History of Science and Technology Q&A (May 10, 2023)

Hi, everybody. Welcome to another episode of Q and A about history of science and technology. And I see that there are all kinds of interesting questions saved up here. Let's see. Okay. IG11 asks, when researching, do you find it's more helpful to stay close to modern times in terms of content, or do findings from hundreds of years ago also prove valuable? You know, it's an interesting question. In different fields, there are different amounts of look back, you know, I have to say, like a field like philosophy, you know, the things that one would debate today often are the same things that, you know, would have been debated by Plato 2000 years ago. It's a field where many of the core questions one's still turning over the same kinds of questions today. In a field like mathematics, a theorem proved by Gauss is still in the early 1800s, for example, would still be a theorem one wants to use today. Although the questions one might ask might be built on things that were from the last 10 years or 20 years. Let's say different fields work differently. In medicine, for example, if you write a paper and you've got references to papers 100 years ago, people will look kind of like, oh no, that can't possibly. You know, how can it be relevant? You know, it's, in a sense, it's almost a badge of honor that the references are to papers that are incredibly recent, like the last year or so. Whereas in something like physics, I would say that there's a certain stylishness to being able to have a reference to something that was 50 or 100 years old, something 2, 300 years old. In physics, it would be like, no, that was a previous paradigm. I think one of the things that happens is there are fields where there are paradigm boundaries, like there are major methodologies that get introduced and the things that were part of the field from before that major methodology shift might not be so relevant. Not least because the ways of talking about things and the questions that are asked have shifted as a result of that change of paradigm. An interesting area would be something like AI. Is it the case that if you look at a modern AI paper about LLMs, you will find a distribution of references from ones that are about things that came out two weeks ago to things referring back 50 years or more. That's an interesting case. Well, I think for my own personal activities, I tend to be much more interested in things where they are, in a sense, they have been slow moving and where I think I might be able to make a contribution and the kinds of the things that I'M building on might be 50 or 100 years old, rather than, oh, I'm building on the thing that was just done last week. That's a personal preference. A lot of the science that gets done, for example, in the academic world is very much building on things that sort of just happened. Quotes last week. Although, as I say, different fields have these different kind of time horizons that they typically adopt. And I think in terms of some of the things that. Or things I've been interested in about fundamental theory of physics or things about metamathematics and so on. It so happens that both of those areas, the big bursts of paradigmatic change and activity happened about 100 years ago. So for physics, they happen between about, well, the very end of the 1800s and the 1920s, 1930s, the creation of quantum mechanics, discovery of general relativity, the kind of the general mathematicization of studies of physics that had begun in the 1600s, but really blossomed with things like Maxwell's equations and so on in the mid to late 1800s. And I think the. So, but then that was sort of a flowering of that mathematical paradigm that happened in that period that extended up to the 1920s, 1930s and then after that. Well, there's certainly lots of great work that's been done, but it's a little, it's less paradigm changing. And I think what we're doing now with our physics project, for example, is another kind of paradigm changing moment where we're building directly on sort of the last paradigm changing moment, which was in this period about 100 years ago in metamathematics. The work that I've done there, in kind of trying to understand sort of the big picture of metamathematics. What I'm calling the physicalization of metamathematics, again builds on things that happened to occur around 1900 to 1920. I don't think there was any. I don't know whether there was a correlation between what was happening in physics back in those days and what was happening in mathematics. I mean, to be clear, what happened in metamathematics was this realization that there was an abstraction of mathematics itself, that mathematics, well, talks about abstract things, but you could sort of lift above mathematics to say what was the meta structure of mathematics. It's about theorems and proofs, even saying that without talking about what the content of those theorems and proofs were. So in a sense that was a sort of a lifting up beyond the, the, the actual content of mathematics to something above that. And I'm not quite sure whether that was, I don't think that was quite what was happening in physics at that time. So I'm not sure that those two developments were that correlated, but it so happens that they were both occurring at about the same time. And in my own efforts in recent times, I've been sort of building on, on both of those things from 100 years ago or so. Now I would say that the things, you know, I'm looking at essentially the science of LLMs. What's that building on? Well, it's building in part on incredibly recent stuff. The, the, you know, the discoveries made with what ChatGPT is and what it makes possible. Things that are sort of, you know, week by week, people discovering things about scaling laws for neural nets and things like this. That's very recent stuff. But then also some of that science goes all the way back to antiquity and kinds of things that people like Aristotle were thinking about, about what can one say about the overall structure of language and meaning and things like this. So at least for me, there's some rather wonderful cases where one has both things that are very much of the moment and sort of discoveries being made now and sort of reflecting back to things that date from way before modern science. And I would say that some of the things that I also find interesting, some of the things that have been revealed by our physics project and so on about sort of fundamental questions about how the universe works, some of those questions haven't really been talked about during the modern period of science. So questions like why does the universe exist? That's something where there is work done at a sort of philosophical and theological level, back from, you know, a thousand years to more than that and maybe 500 years ago. But it's something that is, those are a type of question largely untouched by modern science, which we start to have things to say about. And I suppose that one of the things that can happen is a given paradigm will have certain things it can say. There may be questions that were marooned, were left behind by that