We don't know where the boundary is between life and not life. I think artificial intelligence is life. Um but it would be a fundamentally different question about whether you want to assess whether it was alive or not. >> Sarah Walker is a theoretical physicist and astrobiologist who's working on some of science's deepest questions. What life is, where it came from, and whether we'd recognize it elsewhere in the universe. You know, kind of like in artificial intelligence. She has co-developed assembly theory that is meant to distinguish living systems from non-living ones. So, we're going to be having a conversation about the non-living systems that we see around us. And she also wrote the book Life as No One Knows It: The Physics of Life's Emergence. She's appeared on Joe Rogan multiple times, Lex Freeman multiple times, and now I'm excited to pit her up against Wes Roth. Thank you so much for being here. I guess let's kick off our discussion. It's such an interesting conversation. So I guess let's start at kind of the intersection uh of of kind of what I talk about what you talk about which is you know is AI life? I know it's kind of a broad question but maybe let's talk about that. >> Yeah I I think obviously it's on the tip of people's mind about how intelligent or how alive these technologies that we're making are and it's a very deep question because we actually don't know what life is. And so I think a lot about uh fundamentally what life is. Um coming from the perspective of like trying to study the origin of life and look for examples of alien life like I've done in my career. Um it gives you kind of a different lens to think about the question of AI being a light life. And I think the easiest place to start with that is really to think about um you know like there's kind of this fixation that life is going to be defined in terms of fundamental units and um and that like a cell is alive or um you know like we have edge cases like viruses and things like that. But the way that that I think about life that is actually I think more constructive is is more from this like first principles fundamental physics approach and thinking about the nature of things that require life in order for them to exist at all. Um and so uh you know thinking about the origin of life there's like a huge you know number of possible molecules that could be created but we only see complex molecules in living systems. And what we have seen over the entire history of our planet is the construction of all of these forms, whether they be cellular, multisellular societies or the technologies that we produce that seem to be only produced as the product of an evolutionary lineage. And so from that perspective, I think about artificial intelligence certainly as being um a signature of life. like it's something that doesn't emerge in the universe unless you have a long evolutionary history to create something like an artificial intelligence. So this is very counter to sort of traditional conceptions in physics that might say things like Boltzman brains can spontaneously fluctuate into existence which I don't think is very helpful way of thinking about things. So um so from that perspective I think artificial intelligence is life. Um but it would be a fundamentally different question about whether you want to assess whether it was alive or not or like in what ways we should assign agency or those kind of things. >> Yes. And I guess and Dylan, thank you for joining us. >> Yeah, I don't know. We we probably cut this part out, but yeah, super excited to meet you. Like definitely fan of everything you said. And I can't believe assembly Theory explains how we got to here right now. The three of us chatting and couldn't be more excited for the defining defying the odds to be here. So these questions we're talking about are of course like very interesting. I think lots of people are kind of discussing it right now because it does seem like AI is question is getting us to question a lot of things that we we thought were kind of obvious. Um life and what is thinking, what is consciousness, like all of this stuff. But I guess maybe let's take a step back because kind of how you're defining life is um a little bit different maybe from how people have heard it before. So just so people understand and there's also difference between being alive versus not alive, >> right? >> And you know we've we've talked to Lee Cronin so hopefully if people haven't seen that we have that episode previously that maybe kind of gives an introduction but maybe let's start there. So what is a life and is why is technology part of life? >> And with that definition maybe you can point out what's kind of wrong with the NASA definition because I feel like that's one of your great your greatest like gripes to the bones to pick you know. Yeah, totally. Um, yeah, and I think in order to talk about anything as life, AI or a virus or fire or any other thing that people want to assess in the universe, a a you know, atmospheric pattern on an exoplanet, you have to have a rigorous understanding of what you mean when you use the term quote unquote life. And I think this is actually one of the biggest challenges in science, particularly in astrobiology. Um, so, uh, you know, Dylan, you're pointing out this sort of canonical NASA definition. Life is a self-sustaining chemical system capable of Darwinian evolution. Obviously, AI wouldn't satisfy that definition because it's not a chemical system and it's, you know, there's debate about whether it could be considered self-sustaining or capable of evolution. Um, but that that definition actually has a lot of problems even when you apply it just to living things. Um, uh, that people would generally accept as alive. Um, so you can ask whether all life needs to necessarily be chemical. Um, a lot of people think that life is much more about informationational patterns or um or that, you know, like you do want to ask questions about AI and silicon technologies if they're alive and it's not really the chemistry that we're thinking about as mediating the processes that we would consider living in those substrates. And then there's the issue of like self-sustaining and Darwinian evolution and there's all kinds of counter examples to those things. I, for example, as an individual human being, am not self- sustaining. I'm, you know, I have to go to the grocery store and, you know, like I I'm dependent on all kinds of societal infrastructure to support my existence. If you drop me into, you know, the middle of the forest or something, I wouldn't I wouldn't survive very long. I'm not a survivalist. Um, I don't know about you guys. No, me neither. >> But I think I think most modern humans actually as individuals are not self- sustaining. Societies are self-sustaining. And so, you know, this this leads to the question of like have individuals lost their autonomy? Are they no longer quote unquote alive? and then is like a society as a whole. Now the living structure the thing that's actually persisting as a identifiable unit that has a self sustaining property. So self-sustaining is actually really quite hard because you have to dry draw the boundaries around the system that is the self that's doing the sustaining. Um, and so it's, you know, if you go into any definition of life, basically you get into these kind of semantic debates that become very deep rabbit holes about what do these words mean and how do they apply to different systems. Um, and so, you know, the resolution that biologists have taken at least in introductory biology textbooks is just to have like a list definition. You know, life is a self-replication. Life is metabolism. Life requires boundaries and and units that are are compartmentalization like a cellular life form. But when we're thinking about the possibilities for alien life, we don't know if any of those things are actually features of all life in the universe or not. And many of those features don't describe every living thing on this planet. And there's, you know, examples that people would find confusing. Like, um, I, you know, when I wrote my book, I had like this whole introductory section on defining life. And I really love this example from Carl Sean that if people used these biology textbook definitions and they came to Earth, they would think cars were the dominant life form because they meet all these criteria. And I think most people are not really satisfied with thinking technological artifacts are life. But as you're pointing out, AI is now challenging that conception because people are interacting with it. they feel like it's an animated thing. Um, and so we want to prescribe these properties of life to it. But again, it goes to like we don't know where the boundary is between life and not life. So all that to say like my approach to the problem and and what Lee Cronin and that you already mentioned and I have been working on is whether we could start from first principles and I'm trained in theoretical physics and I'm interested in whether life is actually a category of nature that the universe recognizes outside of us, right? Like, so there's ways that we as humans describe the world. And then there's the way the world actually is. And science does its best work when we build abstractions that allow us to understand the world outside of our experience. That's sort of the way that Popper described it. And I really I love this kind of idea of a visual. It's like we have our conscious experience of the world and the way we interact with it. But what we're trying to do in science is actually step outside ourselves and understand the world as it is, not as we perceive it. And when you do that with the phenomena of life, I think basically what we're trying to do is develop um whether there are physical principles or laws of nature that actually describe what life is fundamentally. And you have to have some kind of hypothesis or conjecture about what life does uniquely that you would need to build a fundamental physics around. And so this is where the ideas of assembly theory came in that we were thinking that life is the only mechanism the universe has to generate very deep complex objects like like deep in terms of the amount of construction history or causal possibility in the object. I can explain what that means in a little bit more detail but I know you've already spoken to Lee um at scale. And so we we conjecture that the abiotic universe can only generate so much stuff and most of the possibilities you imagine actually can only come about because there is a living architecture that has memory history information processing and through the process of evolution can actually discover invent new material realities and and actually build them and realize them and AI is one such thing. Um, but because it's part of the human lineage and has all of our cultural history now embedded in this artifact, you know, it looks very alive because there's this 4 billion year history of evolution behind it that was necessary to generate such a structure. >> Yeah, >> that makes sense. So, just so so I'm clear and maybe um because the the way that this really made sense for me is that, you know, we kind of draw circles around life based on what we see. It's like oh well humans okay that's within the circle animals within the circle >> viruses ah who knows kind of right there somewhere like somewhere you're saying we're almost able to approach it from a different perspective like we can actually assign a number to it assign an actual metric >> and um the kind of I guess the assembly index right so let's say we go to Mars right we take some sort of a mass spectrometer we find these things and we find that certain things there are more complex than a certain threat threshold we you might be able to say okay even if we didn't didn't see a single organism we might be able to look at that and say we know life was here at some point right is that okay >> yeah that's exactly right and for many of the cases for astrobiology of interest we're not going to be able to resolve any individual life form you know for hundreds of years potentially so if you think about exoplanets like these are planets around other stars and the closest exoplanets are four light years away so even if we could travel this speed light. It would take us 4 years to get there. Um these planets, the only data that we're going to get from them is atmospheric um properties, right? So we we'll be able to maybe get the atmospheric spectra and then from that we have to infer the presence of alien life forms on that planet. So there's there's no ability to say, you know, there's a large language model on this planet or there's a cell on this planet. like we have to be able to detect the imprint of life in the molecular composition of an exoplanet. Um and so you know from my perspective obviously like we have a lot more rich data and rich observational experience here on Earth. But ultimately the question of what life is comes down to this really fundamental question of like if I looked at something and I didn't know what it was, how would I recognize it as life? And I think this is a more first principles approach um and why it's relevant to sort of current AI discussions. why astrobiologists have a lot more to say about it. um or people thinking fundamentally about life more specifically and like the kind of approach we're taking with assembly theory is that you want a theory that you can validate in a diverse set of materials across diverse kinds of living systems um and non-living systems and actually be able to say I can measure the amount of life in this system and this this system is you know like assembly theoretically we would say this system is this assembled and it's both assembly index and copy number that gives to um this kind of um I think about it as a depth into a possibility space. So if you imagine that at any point in time there's a certain number of possibilities that can be created. So if you think about all possible technologies or all possible molecules or all possible Lego structures you know only a certain number of them you can actually build at any given time and that