paradigm, unaddressable by that paradigm, where were marooned for 100 years, 200 years, 500 years. And then when you come to look at them now, you're referring back to those questions, even though they kind of miss the modern period of whatever paradigm is being developed. So I suppose that's a case where you can expect very long time sort of things to re emerge. But I think this whole issue of kind of the chain of kind of intellectual development in science is really an interesting one. And I would say that one of the Perhaps less than wonderful features of the machinery of the sort of rather industrial machinery of many areas of science and the way they're practiced in academia is this fact that it really is very incremental in terms of. That there are a lot of kind of forces that, that push towards having sort of incremental progress. You know, you can, you can get that public paper published in the journal if it's really similar to papers that have already been published, and it's just adding one more little extra step, so to speak. Whereas if you're trying to go back to a thing which was, you know, really sort of paradigm changing and reflects back to earlier paradigms and so on, that's something that's much more difficult for the industrial system to assess. And it's much less. The system is much less set up to deal with something like that. Okay, there's a question here. Boy, all sorts of questions. Okay, one from Aaron. You talk about the history of theories of cognition and consciousness. What did the ancients think? What did Godel and Turing think about this? Does Chatgpt disprove Penrose's theories about. About quantum mechanics and so on? Okay, well, I'm not sure I've ever. I've ever thought about this history systematically before, so. But let me see what I can do. This whole question about the idea of consciousness and the idea that there was some sort of spark of something special going on in brains that didn't happen elsewhere was certainly a thing that existed in antiquity. There was the idea of there being some kind of soul, some kind of beyond the mere mechanics of how things work, something that was more abstract, more. And a seat of kind of this process of thinking. But there were some very simplistic theories about it. I mean, for example, there was the idea that there were kind of a pretty good idea, due to the Epicureans and Democritus and people like this, that the world is just made of a limited number of kinds of atoms. And they're atoms of different shapes. They might correspond to the Platonic solids, et cetera, et cetera, et cetera, just a small number of them. And one of them were these fire atoms, which I think were supposed to be long and thin and so on, and that were supposed to be the types of atoms that represented heat. But. But I think they also had the idea that they were a type of atom representing souls, that they were a type of atom that represented that. That was kind of the nature of the mind somehow was associated with this kind of different type of atom to be Distinguished from the atoms that would correspond to water and things like this. And I think that. So they had this idea that the kind of. There was this sort of substance that was the mind like substance. And it was something like. And that was kind of the seat of mind, so to speak, that minds had a substance just like heat was thought to have a substance back at that time. Well then I think some of what this segued into was a lot of kind of theological, religious thinking about minds and kind of the nature of minds and kind of where did minds come from and where did minds go when the actual human or organism died, what happened to that mind like thing? And I think the thing that was realized pretty early was that there was some aspect of a mind that was some kind of sort of abstractly represented thing, the sort of the eternal soul, the abstraction of the mind independent of the physicality of brains and so on. And I think that was something that really was. Was quite recognized that there was something about minds that was different from the. The actual mechanics of how brains work. But how that really works was not something that became clear until modern times. And the idea of computation and all those kinds of things really didn't become clear until the last few decades, I would say. But back in those days there was this kind of notion of there's this soul like thing. How does that soul relate to the physicality of things? And so on. Another issue about souls that loomed quite big was the question of free will and the question of whether somehow minds there were. Was kind of an idea that in nature there was some regularity, there were laws of nature in some way. But then there was kind of something where laws of nature where things happen in the way they have to happen in nature. But then the question was whether minds could do things where they just came up with that idea. And it didn't sort of mechanically happen the way it has to happen in nature. Now one confusing thing about antiquity is the view of fate in antiquity. The view of, of can we know what will happen? No, we can't. It's all just fate. And the idea that this kind of notion that there is a clockwork universe, a mechanical universe, that was an idea that didn't really emerge in a serious way, I would say, until the 1500 1600s kind of time frame. The idea that really there was one didn't just have to, oh, there's fate and it can just come and do whatever it wants. There are actual definite laws of the universe. But I think that this notion of could there be A sort of free will, a. The mind can do anything kind of notion. That was an important notion back even antiquity. And I think, I don't know this history so well, but I believe one of the innovations of Christianity was the idea that there could be a kind of a free will, an ability to choose good or evil or whatever else that was independent of some machinery provided by the world. That that was something that was sort of an independent thing of minds and human minds in particular. That was something sort of separated from the mere mechanics of the universe. I think the kinds of things that. Well, this thinking, I'm probably forgetting a lot of philosophical and theological work on, on kind of the nature of the mind and so on. But, but this whole question of free will which became very entangled with all kinds of questions about, you know, good, evil, you know, how does God make good in the world versus how do humans choose to do bad things, etc. Etc. Etc. A lot of discussion of that that went on from, oh, basically the. Up until, well, certainly. Well, for a long time, but I mean, until certainly hitting the middle of the previous millennium, so to speak. So, you know, 1500, 1600s, things like this. But then there were folks like Rene Descartes in the late 1500s, early 1600s, who were starting to kind of think in a somewhat more enlightened, one might say way about kind of the nature of the mind and those kinds of things. And Descartes famously was, you know, how do we even know, you know, how do we know that there's anything out there? All we know is what's going on inside our minds. The