constrains what you can make in the future. And so assembly theory is basically talking about this possibility space as a physical space and that life is moving through the space of possibilities, selecting structures and making potentially more complex ones. And this allows us a way to detect when life has emerged in any material, not just in chemistry. H okay. So I I'm not sure if you caught this. This is only a few days old, but there was a company called Eon Systems that um just recently has finished putting an entire fruitfly connecttome into a physics simulation. So, we have we have a fruitfly, every neuron that was in the biological version, which went through actual evolution and actually had a chemistry background, has now been sort of digitally copied. It's moving around in a physics environment where it thinks it's a fly. It's looking for food. if there I don't know if it feels hungry or not if the nervous system gives it those signals but either way they would be all digital. Um maybe you could kind of tell us like does that kind of swapping just work with assembly theory or what kind of things are coming out of it? >> I mean I actually even had an allergic reaction when you said it thinks it's a fly. I was like >> okay yeah maybe not. Yeah the >> your mental model thinks it's a fly but whether the model thinks it's a fly is another question entirely. And we are so bad at actually embedding our understanding of the world in other cognitive agents even just with animal behavior. Um uh so I I think that's but there's also another example which is like which was also in the literature this week where there was a simulation of like a whole cell um and you know like so like there's a a simulated cell and undergoing division and so I think you're going to see more and more of this because people are so excited about the computational power we have now that there's some conception that we're taking physical reality and we're transporting it into the computer. Um, but I think that's a completely um uh false uh kind of statement. Um, and I I don't know the best analogies um to describe how I think about this just yet, but it's like it's like we observe the world and because you can replicate your observations in another system doesn't make that system the same thing as what you what the original thing was. Um, and so it's it, you know, I can give an example actually with fireflies because I was just talking about this yesterday. Um, but my lab had a paper out on like a firefly model for SETI. Um, which was supposed to, you know, like we were basically using um, fireflies as a nonhuman communicating system to model the possibility space of alien intelligence. Because every single model that's ever been put forward for aliens and how they might communicate over interstellar distances has been based on humans in our technology. And we obviously don't know what aliens are like, but at least having a non-human model allows us to expand the thought space that we're working in. And so we used a Firefly like it was actually a model developed by Firefly Communication and put it against a pulsar background. So pulsars are are flashing um you know in in out in the universe and they have very regular patterns just like fireflies. And then you could show that something that was ev evolutionary adapted against that background to like distinguish its signal like a firefly would try to be against other fireflies in the same community would actually be a distinguishable pattern. Now the reason I'm bringing that up in this context is um it got me down this deep dive trying to learn about how fireflies perceive their own signaling. And you know we have this very regular pulse p pattern that we see but actually like we don't understand how the fireflies interpret that symb sign signal. So like the state-of-the-art is actually to try to get in the firefly brain and understand its own conscious perception of the world and what actually sensory in input is it getting um and how is it responding to its environment based on like the mental model that it's building of the environment. And you know like to try to do that for all these diverse organisms is actually super hard. So I see a lot of a lot of work where people model a system but they model it from the human labeling the data. They don't model it from the intrinsic perspective of the organism or entity that they're trying to study. And I think this is a really really hard conceptual leap. So and it gets into the issues of consciousness and things I know you guys talk about but you know it's one thing to ma model pattern behavior. It's another to say that behavior has the same experience of the world as um you know as another entity displaying the same behavior. You even know this comparing two humans. It's like two people can act exactly the same and have totally different feelings about it, right? So it's like it's not it's not a non-obvious feature of reality. So I don't I don't know why people are ascribing so much behavior to simulation or so much human agency, human cognition, consciousness to to algorithms except for the fact that it's the first time in human history that we've seen these particular things represented in external environment and not just in our minds, right? But it doesn't mean that they are our minds. Um, and most of human history has actually been about the externalization of our thought processes in our physical environment. Um early examples are even like when we first built physical structures and we could remember like and you know like actually human culture started because we built structures in our environment and they persisted long enough that like the next time the human migration pattern went there they started to have a memory that there were people there and they could build up a culture over time because they had things that were longer duration than just a parent teaching a child. Right? And then and then we get written language we have even longer generational time scales. So these technologies that we building whether it's like physical buildings in our environment written text or a machine learning algorithm or a simulation of a fruitfly are just acquiring more and more knowledge that we've built into um our own minds and our environment and allowing us to see those in higher resolution and actually externalizing you know like these models we have. so we can understand them better. Um, so I think the interesting thing for me about the fly is is that it raises those kind of questions like how would you know that this thing actually thought it was a fly or you know and and if you can if you can ask better rigorous questions then then that gets more interesting. >> Yeah. I know sometimes I feel like I'm less conscious in the morning than I am until after my coffee and and throughout time it's kind of changed. So >> coffee addict I get that. >> Yeah. Um but if we do uh so I was thinking maybe we could like a little thought experiment though but if um we keep kind of moving down the route where the fruitly gets more and more kind of onetoone simulation just obviously just simulation it's not alive or anything but it it does have more and more of a nervous system that landscape that it's in feels closer and closer to our reality and maybe millions of years of evolution can take place digitally in a much quicker time frame where you add multiple fruit flies and and they mate and and over time like all sorts of interesting things start to happen. Do you do you feel like consciousness and and kind of higher level like ordered assembly could start to take place and then we could maybe learn about consciousness or kind of the things that we are >> um I I am not convinced that deeper simulations are actually going to get us to a deeper understanding of reality. Um so I I think the the challenge for me is um you know the the universe has clearly constructed uh very complex structures. So, you know, like if I wanted to like simpler than a fruitly, I just want to simulate chemistry, right? Like obviously like a fruitly is made out of chemistry. And I like these chemistry examples because they're a little bit more concrete, but I think these kind of arguments port over directly to the question that you're asking. But if you think you you think about um you know all possible molecules that are are fairly small um you know like a amino acid size like very small molecular building block used in life and you want to say I want to simulate the space of all these molecules um there's not actually enough resources in the entire universe to build all those physical possibilities and we run up against like computational limitations all the time that like um you know like why are we having to build these massive data centers and and computational things is because actually it takes physical resource to do simulations. And it's really interesting that the universe has actually generated structures that it can't simulate. And I don't really know what to do with this feature, but like the the universe is like if you think about the fact that the universe has like a finite amount of material in it. That material is the physical foundation of what can be used to build a computer and to do simulations. So you have a finite amount of time, a finite amount of material. that finite amount of time and material actually is insufficient to simulate itself. Um and so I think that we will hit with any system uh hard physical boundaries which the system actually can't describe itself and and you know this comes from the foundations of mathematics also if you look at like girdles theorem and the foundations of computation they were actually like computability as a theory was built on this idea that there are things that are not computable and we want to talk about computable things not computable things and then even in the space of computable things there are systems that you know like try to talk about themselves and can't and lead to all these sort of like uh mathematical and computational paradoxes. And so it's kind of weird to me that that out of a a tradition that started from deep principles of unknowability, this was what like Alan Turring was dealing with with like the halting problem, things like this has emerged a 100 years later, this like hubris that we're going to be able to simulate all of reality on a computer and think that that's actually like a physically realistic possibility. Um, and I and so I I just don't I don't I don't see that being the ultimate path to understanding things. It's like if you make the map too precise to the territory, you haven't actually gained any understanding. You just have uh the territory duplicated. Um it's like it's not what we do in science. It's like actually science we build we build more foundational understanding and in in some sense uh we build simplified representations um that are deeper principles and more abstract because they actually allow us to understand the system better than if we could just fully replicate it. So so I think I think there's some kind of like cognitive mismatch between what we're doing when we're it's almost like it's an easy way out for us to say we can simulate these systems and therefore we understand them. Um I I just I I don't I don't find it satisfactory on any level. >> Yeah. Yeah, that makes sense. One very interesting sort of uh concept that popped up a way of thinking about what this universe what this universe is is kind of thinking of it as a creativity engine right so if it's generating novelty life >> and part of life is us creating technology etc. So, I mean, it makes sense why, like, for example, a coffee mug. We're all coffee lovers, I'm sure. Uh, you know, that you can you see that shape in some form replicated across the earth there. Oh, there's I see one now actually right there. I don't have one. I have one. >> I can't reach mine, but it's like just off screen here. Uh, you know why? It's it's useful, right? And so, we replicate it. We found, oh, this shape really does something. And I guess something similar maybe. And correct me if I'm wrong on any of this, but something similar is happening with life. It seems like, right, there's something that's quoteunquote useful that gets developed and it sticks around because it becomes part of the DNA and it multiplies and DNA therefore is like one of the more stable molecules or whatever just because it keeps multiplying if it's useful. >> And so it almost seems like like the whole point is just to build more novelty, more useful things, more complex things. So I don't know. I don't even know if I have a qu a question, but is that what this whole thing is about? Is just this creativity engine to create complex stuff. >> Yeah. Um I mean this gets more into sort of philosophical questions about the nature of life and also like does the universe have a purpose? Um and it's interesting because like very early days of like trying to build theories to actually um understand fundamentally what life is and then to be able to look for life. Um I started to think about life as the mechanism the universe has for bringing new things into existence. So like life is literally the physics of what gets to exist and why. Um and so um so I think um and I so like my conception of of the way the mechanism the fundamental physics of the way our universe works is very different than standard physics and it comes from studying fundamentally the nature of life. And so standard physics would tell you that the universe started from initial condition and we have fixed laws that exist outside the universe and they evolve the universe for all time. And from that perspective, it's easy to think you can sit outside the universe and simulate it because you just have a you know a mathematical equation. You just need the information initial condition and the whole thing can be simulated. And that's not my conception of the universe because it assumes that you have an observer that exists outside the universe that can describe it. And for me the universe is everything. And in fact, the universe is not like a static block structure. It's the emergent property that like a mind inside