idea that there is an actual universe out there as opposed to just sort of. It's all in the mind and it's, and it's just an internal story. It's all just a matrix like video game like thing where we are. Where the sensory. Where we're just a brain in a box or a mind in a box and sensory data is being fed to that and there isn't an outside to the whole thing and the whole sort of how do we know we exist? Well, you know, I think, therefore I am type, type, idea and so on. But this notion of whether there was an outside, whether there were other. Whether we were just one mind in a box and there's nothing outside and it's all just for our benefit was. I mean, that's still an interesting question. I think it is. The main observation is that we know what's internally happening in our minds. We can see other people and we can see that something comparable to what we know is happening in our mind seems to be happening in that we. Their external actions make it seem like what they're doing is something similar to what's happening in our minds. Now it's going to get a little bit weirder because as we get better sensor information about what's happening in brains, as I'm sitting here yakking at you, it could very well be the case that at some point in the future I'll have some display that's showing, okay, this part of my brain is lighting up and this is happening. And these are all the flows of, of thinking that are going on in my brain. And that's what's happening inside, so to speak. And when I'm seeing somebody else, I could see the same kind of thing. And it will, it will feel this idea that there's a very different thing inside than things outside. I think that's going to decay because we're going to see more and more. We're going to be able to externalize what's happening in our minds, so to speak. But any case, the, the thing that, this question of kind of what's going on in the mind, and I think Descartes actually made the statement that one day we will be able to understand what happens in a tree just as we understand what happens in a clock. In other words, he had the idea that in the end everything will be sort of mechanically describable. And that was a tradition that kind of continued of, okay, science is starting to work. We're starting to be able to have these kind of formal descriptions of the world. We're going to get everything formalized, including how minds work and so on. Well, it didn't quite work out that way. I think what happened then we're going into the late 1800s and things like that and. Well, in the 1800s, there had been the beginning of experimental psychology. People like Helmholtz had worked on that, Fechner, some other people forgetting some of the more famous names, William James, another person. But there was this idea that there was a science of psychology, a science of thinking. You could have, oh, things like the, you know, as you press, as you, as you give a higher stimulus, it's sort of a logarithmic effect in certain cases, things like that. The idea of kind of noticing optical illusions, the idea of seeing how minds work. And that turned into people like Freud and so on, also trying to kind of make a science of the mind. And I suppose that was sort of a branch of this whole kind of idea of what became cognitive science, psychology, experimental psychology and so on. A science of the mind. And by 1900 or so, people like Ramani Kajal had done microscopy on brains. And it was kind of known that there was this kind of wire like mess inside brains. And by that time electrical machines were starting to exist. And it's like it's all wires, it's got electricity inside it. We can kind of see what's happening at some level. And that, that became sort of a, a, a general motif of it's all kind of a, an electrical machine. And that then segued, I suppose, into things like the cybernetics movement where people were like, it's, it's all, it's all electrical machinery in brains and that's what's going to determine what's happening. There's still was a certain strand, I would say, the, the electrical machinery part of things, which was sort of the cybernetics world. There was then a philosophical discussion. There was then also a religious part which was the kind of, look, a soul is something very special and you've got all these wires and so on, but you're never going to get to a soul with a bunch of wires. And that's sort of a piece of kind of that doctrine was there's something very, very special about souls. And in a sense it's true. And in a sense we now understand scientifically a little bit more about why that's true. But I think the thing that then emerged, this is probably by the 1950s, there was sort of a general belief that there was kind of like, well, brains were kind of a little bit like emerging computers. And there was this kind of philosophical theological belief that there was something different about minds and about inanimate matter and so on. But it was all a bit vague, I would say, that going into that. Well then there were things like, well, the question asked about Alan Turing and Kurt Godel. Let's talk about those folk. So one of the things that had happened in mathematics was the idea that mathematics could become mechanical mathematics, which had been viewed as an activity of the human mind, something which was almost an artistic kind of creation activity by the time of people like Hilbert, around 1900, David Hilbert and so on, really trying to sort of mechanize mathematics and say, well, no, actually mathematics is just this very mechanical process of going from these axioms and these theorems and deriving other ones, and eventually they'll sort of just be a machine that does it. There's a famous quote discussion between Henri Poincare A kind of much more, I would say, aesthetically oriented mathematician and David Hilbert a much more kind of machinery oriented mathematician. Poincare I think said about Hilbert. It said, there's a machine in Chicago where pigs go in at one end. It's a rather horrible thought, but pigs go in at one end and you know, hot dogs or hamburgers or something come out at the other. It's a giant industrial machine. And that's what Hilbert is doing to mathematics, that it's just like it's going to be a machine where axioms go in at one end and all the theorems of mathematics come out at the other end. There's no room for a mind, there's no room for something beyond the mere mechanical. Godel didn't believe that Godel was in some ways a sort of, I would say a very much more soul, humanity oriented kind of thinker who really did think there was something sort of in the mind that was beyond the merely mechanical. And what when he proved Godel's theorem in 1931, one of his goals was to blow up this whole idea that mathematics was a mechanically accessible thing. And his idea was something where there is a kind of, there is something about minds that is somehow doing more than the mere mechanical. Now one thing that I think Godel also thought was in the Plato tradition of there is a thing that is mathematics and mathematicians are merely discovering pieces of the thing, of the actual sort of reality that is out there, that is mathematics. Very different from Hilbert who said we can just construct mathematics. We can put in, we can talk about tables and chairs, we can talk about rings and fields, it doesn't matter. We are constructing mathematics as a purely synthetic activity from its kind of axiomatic basis. Godel thought mathematics was really a thing that already existed in some Platonic sense and we as minds and so on are merely exploring pieces of it. And the sort of, the role of the mind somehow is to decide what to explore. And for example, even when he wrote up Godel's theorem, he has some statement where he says, well, you know, it could be that this mechanical way of doing mathematics can only get to certain things and minds might go much further than that. He has a weird quote about ramified theory of types into the transfinite as a kind of what might minds be beyond the mere Hilbertian kind of machinery. Well, so that's what Godel thought. And I think in modern times, in my own physicalization of metamathematics, what we've sort of realized is that, yes, that's really true. There is a real thing out there. It is this ruliad object, the entangled limit of all possible computations. And when we do mathematics, Godel was right. We're just sampling as humans, in our human way, we're sampling little slices of that. And that ruliad object, which is the underlying thing of metamathematics, is the same thing that is the underlying thing of the physical universe. And insofar as we claim the physical universe exists, we also must claim that the underlying thing of mathematics exists, and we are merely seeing parts of it. But let's come to Turing. So Turing had originally, in his conception of a Turing machine, he wanted to find. He was really trying to find an idealization of what mathematics is, in the same way that Hilbert had done and Whitehead and Russell had done and so on. But Turing wanted to make it more mechanical. He wanted to really nail that down. And his original model was to think about how bank clerks take things out of files and write numbers down and erase other numbers. Of course, they don't do that anymore because they're using computers now. But back in the day, that's what they would do with pieces of paper and pencils and all that kind of thing. And so that was his kind of model of what computation might be. And he idealized that in 1936 as his Turing machine idea. And then I think he didn't really. He just thought of that as a model for kind of how mathematical machinery works. The idea that that might be a model that went beyond that. It might be a model of minds, it might be a model even of physics. He didn't get to the. It might be a model of physics. He thought that Turing machines were, at best, something that was a model of mental processes. The idea that things like Turing machines and the computation as defined by folks like Turing might be something that is a matter of relevance to the natural world, to physics. I think probably I might be the person who pushed that idea more seriously than anybody else in the 1980s, actually, because for a long time, people have believed, well, there's computation that involves discrete things and all that kind of thing, and then there's physics, and it's all based on the differential equations that have been invented in the 1600s. And physics is not something that can be talked about discreetly in that way. And that was something. In the 1980s, I kind of pushed the idea that these notions of computation are very useful for understanding the physical world, even though at that time I didn't know and didn't. Didn't have a reason to believe that physics would turn out to be computational really all the way down, but at least as a way of describing, as a way of modeling things. In natural science, computation was clearly useful alongside the equations and things like this. What we now believe as a result of our physics project is that, yes, the universe actually is computational all the way down. So that really is much more set on bedrock than it was at that time. But so Turing didn't get that far. He didn't think that his Turing machines were relevant to physics by 1950 or so. He did think they were relevant to minds. And he did start talking about kind of this idea, the Turing Test and all that kind of thing came from that period of time when he'd gotten to the point of saying this mechanical construction thing seems to be capturing what's really happening in mines. Now, if we go further than that and we ask kind of how did that evolve? How did this idea that there could be a sort of a mind like thing that was mechanically producible AI, that term arose in the mid-1950s with Marvin Minsky, John McCarthy, these folk in the MIT kind of computer ish orbit, so to speak, that was where that came from. And it was sort of also very much tied in with the cybernetics movement that had come out of Norwalk. Weiner, also an MIT person from, from the 1930s and 1940s and so on. But that was a tradition, that sort of AI and its relationship to consciousness, cognition, these kinds of things. They were not that seriously connected. I mean, there was AI and AI went through periods like in the 1960s, early 1960s, where it's like, well, actually the 1950s, people were like, computers are going to make artificial brains. It's just a done deal. Then by the 1960s it was like, well, we're going to do these great projects like we're going to invent automated language translation. We're going to have machines that can go around, do planning and walk around like robots and so on. And they kept on being these, it's going to work, it's going to work. And it didn't work. And so by the 1970s there was kind of a. Well, there was sort of a no, AI isn't going to work. Then the 1980s were expert systems and so on. There were a series of waves of this. But I think the idea that anybody would be serious about saying, we've nailed down how brains work and it's just like these AIs we've got. Nobody really took that, that Seriously, I think what has been progressively shocking, so to speak, is that a lot of things where people said, as soon as we can get a machine to do X, then it's going to be a true AI. And one of the last really big boundaries is when we can get a machine to kind of converse in natural language. That was kind of the Turing test idea from 1950. Then we'll know we've really got AI. And that's what's happened this year. And I think that what we learn from what's happened there is, well, it is just a neural net, and it's doing a pretty good job of generating, you know, human, like, natural language. And that means when we look at brains and we say, what's really going on in brains? Is it weird and quantum and molecular and so on, or is it just a bunch of wires basically with a bunch of neurons? There's a lot of evidence now it's just a bunch of wires with a bunch of neurons. And so, for example, people like Roger Penrose, who is, I think, very much in this kind of almost. I wouldn't say it wouldn't be fair to call it mystical tradition, but very much in the girdle kind of. There's something more going on than the mere mechanical and physical has been interested, I think, in looking for kind of ways that