the universe would talk about is all of the creation that it sees around it. But what is the universe actually doing as a physical system? It's building itself, right? Cuz like there's nothing outside of it to build it. Um, so, um, so I have a a like a sort of sense of the universe as a self-constructing system. And that's why I think this idea of possibility needs to be really considered in fundamental physics very seriously because what the universe is doing is exploring what's possible in generating physical structures that can persist in time, which allows more things to be possible. And that fundamentally is what life is doing. But I also just think it's a more general uh foundation um for what physical reality is. And this gets into reasons why the universe can't be simulated. Um because it it's it's like you can't build something outside the universe to simulate the entire universe. Um and it's just it's not like within the knowable territory to be able to do something like that. And it's actually not the mechanism of how the universe generates novelty. So like when you're doing things like simulating the fruitfly, you're basically encoding the past observations of all the behaviors of the fruitly in the current moment and you can you're building a memory. So like we talked about part of this whole assembly theoretic perspective is that the universe is acquiring deeper and deeper memory to construct more novelty. So I see the machine learning algorithms as being a very deep uh kind of memory for a tech technological intelligence to have but I don't see them as a mechanism for generating novelty on their own or like actually leading to sort of the further scaffolding of the creativity of what our system is doing with humans definitely but like on their own I like it's it's hard to know or on its own I don't like we anthropomorphy Even the way we talk about these technologies are really riot. >> Yeah, it's very diff. Yeah, there's certain words that just kind of like the observer or a person or intelligence or consciousness is baked in somehow. We have to >> Well, they were sold to us that way, right? So, it's like as soon as you know like as soon as chat GPT comes online. >> Well, yeah, exactly. And Claude, right? And I love Claude. Claude's my favorite. But um but I think um but I think you know like it was you know like in the sort of early marketing of these technologies they were phrased in a certain way that was very anthropomorphizing and so people have adopted that anthropomor like even the way that they're trained to talk to humans. It's very much to be user friendly and to give the perception that they're um they're like us even though it it's literally an it. It's a very deep uh uh it that records a lot of information and can you know predicts stuff within the space of that information. But I yeah I think it's um it's really hard to think uh cognitively clearly when human language because it's not set up for this new technological phase that we're in actually like we don't have languages to talk about the things that we're interacting with. Um and so we're getting confused by using words in wrong context. So people do this all the time. Like I I basically think like the word agency has been completely reinvented in the last year or two, right? Um but we but we carry over the concept of the old meaning of the word to the new meaning of the word and that's when people get very intellectually muddled about and we lack conceptual clarity. >> Yeah. Especially when there might even be like intentions behind that to help things, you know, and you're like don't hijack that word. But sorry. Go ahead, >> Wes. Yeah. No, I was so angry. Well, the first time I saw like your AI personal AI agent on a poster in an airport, I was like so so agitated. I was like, you can't take our words. >> I definitely want do want to come back to agency and because I know consciousness is also kind of similar to that. I just want to just one quick thing based on what you were just saying that I'm kind of just popped into my head. So one thing that um I I was very very curious with the development of AI that I've noticed is in all the science fiction books and all the stuff before we tended to describe AI what we thought how we would develop it as this highly engineered structure with you know similar to a Formula 1 car now it really appears like we're not really engineering it we're growing it we're creating the environment so it's some part it's like some emerging part of physics or the universe and that seems to me >> very similar to what you're talking about how at the base level or what some fundamental property of the universe is just to this these things emerge that create complexity. This seems like a new idea. So maybe is >> I think I think I think that's actually a very deep point that you're hitting on. Um and so like with assembly theory and the way we talk about the assembly index, it's an intr like assembly index is supposed to be an intrinsic property of the object, right? So things some things are very causally deep structures like like the three of us are. Um but the point of doing that and making it a feature of the object means that we can do physics and we can do measurement on the object and it becomes an objective feature of the object. So so that's a a really important scientific point. But what that becomes evidence of is that there's something in the environment that can make that thing exist because the thing can't build itself. Um and so so all of this complex structure is actually evidence of complexity in the environment in some sense that it and this is part of what I'm saying about the universe being self- constructing is like I can't there the idea in physics I mentioned earlier about this idea of Boltzman brains like this is a very standard idea very consistent with current um current uh like theories in fundamental physics that a brain can spontaneously fluctuate into existence at any moment in space and time and you know this becomes paradoxical in physics because the the idea is that if the brain you know can fluctuate into existence there's no way of you knowing that you're not such a brain um because your experiences would be exactly the same you just fluctuate into existence and for the flash of the second like you're having all the experience in memory that you would have had um and then you flash out of existence but you wouldn't be there to know that um so you know like so it's it's sort of par like because you wouldn't know the difference observationally if you're a Bman brain or not I think it's just a ridiculous assertion because there's no like there's no evidence in our observational environment that brains form this way. Um and we do have persistent structures but more importantly I think physically brains emerge um in bodies uh on planets along with populations of other organisms that are evolving in the same way. So I don't think this idea of a isolated brain is at all sensical. Um, and part of my reason for pointing that out is it requires this very deep construction history of like, you know, systems that can make cells and the cells are reproducing and the cell, you know, like and then the multisellularity gets selected and even like features of the human brain are so socially constructed. It's like one of the reasons we have a hard time seeing reality for as it is is because our brains in early development pick up on social patterns and social cues because the thing that's necessary for a human baby survival is to be socially adept and actually be cute for its parents and get that, you know, like and be able to pay attention to human signals and not the environment. So, humans are actually really pretty bad at seeing reality because of the way that our social brain developed. And I think some of the artificial intelligence technologies are tapping into that sociality aspect. And like why I like people that are kind of on the spectrum are like autistic and I think they like tend to be better scientists is because they they don't have the social cues so much and they're their their cognitive architecture is actually a little bit closer to physical reality. I think, you know, Lee Lee, not he's not necessarily on the spectrum, but he's like a great example of this where like the way that his cognition works, I think, doesn't pick up on the social human cues as much. And he's very good at actually like understanding the physical world outside of these artificial architectures that we build through human language and social cognition that are very useful for our functioning in a social system, but don't really tell us how reality works. Um, so that was a bit of a tangent, but your your point was very deep. So there's lots of related things. >> Yeah, that's interesting because I can see kind of that connection between like how you view the Boltzman brain and then this idea of like the fruitfly brain and and those things and how you be like, "Okay, it's not the same." And >> now we got a good sound bite for Lee Cronin, too. Like, so we got everything we need. But, um, I was >> I hope he's okay that I miss some he misses some social cues, but it pays off in other ways. like it's no big deal, you know? I think >> he's friends, so I think he'll be okay with it. But I, you know, I I I it's one of the reasons that I respect him so much because I think he he's just his own person and like you can really value people like that. >> Yeah. Um >> I absolutely take it as a compliment. Somebody actually has their own kind of thing in the world and sort of knows who they are, everything. >> Um so I was going to maybe talk about um you have one of the more interesting takes I've always thought on AI replacing us. So it comes up all the time like on on our audience like are people not going to have jobs? Are they not going to have purpose? Like AI is going to replace us. And I've heard you actually describe it before as more like a category error. It's almost like you're not stepping back big enough because you said it's like cells fearing multisellular organisms. Like >> maybe you could just explain maybe a different way to think about the disruption AI might come bring to humanity. I mean I think so so on the shorter time scale like just human human cultural evolution in the last few hundred years like anytime we invent a new technology there's always a job displacement but it doesn't end up being a net bad for humanity because like it opens up new niches and new new new kinds of things that we didn't even know we could do. So I think one thing is like because our our lifespan is really short um and we get kind of accustomed to the world functioning a certain way uh it can feel radical within our lifespan that things change so much and oh my god jobs aren't what they were 10 years ago or they weren't what they were uh you know um uh you know 20 years ago. Um but it's it's kind of crazy when you start to think about the fact that the entire human population turns over about every hundred years. So no one alive today was alive 100 years ago. Um and you know nobody alive at that point was alive 100 years before that. Um and so you have this entire turnover of the world population on a typical time scale of 100 years. So like we don't really have that longevity of experience to really realize that these kind of radical shifts are maybe more normal on the time on on the trajectory of of human evolution which I think is kind of an accelerated example of evolution or on the template of like the biosphere as a whole over billions of years. So um you know obviously there are major evolutionary transitions that have happened in the past history of our planet that were maybe quite bad for some of the species alive at that time but not universally bad for the evolution of our biosphere as a whole. And so one example is um you know when you you evolve multisellular organisms uh from unisellular organisms there's a lot of trade-offs there right like a unisellular like a single cell now no longer has any autonomy and in fact the multisellular body plan is such that almost all of the cells in your body can no longer reproduce right but but our species reproduces because we have a germ line and those cells can reproduce um and so there's all these kind of like new cooperative layers that emerge at different scales and obviously I'm very happy that uh I'm the entity I am although I do find it really shocking that every single organism on the planet still goes through a single cell. Um you know like every single multisellular organism on the planet at one stage in its life cycle is a single cell which is like totally radical that we have to go that far in history evolutionary history to to actually produce structures like us. Um but so um so if you so I I guess you know where I where I'm going with this is I tend to think about life on the many billion years time scale. And so where we are in this sort of like current hundredyear moment that everyone alive that's a human that you can talk with is experiencing is a very very very tiny slice of that. And if you try to think about the bigger picture of like what is happening with life large over billions of years and like what is the purpose of life and the progression of life like purpose is a hard word to use in science but I'm just using it because I think that's sort of how intrinsically how we feel about these questions just as humans. Um it seems to be the case that our planet is the most creative planet in the universe that we know of in terms of like the structures it's generated over time. the novelty that's become apparent and the kind of possibilities that open up. And so for me, I don't see, you know, it's hard because obviously a lot in our environment is changing. People are losing jobs or people are having existential crises because they're, you know, able to talk to a chatbot. But in the grand arc of like, you know, what's coming next, these technologies are enabling a lot. Um, and presumably they're actually going to make human lives a lot better if we look at the sort of trajectory of what technology has done in the past for humans. Um, and and it's going to change the