brains go beyond the mere mechanical kind of clunking along of neural nets and looking for that in terms of quantum gravity and things like this, and looking for the possibility that there is some effect beyond the physics we know, beyond the mechanisms we know that can contribute something special to brains. And, yes, I think that now there's a lot of evidence against that from the success of ChatGPT and so on. Now, I will say that the. What is it? Objective reduction? Is that the term? There's this question that arises in quantum mechanics of sort of the big deal of quantum mechanics. Well, in our models of physics. Now, I think I'm finally getting to the point where I can say I really understand quantum mechanics. It's all about this fact that there are many possible histories for the universe that branch. They merge. This is this kind of idea. When we look at a slice across all these histories, we get this kind of branchial space of possible branches of history. And we. When we're observing the universe, we. Our minds are, in a sense, also branching and merging, and we are extended in branchial space. So this question about sort of objective reduction is the question of whether somehow there is a necessary channeling of all these different branches of history crushed down into one branch. Because we believe that definite things happen in the universe. And I think we believe that because we are actually spanning many branches of history and we just in the way that we think things happen in the world, are conflating all these different branches of history. But there's a different possibility, which is what I think Roger Penrose, not sure that he knows quite this is what he's doing, and I've been meaning to point this out to him. But the thing that is really quite interesting is in physical space we're used to the idea of black holes where there's a big lump of mass and energy and that causes a great enough gravity that we get a black hole. And sort of everything is kind of things flow into the black hole and nothing flows out of the black hole. And the black hole is this kind of unique place where for example, time can end and sort of one is in a situation where sort of things go in and then everything stops. Well, in branchial space you could have the analog of black holes as well. We don't yet quite know how those work. But I think that's what objective reality reduction is about. It is about the singularity that can happen in branchial space where these many threads of history all are sort of sucked together into this one singularity. And interestingly enough, Roger Penrose, one of his achievements is this so called singularity theorems, which say that there is inevitability to the formation of a black hole. When there's enough energy momentum in a particular region of space and you get these sort of trapped surfaces where light can't escape and so on, it inevitably forms singularities. And no doubt the same thing is true in branchial space. And no doubt that same kind of singularity idea would lead to this notion of crushing down to a single state. But I'm not sure that that's what happens. I don't think that's an important effect of brains. You know, I think it's a funny thing because, you know, I've been interacting with some philosophers and you know, lots of techies about AI and things like this. And it's still kind of the philosophy and the science definitely are not all on the same page yet. I think that there's still the discourse on the philosophical side continues to be as it has been for thousands of years, continues to be very much at the kind of we can give an argument with words kind of level. On the sort of the AI techie side, it's like, well, we put a bunch of neurons into this thing and suddenly it starts producing text. And we don't really know narratively how that works. So there's really a disconnect between these two worlds of the kind of the narrative description and philosophy where you can make an argument with words and the kind of. And we made 100 billion neurons and it did something interesting. But we don't really know, we don't have a bigger picture about why that's happening. And I think that's an important piece of what one might call sort of LLM science that is an emerging possibility for the next few years to bring those things together, to be able to take this sort of what has been described in words philosophically and on the other side, look at kind of the actual mechanics of what's going on, the physics of thought, so to speak. I would say also that when it comes to psychology, I mean these different levels, there's sort of a psychology of the phenomenology of how brains work and how people do things and so on. And then there's the neuroscience going down to the level of how individual neurons fire. And it's been a big mystery in neuroscience. What's the kind of intermediate language of neuroscience, the description that is above the level of individual neurons and below the level of psychology. Well, we have the same issue now with LLMs. We've got the low level description of we can look at the individual neurons and so on, and we've got the high level and it talks to us more or less and we still don't have that intermediate layer. Knowing what that intermediate layer is like will probably give us a great science of neuroscience. It will probably also give us some interesting insights about psychology. It will probably also give us important insights about AI and so on. And you know, who knows what the, what the mind hack that turns out to be the great thing you can do in prompt engineering for a chat, GPT or something will be. But that's something that can come out of that kind of science. Let's see. Oh my. Lots of interesting questions here. Boy. Eigen is asking about Aristotle, Leibniz, Skirdle, people like me. How was were people able to understand the idea of universal computation? What reached that insight? That's a very interesting historical question. So, you know, universal computation, the idea that things can be programmable, that you can have one machine that can do all these different kinds of computations, that's an idea that I feel is sort of an idea that happened in the 1920s and 1930s. And it's bizarre that it didn't happen earlier. And it is a lesson to us in the kinds of things that the kinds of thoughts we cannot form because we just don't have the right paradigms. And it's kind of a lesson to us today. What things are we not understanding today which will seem utterly obvious in the future? Because the fact is Aristotle, could he have understood the idea of universal computation? Could he have understood the idea that there can be a description? So one of the key things is this notion that code and data can be the same thing, that the data, the things, the sort of, the dead facts of something can also represent code can also be instructions for doing things. So the question would be, I mean, what would Aristotle have said that could have led to something like that? He could have said that the. Yeah, actually he could have had a way of thinking about this. Let me give an example. So from the early days of. Well, I don't think Aristotle quite