landscape of how we interact with each other and also how we interact with our technology. But it's not fundamentally different than I think things that have happened in the past. And I'm I'm really confused when people think this is just some major level shift and like nothing like this has ever happened in the history of the universe or the history of the human species or the history of our planet. like suddenly there's an artificial intelligence that just popped out of nowhere and it's going to change everything and like doomsday is going to happen and you know th that narrative is not new like people saying that kind of thing is not new because when new technologies have emerged in the past people have traditionally been scared of them they have had you know dreads of existential threats my favorite one when LLMs um you know started becoming very popular that I didn't even know about was like you know in ancient Greece they were like afraid of having people read because like and write because they didn't want the dead talking to them. Um you know so I I think you know in photographs like when photographs were first you know becoming uh you know a prevalent technology people were afraid they actually took your soul and that was what was recorded in the the image like they were you know like soul stealing devices. I think I think this idea that our technologies like what we do when we externalize our mind and the environment and now we can do all these new things with our environment that used to be internal cognitive processes or things that were only related on human communication. Like stories were always oral and now we have written stories so we can read people 2,000 years ago without having to go through every voice in between. Like these kind of things, they're traumatizing. They're existentially traumatizing but um but they are part of the human story and humans are uniquely poised to develop such technologies and they end up leading to all kinds of brilliant things that our species could do that we couldn't do before. So I I'm I'm a glass half full kind of person I guess. Um, but I'm not I don't I don't feel threatened by uh AI intellectually. Like a lot of people that think they're going to replace like the stuff about them replacing physicists I find so ridiculous. It's you know like it's like the new like I don't know you guys see this stuff in the news. It's like oh yeah they can do string theory and they just did this theorem proving thing and they're going to replace all like physicist is the smartest job and they're going to replace all physicists therefore all intellectual work is done. Is like kind of you know one of these narratives that I I just I find it insane. I'm like, clearly they don't know what physicists actually do. Um, so, um, so I I I don't know. I I I I I think humans are, um, an incredibly adaptive species, an incredibly brilliant kind of physical configuration for a system to have. I'm very grateful I exist. Um, but I don't I don't limit the potential that we have or how we integrate with the rest of what the planet is doing. and the fact that I, you know, we live on an incredibly alive planet and it's getting more alive all the time the more creative we are and the more things that we actually build. Um, so it's I yeah, it's not the way other people think about it, but I have reasons. >> Okay, so we still don't understand how distributed brain activity kind of creates a unified conscious experience. Um, if we can crack that in biology, like what do you think it might tell us about AI and if AI could ever have the same thing? >> Um, well, I think it would answer that question, right? So, so but I I but I I almost wonder if the way that we're approaching it is not from the right perspective, right? So a lot of um um a lot of the work on trying to understand consciousness is like basically picking apart the human brain and looking for like the neur neural coralates of like what we call conscious experience and you know the best way to understand if something's conscious is verbal report out even now which is one of the reasons that LLMs are even part of the discussion right because if if if an entity in the past before all of you know the stuff that's happened in the last few years if something could state it was conscious. That was like the best signature we had of whether it was conscious or not. But that was never a good scientific criteria. It was just this this is what we can do to start to approach this question and maybe build better questions from it. And so I think this is something that is really hard to understand about the process of science is that we get stuck in paradigms for a really long time where like we are banging our head against a wall asking the same questions and we have a really hard time trying to formulate the best way of asking them and consciousness has been this way for decades just like um you know the problems of alien life have been this way for decades. So I draw a lot of analogies between the the sort of hard problems of consciousness and the hard problems of life and what people in the neuroscience field have been dealing with with trying to formulate consciousness and what astrobiologists have been trying to do with life detection and formulating life. like the if you if you forget about the questions and you just look at like the kinds of approaches and the way that people have to ask questions when they don't have good conceptual foundations or theory to guide them like it's almost identical and it's usually by by recognizing that there's a phenomena you want to explain whether it's consciousness or life and then finding a coralate for that phenomena whether it's a pattern of neural activity or a DNA molecule that you know like so far we only find in in biology on Earth um and and you go use that basically as a proxy for the phenomena. And so I would and so so the way that you're posing the question Dylan is like well if we understood the patterns of the neural activity in the brain we would understand consciousness and then we would be able to understand if an LLM was doing that. But basically what you're doing is you're like you're pattern matching from this physical system some way that you've labeled the data and you're calling that consciousness and you're moving it over and saying like does this match that data? And in bios signature science we do the same thing where it's like well earth has these sort of atmospheric gases. If I find them on another exoplanet uh you know I'll have detected life because the pattern matches but >> I might not be extract am I not abstracting enough or generalizing enough? >> No that's exactly my point. Um so I think I I think the way that I would appro approach those kind of questions is you actually have to have a conjecture or a hypothesis about the phenomena you're studying and you can't use the phenomena to explain the phenomena. So you you know you can't take consciousness as fundamental and then ask questions about consciousness. You have to say what are the properties of the physical universe that allow me to talk about a thing like consciousness existing. And so this is the way that I've approached the question of life is like I can't assume life to explain life. I have to say well if I forgot about my human categories of nature and I was really just trying to say there's a regularity of the physical world. I want to describe a particular pattern in the physical world. What would be my first principles approach to describing the things I associate with that set of phenomena? And then can I build theories that allow me to test that against the cases that I think are actually relevant? And this is a really hard process to do because you have to do it iteratively. So like when I was first starting thinking about life, you know, I h I had like a bucket list of like 10, you know, things that I think are really relevant to uh what life is about. And they would be things like um you know, information plays a causal role or you have to have dynamical systems where um the states and laws are not separate. So, I I talked about already like an initial condition of the universe and a fixed law, but like those are human categories. Like somehow they collapse when we're talking about living systems because they're open-ended and the rules are changing all the time as a function of the kind of systems that emerge. So, like there was this whole list of things and basically what it is is just a conceptual map. But you don't want to hold on to any one of those definitions because the whole role of a scientist is to assume we don't know what we're talking about and we want to build a theory that we can test against reality. And what I see too much is people putting too many assumptions in and then looking for the phenomena as they predefined it rather than letting nature tell them what the phenomena is. And we are so in this place with consciousness. It is like and consciousness is like is is obviously much harder because it is our experience of the world. But everyone's very fixated on explaining experience rather than explaining what would be the consequences of whether an entity had experience or not. what does that even mean? Um, and so for me, I have like, you know, my own developing theory of consciousness, which is just basically about like causal depth and possibility space very related to assembly index, but there are sort of testable consequences of that that, you know, eventually we'll be able to work out with like all the other stuff we're working on. It's too long a bucket list. Um um but I think I think even Dan Dennett wrote this. He's like, you know, you don't want to ask the hard problem of consciousness. You want to ask the hard question of consciousness is like what is it empirically um tractable? like what what would what would be the consequence if consciousness is a real physical thing? And right now a lot of people think consciousness does not impact physical reality somehow it's like internal to reality because that's the way that we experience it. But that I I just don't think that's the right framing. I think it's it's it's got so many problems the ontology of the way that we talk about this problem. Um, >> yeah, it it this is so interesting, so deep, and uh this really helps me get my mind around it a little bit hopefully because we've we've seen as we're talking more about AI, I'm really no one knows what consciousness is and we have such different ideas of what it is. I really like how you're saying so like what is the proxy for it? So we can't measure somebody's subjective experience, but is there a proxy? Like what does it mean? Uh can we >> actually Yeah. So that that's kind of what I'm saying. But I'm even saying I'm not sure that that's the right question to ask. Okay. Like that became the hard problem of consciousness because that's the part that everybody thought they couldn't explain with science. >> But I but I think I think >> you don't you you want to you want to ask questions about like imagine that you you were just observing the world and you wanted to pick out what were the conscious entities in the world, >> right? like that that's actually the scient because as scientists we have to we have to do science from the outside the system but obviously we're inside the system which is one of the reasons that I think life and consciousness are hard because you can't take fully that external view you need but but what we what we've done so far is is like like physics is built on this for example this idea that you can observe the universe from the outside and describe the laws and actually almost all of science is built that way and um and it's necessarily built that way um because of the historical development of these fields. But when you get to consciousness, you're like, well, I am the system or when I'm life, I am the system. Um and I think it's still possible to do what we do in science and objectively study those systems as if you're from the outside, but you need theories that account for the fact that your entities are in the systems that are being described. And so one of the hardest things about what we're doing with assembly theory that I think you know we haven't communicated very well but I think very deeply about is it's basically a theory of physics that it's possible to describe the universe from the inside like there there's no external law in the theory like they're all internal to this objects that are being described. Um, and so I think I think that the the fundamental issue is often in science we don't ask questions the right way because we have paradigms that don't work for these problems. So the question is what paradigm would work for this problem. But that means that really fundamentally it's not a question of the answers being wrong for the questions. It's like your questions are wrong. >> Um, and so I think consciousness is in that space. So I don't even know how to articulate exactly what I think about it because like the language we have is so limited um and it's stuck in these ways that we have of thinking about things now and I really all I'm saying is like there is a way that we talk about things right now but we need to get to some other space entirely to ask questions about the nature of this particular phenomena because these ways are not working and I don't think that doing things at scale having better computational power is going to fix this problem. I I just I don't see that happening and I don't see happening uh that you know I think what we'll do by this you know neural mapping to algorithms is we'll build better algorithms but we won't have any deeper insights into the nature of consciousness. >> Um this be a kind of quick answer if it doesn't work but what do you think about the word intention? Does it bother you like it has too much assumption around it or do you kind of feel like when the universe is sort of moving towards uh like higher assembly objects that that is an intention? First off, I should give the caveat that all words bother me. >> Okay, >> I have a little bit of that as well. Yeah, I know what you mean. >> Um there like just words are so radically imprecise to the concepts that we we're aiming to communicate. And um and I think people think words carry much more meaning than they do. Um so oftentimes when I communicate with people, I don't look at a specific word. I look at like sort of the pattern of what they're communicating um to try to actually understand the meaning that's being conveyed. >> Does that annoy people when you do that? cuz I annoy people when I do it. I'm just wondering, do you get that same reaction >> or no? >> Um I I think um probably >> Yeah. Okay. >> I think it's I think it's also quite hard for people like like you know it's like people be like what do you mean words don't have >> it's usually that they haven't thought that deep about it. They're like I don't know. I've just been using that word like I'm not really sure what I mean. Maybe in this context or that one. Th this is another like I mean why do LLMs work so well is like the way the sort of structure that we have for language is a set of associations that humans use as pointers to objects in the physical world and you build a common semantic representation in your mind by communicating with someone that understands that pattern and physi like what what happens with kids you have to physically point to the object so they understand the meaning but humans get into these abstract spaces with our words where there is no longer a physical object object for us to point to. So the the more abstract our language becomes, the harder it is actually to say that any two individual humans actually have a shared representation of what that concept is. Um but with the word intention, I don't have actually I have no problem with like uh teological language, which which intention is like goal- directed language or purposeful language um as a scientific language. I think I just want to put the caveat that like these are properties that emerge in the universe. They're not um imposed on physics or the way systems behave by some external designer or intelligent agent that's observing the universe and set the laws of physics in motion. For example, like these are these are properties that um actually emerge along lineages through the process of evolution. So when you say something has intention, all that suggests to me is there's a history of evolution and selection that put intention into that system so that it can repeat repeat sort of at least not if specific past patterns um problem representations of past patterns. So this is like you know like why are we good at generalizing to new things even if we haven't seen them in our past. It's because we have some abstract representation of the problem space that's been presented to us in our past that we can apply to the future. Um so so intention to me is actually a signature of the past um and past selection. And I I think this is also something that's quite hard because there's a lot of language that we use that's future directed like this thing predicts this thing has intention this thing has goals. And what we don't realize is those are statements actually of the history of that object because that that physical structure you're talking about was evolved and selected to have those properties. And the prediction is about the past being recorded and being able to be repeatable. It's not actually a statement of the future because the future hasn't happened yet. This is this is another sort of radicalism of assembly theory, but like I like I actually I'm a I'm a strict presentist uh which is uh philosophically and scientifically, but like as a person, it's really hard for me to like cope with this. But I totally agree with it as like a a scientific platform. the idea that only the present exists. But I think the present is thick in the sense that like we have these causally deep objects that have a lot of stored memory in them. So a lot of the past exists in the present moment as a deep causal structure, but the future hasn't been created yet. So it just it has no existence until it gets generated. And people think that they predict the future, but what they're doing is predicting recurring patterns from the past. Um so >> yeah, >> so cool. Sometimes I love your answers. That's great. >> There's so much so many >> different areas that I want to go into. There's there's so much there because it's Ilia Sutsker was recently saying that that's kind of like one thing that's missing from large language models >> is this idea of intention and he compared it to like human states or human emotions. He said maybe they need emotions like you know how sometimes you just move in a certain direction like people who had I don't know poor childhood they want to get richer and it's not that they have some goal it's just they have this I want to get to this state or they think in the future they'll get to that state so they work in this direction >> um and also the fact that like you're saying how we have certain words that are grounded in meaning like this is an apple this is a rock and then more and more layers of meaning that at some point is disconnected from these little pointers the other thing that kind of was So interesting for me to realize is LM have no actual connection to any re base reality there. They're all just words related to one another cuz they've never you know seen an apple. It's just how Apple relates to all the other words. Um >> yeah. You know what's really interesting cuz you know like I there's there was that famous example of like the Golden Gate Bridge where like the LLM started to build the association of it like and have like a representation of it and and so people started to think like oh yeah we're recognizing the neuron where the golden gate like bridges in the in the system but all of these things so so there is a interpretation right like these systems actually have a representation but again I don't know if the hardware and the computer of like the electrons moving around silicon trips has any awareness of the Golden Gate Bridge. Um, but I think a human interacting with that system would be able to say this is the Golden Gate Bridge. And that's sort of the level of the dichotomy of the question. And so when I see these systems, I don't think about how deep that neural network is in the computer. I think about how deep the human mind is that our our cognitive architecture is so freaking deep in terms of like the the abstractions and the connections we build that we can we can build an algorithm that imprints this huge pattern that human culture has built which is human language and encodes it into a tiny box like the these are like language before LLMs you know is massively distributed and then we have you know over all human minds and all books and now you have this dynamical registry where you can just access human language language and get like these these report outs of it and that architecture is so deep that like like we can see these things in it but it's actually like you're looking in your own mind and like the patterns in your own mind just now physically stored in a device in your environment and I think that is like that's more wild to me and I wish more people were talking about that like I I just think what humans are and what we create is like so causally powerful that like we are completely underestimating the architecture of our own minds as these like amazing explan explanatory machines that are capable of building these abstractions. Um >> yeah, if you're um since you're like kind of building a theory about what life is, but you're also alive and doing it. >> Does the work ever get like kind of surreal? Like you're trying to explain the very thing that you're doing and explaining. >> It's actually kind of funny. I was I was telling my lab, you know, because I have like grad students and posttos I work with and they're absolutely fabulous. But I I was I like I was trying to explain to them that I'm I'm a method scientist, you know, like there's method actors and they like, you know, they stay in character so they like really understand um like the charact like I I have a hard time shutting off what I like what I do for work versus what I am. um as a person um because I and it actually maybe it just comes with the process of being creative cuz like you know a lot of artists have to really embody themselves in their work and I I tend to think of like what I do professionally much more as a creative exercise than I do as like the way people would traditionally describe a scientist. So I have this very embodied experience about what I think the physics is of what life is and how I'm trying to describe it and I do have a very hard time shutting it off. So, it's like if I'm hanging out with my kids and stuff, it's like like that's probably like the only time that I'm like, "All right, I like I Yeah, but it's it's it's it's everpresent." Um, and I'm constantly trying to regularize every observation I have almost into like formalizing these ideas and trying to test them rigorously. And I sometimes also joke that the only reason I'm sane is because I'm a scientist because I can like I actually have to physically force myself to test things. I think I can't go into these like, you know, crazy loops of like building my own reality for myself. Um, because it has to be actually something that works in the real world. Um, but it is I don't know. I also find that part kind of fun. But yeah, >> and and to to that point, that's very interesting. So, I mean, if we're looking kind of like the assembly theory, uh, how you're viewing the world. I mean, is there anything number one that's like really surprising to people from that perspective? they would find very surprising. Is there anything that we can use that to predict the future a little bit? Anything that kind of emerges out of that that is fascinating to you? >> Well, I I mean I think actually it's interesting because is a scientific theory, but I think some of the most interesting ideas for me are like the new philosophies that emerge out of the theory >> and and one of the philosophical but it's also I think an empirical concept is the idea that time is a material property. Um and what I mean by that is is actually this idea of the assembly index measuring the causal depth. So if you measure the construction history, but you think it's a physical property of any object in your environment, then basically what I'm saying is the thing that distinguishes me um from a bacteria or me from a molecule or me from this microphone is actually how deep am I as an object in time, not just my physical dimensions as a spatial object. So you know, like I wouldn't you you wouldn't know looking at me that like I'm part of a 4 billion year lineage. you might just think, you know, like I'm 53 and, you know, like like you you see the physical size, but you don't see the size in time. Um, and so I think about all evolved objects as having actually a a a causal like a causal history that is a physical property. And there is like this entire possibility space that's a physical space just as much as the three coordinates of space and uh r or like uh the time like coordinate time on clocks which is actually more about simultaneity than it is about time in a progression of events kind of way which is a whole separate conversation about relativity and things like that. But um uh anyway, long story short, >> no, this is great. Let's let's let's unpack that a little bit because I'm just so so I mean obviously how complex uh something is, an organism, whatever, culture, technology, of course, time is a part of it. Are there other big things that kind of go into that? Is just like the longer the universe exists, we're going to expect to see much more complex things or is there something else that's driving that? >> Well, it's a totally separate thing. So, so I actually like I have to be careful. Again, it goes back to the use of language. So, I'm saying like time is a material property and I do really mean that, but like the the sense of time I'm talking about is a causal time and not necessarily a clock time. So, there the the fir the sort of first base level that like is important to understand is that every theory of physics that we've ever developed has its own sense of time and its own way that time gets regularized into the theory. just in the same way like energy has different like energy is kind of universal but it takes different manifestations in different kinds of physics. Um but so like Newton introduced the idea of like a clock time and like the universe basically moves through time but time is not like a property of the universe. So like you have a state and then at like at some time and then like the universe has a next state at the next time. And then what Einstein's innovation was was like actually to say that time is relative and so this clock will measure a different time than this clock if they're moving at different reference frames. Um ent like thermodynamics has another concept of time which is the entropic arrow of time which is to say that time has a directionality and that directionality is moving in in in terms of increasing entropy. So there's kind of different concepts of time and it's very unclear exactly how they mesh together for some of the theories not all of the theories. Um, and what I'm talking about is causal time, which is to say that ex like the order of existence that object like things can come to exist has a physical structure to it. That's what I mean by possibility space. It's like you can't be born before your your grandfather, right? Like we're all familiar with that. But actually, there's a hierarchy to existence um and an ordering to it. And so um you know like uh and so that hierarchy is actually what I'm talking about when I talk about causal time and each object exists in that hierarchical structure that that tells you the depth as the assembly index and um and so that tells you something about like the possibility space that actually had to be traversed to create that object and how far into the structure of things that could exist that object is placed. And then that that space we call it the assembly space is actually a physical space in the theory. And