had this. But for example, Panini in the Sanskrit tradition had this idea of writing down sort of a descriptive grammar of Sanskrit and which sort of was. It's a little bit hard to understand. And I'm not completely sure we're not back projecting some things from modern times, but in some sense it was kind of a. This is how you make Sanskrit sentences. It is a procedure, it is written in Sanskrit, but it describes how to make Sanskrit. And that's a little bit close to a universal computer. It is something which is written in the data that the computer manipulates, but it is saying how the computer can do things. So could Aristotle have had that idea of saying we can write a description of what we do and it is written in the same terms as the thing that we produce as our output? I mean, in some sense. Well, okay, so that was a thing. Now something that I think Aristotle did understand was this idea of the equivalence of human languages. It is not self evident that it is possible to translate from one human language to another and that the same abstract thoughts can be expressed in different languages. And there's still argument about. And it isn't even the case that, you know, there is a bit of a difference in complexion between the thoughts that can be expressed in one language and another. That's a whole different thing. I think we understand that much better in terms of thinking about rulial space and all these kinds of things. But, but you know, this notion of interconvertibility between languages, that was something which was understood in antiquity. And again, another hint towards universal computation. Leibniz, mid to late 1600s, he got a Lot closer. That was. I mean, there was a kind of a. Well, this idea that there could be a representation of thought, a concretified representation of thought. There were other people who talked about that. Ramon lull in the 1200s, a Catalan thinker, had this very bizarre kind of way of. Of having the sort of somewhat mystical kind of sort of things he would write out, which would be kind of like things for constructing sentences. You know, it's kind of like a Mad Lib type thing where you have, you know, a few words, then another word, then another word. And it's kind of a construction kit for sentences. It's almost like he thought of it, I think, as a construction kit for thought. Well, again, it's. It's going towards the data is code, data is thought. Leibniz, same kind of idea. He was an enthusiast of Raman ls, I think, and he got to the point of having this notion of a characteristica universalis, some kind of logical language. He knew about logic, that some kind of logic language, sort of mathematical like description, that could talk about everything. And that again got closer to. To the idea of you can talk about everything in some kind of form that is sort of concrete and specific, and yet it can talk about everything. And that's again a kind of a complexion of universal computation. Now, in, I think, at the time of Hilbert and the sort of the flowering of metamathematics around 1900 and so on, then there was. Well, in mathematics we will have this framework, this meta model of mathematics which can talk about any conceivable mathematics. That became another, much more concrete way in which one might have a single framework that could talk about everything. Well, there were some attempts at that framework with primitive recursive functions, things like that, that didn't work out. Then in 1920, Moses Schoenfinkel came up with combinators, which as we now know, are in fact universal in that sense. And then Emil Post came up with Post systems, then Godel came up with general recursive functions, Alonzo Church with lambda calculus, Alan Turing with Turing machines. All of these were descriptions which went from a single kind of descriptive framework to kind of the implication that it might cover all of mathematics. Nobody was sure it would. Nobody knew that for sure. It was all a hypothesis on the part of all those folk. And one of the big discoveries was all those different hypotheses about what might describe mathematics in general turned out to be equivalent. And that's a big clue. And then this notion that, well, with Turing, the thing became sort of more concrete. And when Moore got the idea of a machine that was programmable and so on. And that's where this kind of, that's where this started to emerge from the notion that that might extend to the natural world. As I say, that was a notion that really took until the 1980s and was very much firmly, I can even say in the 1980s, even into the early 2000s, it was a thing that was widely not believed. And I think at this point we're getting to the point where, yeah, that really is the way things work and have to work. And it is in a sense remarkable that this particular idea of universal computation that emerged both from mathematics and then from almost the thinking about machinery that Turing did, that that turns out to be something that's relevant to the, to the natural world. Let's see, Eigen is asking about von Neumann, John von Neumann and his work to understand models of computation in the mind. Von Neumann's last book was this thing called the Computer in the Brain, which I think von Neumann was an optimist and an enthusiast and said, we've got brains, we've got computers, computers are going to be like brains. And in fact, his understanding of how to make a practical electronic computer, he was kind of borrowing from McCulloch and Pitts and their work on neural nets from 1943. And they in turn were borrowing from Alan Turing. The idea of universal computation and the idea that one could get a neural net to do the same things a Turing machine could do and to do sort of arbitrary computation. And that's actually, I think, how von Neumann even he, he knew of Turing the person, but he didn't understand Turing machines until he kind of went through that back route through neural nets. And the final, I mean, von Neumann was, you know, I would say he had a lot of ideas about what will be important. For example, he was very concerned about reliability of components. How could you make a big electronic computer or a big brain that hold its components in it even when some of those components might fail, why did that not cause the brain to just freeze up or the computer to just freeze up? He was very concerned about issues like that that turn out to be kind of a non issue. And the things that, you know, I'm not sure that I would say, and maybe I'm being unfair here, trying to remember what's in the computer in the brain. It's a little thin book of. Based on some lectures he gave at Yale. And he wrote it when he was very much dying of nasty cancer at the age of like 53 or so. And I think that. I'm not sure quite what he contributed to that whole discussion, except that he did the important sort of piece of arc of intellectual history of merging sort of through McCulloch and pits and neural nets, merging kind of the Turing machine story into the practicality of actual electronic computers that were being built. Let's see. Oh, boy. So many interesting questions here, you guys. Well, Jamie is asking, has the importance