so just like my body can move you know like in an xyzcoordinate space physical objects can move in this assembly space in terms of like this object can be created because this object already exists or this object is now impossible because I went down some other path. And so um uh and it so that's very abstract and it's not like directly perceptible to our our physical experience except that we know what it is to have a creative act and to like actually create something that didn't exist before. And you know the our mental architecture has a map of that space just like it has a map of physical space. And um and I I I I think it's just as physically real. Um, and I think people have not appreciated this for most of human history, but people also didn't know that like space-time geometry was curved for most of history. Um, so I I don't think it's like it's a huge leap in the history of science to propose that physical spaces exist and they have measurable properties. Um, but where we stand in science because we think we know all of physics, people think it's a huge leap. >> Um, >> can I thank you for that? Yeah, I appreciate that. Oh, I'm curious. Could um so if we go all the way down to the perspective of like a single hydrogen molecule where it's >> I guess about as simple as as can be and you take one from the big bang and you have um whatever I think 13 billion years or something since then and it has a position that it's been in like it's been in different states which is more of a geometry >> like >> but it's not become more complex in any way does it? It doesn't have any memory of any of of that though. So >> it got no memory of where it's been. >> You you I mean you couldn't recover that. You you can say it had a history of different positions, but you can't recover that from the object, right? Like the hydrogen doesn't encode that. So >> you would need something more complex to actually look at it, >> right? And that's why things like us get really interesting and why I think we're not just emergent spatial patterns, which is the way that people would normally talk about life. Like one of the paradoxes is like of life is like life is an emergent property. It's not a part. It's not a feature of its parts. And people would use that to write off that life can't be fundamental because particles and molecules are fun. Like, you know, you go to molecules and then you go to atoms and you go to elementary particles. That's fundamental because I've removed all the spatial complexity. And but really what you've done is just remove all the memory and time. And what I'm saying is the memory and the the that causal structure is actually just as much a physical space. And that's the physical space that defines me. Not just looking at my where my atoms are in a spatial configuration, which is the way people would talk about it entropically. It actually is the configuration that's assembled over time that becomes the thing that we call life. And the fact that that entire structure is still, you know, like I'm a compactified 4 billionyear history in a 5-ft object. Um, you know, so it's like like there's a lot of like a lot of causal depth to make have the universe make something like me. And so you and I are capable of talking about the history of that hydrogen atom because we have enough recorded history to actually even encode in our neural architecture that such a thing is a possible statement about the world. But the hydrogen has no knowledge of that. Um and so if I was an alien like like you just you just can't recover that. So so it it is worth talking about what information is actually intrinsic to an individual object versus what we project on the object because we actually have that depth of information. And that's sort of the main distinction that I'm trying to make also with AI and other things is that like we do not appreciate that as far as we know we are the deepest structures the universe has created in terms of evolutionary causal time. Uh and therefore everything that we see in our environment is like you know encoded in this very deep neural architecture we have in our brain and in our body. Um and it's wild that we we can't see the universe outside that. >> Okay. So what questions about like life or the universe do you think are genuinely unanswerable? >> Um well I think in some sense that the like the future will not be fully predictable ever. >> Uh that's one and I think that's very counter to the way that people think about things because I think people think if you have higher enough resolution and enough data you'll be able to predict the future. And I think that the universe is intrinsically creative and as it as you create more possibilities the structure gets deeper. But there's also like the future horizon of what could be created is always bigger than the past horizon. And so the universe actually can't encode its own future. It doesn't have enough like computational capacity or memory to encode its own future. Um so that's one. Uh I think also with alien life uh you know because I think the possibility space is so large I think it would be very difficult for us to predict the exact instantiation of aliens until we actually encounter them. Um, and so, you know, there's a lot of conceptions that life on other planets is going to look like life on Earth. But I I think I think this space is so large that when you start to evolve new structures out of chemistry, the potential space of what they could look like is so different that I don't have any anticipation of another planet following the same evolutionary trajectory that ours has. So, I think a lot of the things that exist on this planet exist on this planet alone and will only ever exist here. Um, and that's a a really sort of also a radical reframing of the way that people think about things because I think we just think we're common. Um, and you know, like I mean even in physics we talk about like the fact there's there's there's a copy of you somewhere in the multiverse and I just don't believe that. I think like you only exist here. Uh, like this is this is it. Um, uh, but um, but it's interesting actually because historically people thought the universe was really small like both in size and in time. Um and so uh you know I really like uh this historian uh Thomas Moyahan's been writing about the history of how people think about uh like depth in space and depth in time and depth in possibility and you know makes this point about even a few hundred years ago like I forgot who the person was but like literally writing at their desk saying like you know in several hundred years I'll be sitting at the same desk writing this again because they thought that the history of the universe was so temporally shallow and like the spatial configurations would just repeat themselves on such a short time scale. Like the fact that humans have the recorded history we have now is the only reason that we can even conceive how big the universe actually is because we're recording more and more history as we go, which means we're seeing a much uh bigger and bigger glimpse of how big the universe actually could be in terms of the potentialities. And I think this is like one of the greatest shocks of human evolution and also why LLMs are the latest existential trauma in that kind of space because we're like, "Oh, well that like human language is actually like this incredible like deep structure and I just like and I could you couldn't even see what it looked like because it was just in our minds and in between us and now you're seeing this whole extra dimensionality of how big a structure that actually is um with large language models. So I think I think we have very limited perceptions. uh they're evolved on a planet over four billion years so that I can see what I can see. I can hear what I can hear. But the human mental architecture is is recording that and our societies are recording that and we're you know like and we're externalizing what's inside our mind and also like seeing more of the world. So it's like that that feedback process is um really incredible. So, I'm going to ask a question and I think we both we have we're sticklers for certain words and how what they mean exactly. So, you might not like the words that I use for these this question, but that's >> right. I won't be offended. I promise. >> I just broadly want to ask the question words. >> Some some of these are are are not going to be correct. I know it. But, you know, with this causal time, I mean, I missed that question, the that word that you used earlier. So it's almost like maybe we can think of it as like tears or sequences of events. So step one 2 3 4 5 right and maybe step one was bacteria, step five is humans. Uh and then as we look back we like just broadly speaking and if we look back we can kind of see this progression. It's more and more complex over time and of course we're assuming by step 10 it's going to be that much more complex. Um, and so I understand you're saying it's impossible to predict what step 10 is without playing it out in the real world and letting the creativity engine sort of like build it in real time. We can't fast forward that progress. >> Um, but is there do you think over time it's possible that some laws will emerge that allow us to approximate like okay by step 10 this is the technological and biological sort of complexity. This is kind of what is there anything that we can kind of uh uh you know look into the future. >> I think we can predict the near horizon of where we are >> when you when you're when you're deep in the space of pos because you have to like each step like in assembly theory like when you take an assembly step you're actually taking a step to construct a single object in a combinatorilally explosive space. So like there's an exponentially growing number of possibilities. Um and so um so when you take a step, you're taking a step into an exponential space and maybe realizing one thing out of that space. Um but it means that you can you can you can kind of predict the near future. But like the the ability to predict actually, you know, dramatically drops with how far in the sort of future of the kinds of objects that you want to predict is. So um and again this goes back to a resource thing like maybe if you had more resources and you actually understood the causal mechanisms that construct the space which is which is a challenging thing for us to develop theoretically. We worked it out really well for molecules and planetary atmospheres and minerals but like trying to do this generally is quite hard. um as you're you're you're it depends on um like if even if you know the mechanism uh you don't have enough resource to predict uh you know like the full like what the full space looks like in the future because it's too big and so >> so if I if I can predict the next step you know then I have a cominatorily explosive possibility space after that step and then after that step and after that step and So you're just going to run out of computational resource. Um it's just and the way the universe handles that is it actually stores it in memory and then it truncate you know like if like once you have something selected you actually you you kill off a whole bunch of things that are just never going to happen. Um because they're they they become increasingly impossible. I like Steuart Kaufman has this concept of the adjacent possible which is like you know like the possibility space that exists around like what's present in the biosphere. um which I really love. But I also think in assembly theory about the adjacent impossible that once you you have a path that gets deep enough in this space, there are just some things that like there there's no possibility for us to go back and like resurrect all of history. Although it's interesting as you get deeper, you can start resurrecting objects like mammoths and things like that because you're you're deep enough to actually encode memory to like bring some stuff back, but you have to have some kind of recorded memory. Um, >> is it is it kind of like a is it kind of like a light cone where like the higher the like the higher the object is on the assembly index, the more possible states the light cone can kind of capture or predict? >> Yeah, it it actually um so light cones are a great analogy because they're basically built on causal limits. I think relativity is actually like causally a flat theory which is like you only care about like photons and their causation but you miss the rest of the structure of like the fact that existence is actually layered and and so I actually am doing a lot of work on like trying to think about uh relativistic assembly theory which is really totally probably the most batshit thing I think about but I really like it. Um, but anyway, the the point um that I that I wanted to make is that that that is a good analogy because it's a causal space and when you think about like this future horizon, basically you're saying I'm here. What could be causally constructed from where I am? Um, and so it's like there there is like any any any configuration that you observe in the universe has a possibility space that exists local to it that tells you what can be created from it and what was the past history of its creation and that's actually what the assembly space encodes but it doesn't mean that you could predict exactly what would happen. Um, and that the future horizon of the predictability as I'm saying you know drops off because the space just gets so large. Is it limited in any way by stuff that could happen in evolution and stick around? Because we're probably not going to see some giant leap in abilities because it's usually like these kind of directional changes. We see some convergent evolution like similar proteins and eyes of like octopus and a human like things like that that seem to be like almost like there's not like there's certain like pathways that evolution takes that are more common. Does that change that in any way uh shape or form? No. It's the same question you asked earlier about like you know if you had the constraints it would make I forgot exactly the way you phrased it but I had the same visual in my head. I was like oh here's the example um uh like if you have the constraints you're going to reproduce the same system and so eyes are produced because you know there are certain physical constraints in the architecture and a certain sort of um template in the evolutionary history that like that is the thing that gets generated because there's only a certain kind of like number of ways to make an eye like a proton receptor. Um and so that's imprinted by like the the environment of the structure the the system evolves in and the environment includes everything from the laws of physics to what other organisms might be present right. Mhm. >> So, you know, the pathways can be limited, but the they're they're what I mean is like, you know, in the space of possibilities, these things have to be co-constructed. So, things can exist if there if the environment has the right memory and templating for that structure to be able to be created. >> Y >> um >> that makes sense. Thank you. Yeah. So different environments, completely different >> pathways. But yeah, >> it's an interesting feature of your cognitive architecture that you're like, you know, like like I always look for these kind of patterns when somebody asks the same question different ways. And I do this with like with my students. It's super funny because like I have, you know, all these like brilliant PhD students I work with. And it's so interesting because they'll pick like what seem like diametrically different questions um to work on scientifically, but it's really fascinating because like if you dig down into it, like they really just have like one core question they're trying to ask and they don't even like realize that they're actually just like statement on this one thing. >> Yeah. >> Um >> because I just have a bunch of polymaths I'm hanging out with over here. >> I can't wait to just use that with Wes, too. Like, oh, that's an interesting feature of your cognitive architecture. >> He's going to use it as a insult. >> Yeah. >> That's funny. I don't know. I'm always just curious about the way people's brains work because I think um it's it's interesting the way the it goes back to what we were saying before. It's like people communicate things, but you actually have to like to build a model of what they're communicating. you have to understand like what what are they actually so I'm just I I I I cannot shut off my my uh my cognitive architectures desire to find patterns >> and then to build a more unified abstract representation of those patterns. It's just the way that I work. >> Yeah, it could definitely be a positive that that's an interesting feature of your cognitive architecture. You know, maybe it's Yeah, maybe it's industrialism or entrepreneurship or pattern matching. One really interesting thing that I found recently is multiple people told me this that they are finding that LMS now are able to help them talk to people that have specific either personality disorders or some specific features that they're able to kind of like translate almost like oh they said this this is kind of what's happening. So it's yeah this is going to get more interesting to study I think. So, uh, yeah. >> Well, I think I think there that also goes the other way where I like I know a lot of people that are trying to use them to figure out how to communicate with like people regularly, right? >> Just like like what would the average person respond to this? Like how do I communicate to people in the way that I actually intend to? So, I I think they're excellent devices for that actually because they're like they're patterned on human language large. So, you know, if I I used to use social media for this, um, but now I'm finding I increasingly using LLMs, which maybe is not the best thing, but like what would be the consensus representation in other people's minds of this idea I have? Um, and so I used to do these and I I I I probably will go back to it at some point, but I just very tired of social media right now. But like like I purposfully like write my my um like posts on X in a certain way and then I would like read all the responses because it like allows me to see you know through this prism of social media like how does this one idea interact with other people's minds and I I'm very careful about how I linguistically represent it to try to see and I LLMs are a much flatter representation of that but they're a much quicker way to get like a quick inference. So, I don't think I get as much information out of them, but I think they they can serve a similar um utility. >> Absolutely. >> Yeah. >> Have you had like a an emotional experience with one like where you kind of felt like it like a poem or something written by an Ellen like actually touched you and you had to remember like no this is just this is not another person? >> No. Because I don't construct meaning from words. I construct meaning from having a model of the mind that's communicating them or a model of the social system. And for me, like there is no underlying model that has any conception of what the words mean for an LLM. So like I I just I I I I use them as a proxy for human language and like what would the structure of human language be based on the way that like you know the model is trained for like this thing I'm interacting with. I don't have I don't have any emotional model of an LLM. Like it it just I I just don't think there's anything there. Um >> what does it seem like so some of the people that are kind of thinking about long-term relationships with like avatars or using that as kind of >> I totally understand why. Um but I but but as I said be because of the way that I've trained my own architecture on interacting with the world because there's certain kinds of patterns I want to understand. I am not sensitive to the way that LM LLM's communicate, but other people that need emotional support in language in a certain way or um or you know want some kind of feedback in a ver like in a in linguistic representation and they actually find that to be the thing that's fulfilling for them then it it completely fills that need. Um but that that's not actually the way that I interact with people or the way that people communicate. Um, and I have actually purposefully um I because I I do science and because of the science I work on, I I'm much more hyperacute to that than most people are because I have literally my entire adult professional career had to get my mind to look not at the way that people represent symbolic representations, but to try to look at the meaning behind them to try to figure out like what is the actual theory that would describe life. And maybe I'm particularly attuned to this anyway, but I've never been really sensitive to like the meanings of words. Uh even like, you know, actually it came up kind of um early in my career uh because I was never really um sensitive as a woman in science. Like a lot of people would make this big deal like, oh, you're you're a tough theoretical physicist woman in science. And I never thought of myself that way. Um, and part of the reason is like I and I I I don't know if this is true, but like there's some social cues that I think other people are sensitive to that I just like I just completely like they just completely miss me like words people say like that are, you know, because I I just like they're just irrelevant. I don't like or they they're just not like they're not hitting my cognitive architecture in that emotional way. Um, but I can totally see why it could be useful for people. Um, but I think it it's just like people have to have caution with that obviously because just in the same sense you have to have caution with a therapist or you have to have caution with like an emotionally dependent partner. You have to be sure enough of yourself and your relationship to yourself to have a meaningful relationship with any other entity. And so I think the worry I have there is that it's just, you know, like if people are becoming codependent and they're not, you know, like the codependency is not actually building them to be dependent on themselves and actually have emotional support from themselves, that's bad. blanket bad whether it's an LLM or another human being doing and I think certain people have sensitivities to building those kind of relationships because they've had a hard history and that's just tough. >> So in your free time you're not watching like romcoms and the Bachelor and Love Island and stuff and >> Oh actually some of them I do you know sometimes but I yeah I you uh actually have a hard time watching uh television these days. I don't know why. Um but I do like some things. Um, >> I like Emily in Paris because I like fashion a lot, but like it's a, you know, it's like >> Yeah. So, there's like some things, but no, not The Bachelor. I mean, I'm a huge Mr. Beast fan. Um, >> that's cool. Yeah. 10 people in >> mostly because I think he's so intellectually like he's such a genius. Like, I just So, I actually I watch them not because of the the clickbait value that they have, but because I'm I'm trying to think about like him as a creative person that generated this that's engaging so many people. So it usually for me to find enjoyment and entertainment I have to think about it across multiple levels and I I always just want to know the mind behind the thing. >> Yeah, you can probably get a hold. Have you met him? I feel like you might know the kind of people to >> Yeah, I haven't met him but yeah. >> All right, >> probably at some point. I I hope so, but yeah, I think I think he's just super cool. >> Yeah, this was it's such a fascinating conversation. I wish we had so much more because I there's so much overlap that I think we have not just with the the science but just like the personal stuff that you're talking about not or whatever like how you perceive other people, how you perceive words that's so fascinating because a lot of these things that you're saying is like really resonate with me because I have kind of these analogous things that that I notice. Uh so I just thought that was absolutely uh brilliant. Um we want to be respectful for your time. How do you have a heart out right now? Should we wrap it up? How do you feel? >> Okay, for a little bit longer. So, if you feel like there's something you want to do to wrap it up, that sounds good. I don't have to >> dash off, but >> Thank you so much. I appreciate that. Yeah, this is absolutely phenomenal. So, um Dylan, do you have anything that we should >> Yeah. Well, I guess I was going to kind of look at because good science has ways to like disprove things like um if uh if assembly theory turns out to be wrong like in some kind of fundamental way like in what ways might it be wrong or disproved or lead to sort of >> Sure. So, um I I'm already mentioned Popper, but I'm a huge uh fan of the falsifiability criteria for scientific theories. Like I don't think you can prove a theory right. Um but you can certainly prove a theory wrong. Um, and actually I don't I personally don't think that any theory is a permanent solution to understanding reality. So I fully expect assembly theory would be replaced by something else if it becomes a framework that you know like actually stands up to the scrutiny that it's under now and becomes uh a framework that we use for understanding life and intelligence in the universe. It will be replaced by something else. Um, so, so I just want to give that caveat because people have this kind of conception of a theory of everything like we're going to get some grand unified final description of nature and I just think that's a artifact of 20th century thinking and not really a realistic way of thinking about the structure of theories or what they do as a human cultural system. Um, so that being said, assembly theory has a lot of empirical tests. The ones I'm most interested in are the robustness of this idea of a threshold in assembly index and copy number like how many of a repeated high assembly object you have. Um as a threshold that divides the non-living and living universe in some sense. Um, and so Lee's lab did this really fabulous set of experiments in a paper published in nature comms a few years ago. Um, where they took a whole bunch of living and non-living samples um, and they did this mass spec experiment measuring the molecular assembly index of molecules in these samples and showed that there was a threshold of 15 above which you only found molecules that were produced by a living system. Um, so there were there were no um no abiotic non-biological things could make molecules of assembly index 15 or higher. Um, and I think so that threshold is actually system size dependent. So, uh, Lee and I just wrote this philosophy paper recently and derived cosmological bounds. Like if the universe had no life in it, what would be the highest assembly index molecule you could possibly find? And it's probably between 40 and 60. It kind of depends on how you parameterize your model, but like I'm, you know, probably about 58 or so, but um but like you could actually develop theory to precisely pin down this number. And like basically anything above that which is you know any DNA molecule or any protein is above that um is something that the universe you know like if you find it in the universe is a surefire sign of life because using the entire resources of the universe with no selection and no evolution you would never produce something um of that assembly. So, so basically you have a range of 15 to say 60 um that you can you can actually make predictions and I think we could predict like for meteorites what would be the highest complexity you would expect um for minerals what would be it we we have some stuff that we're doing for planetary atmospheres now where we've applied assembly theory to planetary atmospheres and it's actually like really compelling because you don't have to put any knowledge of any planets in and if you take all of the known planets with atmospheres um and their properties and you you put it's an assembly theory framework you can see that earth's modern biosphere is the only one that has high assembly