of areas of science shifted in history? What was the main focus of science 500 years ago, 100 years ago, 10 years ago? You know, science progresses by sort of the invention of new methodologies which then open up new areas. Very energetic things happen. You know, whether that was chemistry in the late 1800s of we can make all these chemical compounds to physics in the beginning of the 1900s of we've got all these mathematical methods, we can do all these kinds of things to computer science in the 1970s, 1980s, 1990s, we've got all these algorithms. We can make two kind of things about AI where we didn't know it was going to work. But now we have this kind of large language model thing. What, what keeps on happening is there are these kind of methodological advances. Things open up, very intense activity for a short period of time. Then that kind of area becomes much more standardized and moves forward. What can happen, I think what drives science more than anything is these moments of methodological advance, which are sometimes a matter of experimental method, sometimes a matter of a new way of thinking about things, a sort of paradigmatic change. Those are the things that ultimately are the big things that drive science forward. And they happen rarely. And when they happen, they often happen in a way that defines a new science. And, you know, I wrote this big book called the New kind of science 20 years ago now, 21 years ago actually, and next week, the. Which was an attempt to kind of define a new kind of science based on sort of computation as a core way of thinking about things. And that's an example of kind of a paradigmatic change of the kind that's happened a bunch of times in the history of science. And I think that the. This is, you know, it's very hard to predict these paradigmatic changes, but they often define they're not. They're off. Sometimes they are, we've got an existing science, we know what it's talking about, and boom, there's a new method, there's a new experimental method, there's a new theoretical method, and the existing kind of content area of that science suddenly wakes up and exciting things happen there. But sometimes the subject matter is different. You know, in, in thinking about LLM science, for example, the subject matter of that science is really going to be different. It's not going to be psychology, it's not going to be computer science, it's not going to be physics. It's going to be a different kind of thing. And that's, you know, that's a, that's a way that advances happen. Now, when you say what kind of science was done? Well, not all the sciences that we have today even had a name. They weren't even a thing one thought about. Computer science wasn't a thing one thought about 100 years ago. It wasn't a defined kind of a thing. And, you know, that's something, you know, that's what happens with the methodological advances is sort of new areas get opened up that become new sciences. Now, I would say that the other thing that happens is the application of the science and technology. There are areas of science that get sort of enmeshed with technology, and that's a major driver for sort of more development in those areas. And that's, I think, something that, well, we see in some kinds of things. We don't see in other kinds of things. We see there are areas of physics that have become enmeshed with technology and are really zooming forward, like condensed matter physics, for example. There is a physics that have not gotten enmeshed with technology, like particle physics, which have a different kind of, a different kind of story to them. So you know what, the focus of science worth remembering 500 years ago? Well, let's see, a little bit more than 500 years ago was Copernicus. And that was a key moment in science because there had been this question of was science just natural philosophy? Was it a matter of debate? Was it a we can just use our senses and debate what's going on? Or was there some sort of formal tower where you could say, well, I got this calculations and they show that the Earth can be going around the sun and so on. And even though that doesn't seem to be our common experience, and even though by pure sort of human argumentation we wouldn't conclude that, then there was this notion that there is an abstract formal method that concludes that that's a possibility absent, different from our kind of everyday way of thinking about things. So that was sort of a major change. I think there's a similar change. Computation is another such change. I mean, I think the notion of kind of the intuition of computation and things like computational irreducibility and so on, they are a very. They're sort of a major change in the way that one thinks about thinking about things, so to speak, just as Copernicus was a major change in the way one thought about thinking about things. From we can just argue it out to there are formal facts about the world which are not amenable, which are not immediately accessible to kind of our everyday experience. Well, let's see. There's so many questions here. Paula is asking, is there a connection between these advances to science and education? Does education evolve with these changes? Well, sadly, not really. I mean, education. Look, we're still in kind of early days for very widespread public education. Early days. 150 years, 120 years, depends on what part of the world you're in. But 100 and a bit years in many places. And, you know, there's been a sort of definition of what is the thing to communicate in education. Now, if we go back even to antiquity, there was a certain set of subject matter. Aristotle really was the person who, more than anybody, sort of captured the list of these are the things you should learn. I mean, some of what he said was totally goofy from modern point of view, but he was sort of the encyclopedist of the time. It's just, go read the encyclopedia. It's kind of, by the time you've got the encyclopedia and the Bible or your analog of that, you're kind of your set for what you have to learn. Those are the things, so to speak. And I think the education moves kind of shockingly slowly in many ways. And in a sense, education should be, and is about, you know, we have a human society, and when born, we don't know how to operate in that human society. And particularly as society has gotten more and more sophisticated in all sorts of ways, we have to kind of be taught. It's the way our species has evolved. We have to be taught how to operate in human society and so on. And in a sense, education is kind of this normalizing force that's like, okay, everybody, you know, in Place X, you know, all kind of learn the same way to operate in human society. And people can get very upset about, oh, you don't want everybody to think the same way. Oh, you, you know, and, and in fact, of course, that is in. In contra. In contradiction to the idea that we want new stuff to happen, because if we get everybody thinking the same way so they can all operate in the same society and sort of live happily with each other and all that kind of thing, then you're kind of driving something towards alignment, parallelism and thinking. But