and high potential possibility space. So I guess what I I I want to see is more ways of testing this conjecture of this threshold and how robust it is in different physical systems. So in a molecule, an individual molecule can be the signature of evolution. An entire planetary atmosphere can be the signature of evolution with this theory. A mineral could be a signature of technology because it's precisely designed um you know like a silicon chip. Like if I looked at a silicon chip versus like a rock, how do I know the difference? Assembly theory can tell you why a silicon chip is a deeper constructed object than the rock. um if I look at a large language model. So I think what what we're trying to do right now is basically say we have a theory of physics developed in molecules that tells us that there is a firm experimentally testable threshold above which uh we have definitive evidence of life and I want to test it in as many systems as possible. I could have said that as the short version before I went into all the detail but there you have it. That was a short version. >> That was great. Yeah. So, I mean, good. >> You sure? Okay. I feel like I've definitely monopolized the questions, but I guess Okay. What? No, you go. I'll make sure you go next. But the I guess I was saying this theory is like pretty ambitious and if it um >> does kind of hold up the test of time maybe for the next decade or so and you feel like enough of these tests have kind of been done and it still is like resilient. Um do you do you feel like you might be in for a Nobel Prize? >> Well, that's such a funny question. Um, I you know I I I would be more excited if I could raise enough money to do the experiments to make life in the lab. So if I can do that in 10 years, I'll be happy. A Nobel Prize is a a nice thing to get. But I I don't it it's it's sort of >> yeah, >> it's an after the fact thing. And also the other thing about the Nobel Prize that's kind of hard is like most people when they get the Nobel P prize like their career is kind of killed in some sense because um because they get asked to do all these like public engagement events and things like that and you'll you'll see like most Nobel Prize winners like a few of them do really great work after the Nobel Prize but a lot of them end up not doing you know like and I ju I just I don't know uh you know I think I think I think the work that Lee and I are doing you know always always our goal is to have it be of that caliber but But I think you know for for neither of us is that the goal and if that was the goal we wouldn't be really doing science. >> Yeah. Um so I think it's really clear to make that distinction because so many people in science are actually like there there is like there is a career path in science right where you check all the right boxes uh you can get in the nationalmies and you can get a Nobel Prize and this stuff and if you have the right peer network and you're you're working on the right ideas and I think both of us are just so fundamentally interested in how the world works that there is no other question and so the yeah so um so I'm very uh flattered that you would ask that, but like I I just want to really make clear about um you know what is it about because like what we're doing is hard and it's under a ton of scrutiny and very critical >> and I know it's like a very tough question to ask somebody like how how great are you you know like are you that great like tell me >> right now it doesn't feel so great you know like I mean it's you know the the flip side of it is when you propose new ideas in science um you know science is like any other part of human culture people have um they have confirmation bias and like certain sets of beliefs and if you challenge them, you know, the community has a reaction to that. And so, you know, like the one of the things that's really hard right now is assembly theory is very different as a conceptual framework and also as a theoretical and empirical science than what we have in the standard literature. And so there are a lot of people that want to write it off immediately as wrong and not even give us the chance to like actually like like they won't read the science or they like they they you know they they'll adopt someone else's critical opinion um just on on hearsay right so so there's a lot of um ad honan and like you know attacks on assembly theory as being like obviously wrong like this is the statement people make it's obviously wrong um and they don't even look at the empirical ical data or what the structure of the theory is. It just has to be wrong because it doesn't conform to my beliefs about how the world works. >> Yeah. My science >> my gut instinct is that like if what it if thinking about something as an assembly index turns out to be how somebody wants to confirm detection of any kind of extraterrestrial like bio signature or something then all of a sudden it seems like it would be >> kind of in in reach. >> Yeah. I mean, I'm I'm hoping that we'll have observational research programs for exoplanets soon based on the atmospheric assembly stuff, but we still have to get like the first paper out on that. So, um like we're not even at the observing point. We just have done the theoretical work and I think it's incredibly promising. But I think I I think that's one of the reasons that my answer to you was like actually one of the things we're doing is critically testing this theory in as many ways as we can. And um I I don't like this idea that I work on assembly theory and I have some like soap box I'm standing on like the thing that I have done my entire career is I want to understand what life is and I started you know as a posttock and even as a grad student wanting to build a theoretical physics that would help us solve the origin of life like that is my career goal and assembly theory has emerged for me out of that desire and the the fact that like that is my only career goal. And so, and Lee was the only person that I think was as crazy as me to care more about the problem than anything else. Like any anything that we were taught or thought before. Um, and so, um, so assembly theory is for me the current incarnation of that is like is basically building in every idea I've ever had about like what I thought and like all of these colleagues, we have amazing set of colleagues working on it. Um, and there is a lot of really rigorous intellectual work behind it. But if it's proven wrong, we're gonna move on to the next thing. Um, and so I'm investing a lot of time in right now because I think it's so promising and I really want it to undergo the scrutiny that I think a promising theory deserves. But, you know, it's going to take us time to devel like we have to develop all these new mathematical formalisms, all this theoretical infrastructure, all these empirical tests which require new ways of like actually accessing and and interrogating experimental data. I mean, it's just like it's so many moving parts. um that you know I hope in 10 years that we can prove the things that we say but research is slow slower than the ideas you have so it's always a a long haul um yeah >> yeah this is yeah it's such an interesting and it seems like it's really developing it's as far as I know it's a little bit more recent one of the things that I heard recently so there was an interview with Bla1 Agua Earcus who's working over there at Google are you aware of him >> yep I've met him yeah >> oh yeah oh abs that was an eye openener for me the the BFF experiment. I was like, wait, >> you know what I mean? And it's it seems to really tie in to what you're talking about and just Yeah. I >> Yeah, those kind of things like I have a colleague Michael Lockman that wrote some code like I mean like like 20 years ago, so like they couldn't do it at the scale that Blaze and his team did now, but Michael had this really great thing that like had emergent self-replicating programs in it. So I think like we've known about those kind of things for a while and I really like Blaz's experiments. I think they're super cool. But the issue I have is like there's already so much historical evolutionary information in the algorithm that I I don't really see it as a I see see it as a good window into the origin of life because really what we're talking about is like you go from no information to a lineage that emerges and has memory and and and like and the causal ability to construct itself. And I don't think I think algorithms are already very deep evolutionary objects because they were created by humans and they're embedded in computer architectures that already have to be designed that I think it's it's really hard for me to think about the origin of life in a simulation like that. >> Yeah. And just one final qu sorry no just one point on that though I think the origin of life is a continuous process also. So it can happen in new substrates. So I do think like what Blaze is doing is giving us some insight, right? But it's like you have you just have to understand the evolutionary scaffold and then you open up a new space and you see the origin life happen again in that new space. So I just want to give that caveat. But yeah um >> and uh so just when we're thinking about this like complexity the um um how would you rate like the human brain versus let's say a super intelligence like does one have a higher lower assembly index? I mean, that seems pretty obvious, but you would like, does that make sense? >> I don't even know what a super intelligence is. People throw this word around. I like I I just I literally don't know. It has like no definition for me. What is it? More intelligent than a human on what what scale. Um, so I it it's I I would rather just try to ask first principles questions about what intelligence is before I want to conjecture about some supreme intelligence that surpasses humans. And already you know I I like I like um you know David Deutsch's perspective on AGI which is like like by virtue of the fact that we invented the theory of universal computation like we're already like humans can do any computational operation because we have a model of universal computation. So anything that any of these super intelligences can do we can do. It's just a matter of the time scale to do it. So LLMs are really good at some reasoning tasks and can do them super fast and we're good at some and can do it way faster than LLM and like and so it just depends about the accessibility of the information. So um so anytime that you take this abstract idea of intelligence, super intelligence, universal computation and you embed it in an actual physical system, it's going to have some limitations on it. And the idea of super intelligence and AGIS are discussed about now act as if the physical universe doesn't exist and there's like no upper limit on intelligence. on any given physical system and I just don't think that's a realistic possibility. >> Yeah, the phenomenal. Thank you so much. You've been absolutely great. >> And if you know what, if you could explain stuff to us in a way that we can understand it, I'm sure everybody everybody else will understand it. That's that's the one thing we can guarantee here. Uh but yeah, thank you. >> I hope I understand it. I'm not like >> I'm not sure if I understand anything I say. For whatever it's worth, somehow a lot of the stuff feels intuitively right. I don't know if that makes sense, but a lot of the stuff >> Yeah. No. Yeah. Sure. I do I know exactly what you mean because most of like the way that I I I I think about things is based on intuition. So, >> yeah. Yeah. >> I think intuition is underrated. Yeah. Sorry, Dylan. I didn't mean >> that's Yeah, intuition. Yeah. It kind of gives like when I listen to you too, I get sort of this sense like of a north star. Even though I can never repeat anything I I've heard from your podcast, I'm like, "Okay, I think I learned something and the world feels different." >> Yeah, it's actually very um a lot of my perception on language comes from like obviously talking to other people and understanding how they interact with the way I I think. But I I the the thing that's kind of funny there is I think I I explained to you guys about how I try to read the meaning below the words that people use. And I think that's also how I try to communicate. Although I really don't know what that looks like in somebody else's mind, but it's like a like for me it's very like there's a certain there's a certain representation I have in my mind, a very emotional intuition thing and like it's not very easily to encode in human language. So I just basically try to communicate that directly but I have to use words to do it and I don't even know how I build those layers in there. >> Yeah. >> Um >> yeah, I struggle with that as well quite a bit and I feel like a lot of people don't even think about that. So it's kind of like this almost like speaking a different language when I try to what what do you mean by that word? Yeah. >> Yeah. And then there's also all kinds of things that like I think written language is a fundamentally different language than spoken language and but we think they're the same because we use the the sort of same words and same symbolic representation but I don't know about you guys like I get a text from somebody and it's like >> like they all read angry to me like if I'm like you know but if I talk to the person it's like very clear they're not angry. So, it's it's kind of like it's like I I like text messages are really hard for me because I don't have the emotional like input of like the modulation of the voice. So, I have a really hard time uh reading text messages them. But, >> no, I definitely Yeah, text and emails. I feel like they slowed down the world somehow. They're like tough for me, too. >> Yeah. >> But it's good. And you made me a presentist. I think them now like seeing everything that I think is coming up in the future is just some pattern from the past I'm projecting. So, >> yeah. >> Very good. >> Totally. But there is real novelty, too. So, that's also exciting. >> Thank you so much for for for sticking with us so long. And uh we'll see you in the next episode.