yet, if you want new things to happen, better be the case that somebody says, oh, I can think outside of that box. And that's in a sense one of the paradoxes of education. It is both useful to align and useful to allow a lack of alignment. I think that hasn't gone that well. And I think as we see kind of more pressure on the importance of education in sort of the, at least in the sort of some parts of society, so to speak, there's much more of you've got to run down this racetrack to get to the other end. And the only way you're going to get down the racetrack is to go straight in that, in that straight line. And the idea that there may be other things you could look at on other sides is something you've got to ignore because you've got to get to the end of the racetrack, so to speak. So I think the thing that what we need to see and what we can see is a change in education, but boy, is that a difficult thing to achieve. I think the very exciting thing that kind of LLMs have brought to us is a possibility of total broadening of access to this computational paradigm. It isn't just going to be the sort of high priests of programming and things that have access, and this has been my lifelong effort of mine, is to try and kind of broaden out the access to this very powerful idea and technology of computation. And I think that LLMs help in that broadening process in a fairly important way. And it's like, what is post LLM education? What is it like? We don't know yet. I'm thinking about it. I think it's very interesting question. There are things that very suddenly don't need to be taught the way they were taught. There's much more possibility for personalization. There's much more possibility for. You don't really need to learn that you can just ask the machine to do it. That's been a thing that's happened. I myself have been responsible for some of those changes in a bunch of areas of more mathematical, quantitative kinds of things of you don't have to do that integral by hand, a machine can do it for you type thing. And it's kind of the, the progress of, of society, of civilization and tech, you know, is, is really is measured in part in the progress of technology. What can we automate? What can we not have to do ourselves? What can we just take to the Point where we have to figure out what we should want to do, but then where it can automatically get done. And then to what extent do we need to know sort of how the car works and how much do we just need to know how to drive the car? And I think we're seeing another great moment of we just need to know how to drive the car. We don't have to know how to, you know, how it works inside. We don't have to, you know, spend every time we're going to go on a drive opening up the, the engine and, you know, rearranging wires and so on. We don't have to do that. We can just get in the car, know, switch it on and start driving kind of thing. And that's the thing that increasingly we're able to do with computation. And that's a thing which we should be teaching future generations how to do because that is the, that's the place where we both. It is a much broader thing to teach. It is a much less, you know, if you don't happen to have a mind that's well oriented towards the, the detailed machinery of, of, of programming languages and so on, it doesn't really matter anymore. It's something where if you have the ability to think in general, you're coming out ahead. You don't need the specifics of, oh, I need to know this very formal, detailed, technical thing. So I think it's an exciting moment. It will be interesting to see how that evolution takes place. I have to say my observation over the last 40 years or so of the ways of education is that it's awfully slow moving and there are so many forces that in part serve this other need of education, of sort of keep everybody aligned, which is something that almost by definition has to be a kind of slow moving activity. Let's see, maybe a couple more questions and then I should. Move along here. Let me see. Katie asks what's been the most important invention that has improved research overall? Well, there have been a bunch of things I think in the. I would like to think that computation as a paradigm for describing things and as a way of automating things has had a contribution. I would say it's had more of a contribution in terms of the practicality of automating what was done by hand before than it has yet had in terms of a way of describing things, you know, as mathematical notation had this way of crispening up how to describe things mathematically 500 years ago or so now. So computation at least my efforts in sort of computational language are a way to kind of crispen up the way to think about things computationally. That has happened to some extent, but there is a big tsunami of activity that will happen in the future on that as people really realize how to start formulating everything in computational language terms. Hasn't quite happened. The wave hasn't really quite come through yet on that. And it's more been the practicality of using computer to do things. Now there are other pieces. The fact that sort of communication of scientific information is a bit faster, it's a bit cheaper. It's, you know, back in the day, there were, you know, when I started doing science, it was a question of, you know, there were journals and they got mailed to place, and you go to the library and you look at them. And then there were preprints and they got mailed to places and you went to the library and you look at them. Or sometimes you'd get them in your own mailbox, physical mailbox, that is. And then even at that time in the 1970s, there were already online abstract services where you could go and dial up to this thing and you could. Well, once you had access to a computer, you could go and you could start searching abstracts and things very much like what you would do in a modern search engine, but just abstracts, scientific abstracts. For some bizarre reason, only the druids used it. And by druids here, I mean librarians. You know, people would go to reference librarians and say, can you look this up for me? And I. I realized around, I don't know, 1978, 1979. Wait a minute. It's just some system. You just log into the thing and you go and look it up yourself. And it has a slightly arcane syntax, but it's not. Not that arcane. And so, you know, I got myself an account and. And, you know, you could go off and you'd start lugging these things up. And a lot of things I discovered in the, well, early 1980s, you know, I was able to look up all these references online, and I'm not sure if I'd had to go through the sort of extra layer of librarians, and things would have been much more difficult. I mean, this whole thing about being able to do computations yourself as opposed to going through a loop of having a computation expert doing them, that's been important. I would like to think that our Mathematica system that came out in 1988, I would say it was actually fairly dramatic, that it really went from the point where theoretical and Mathematical scientists always had to have a programmer in the loop kind of dealing with that nasty oily machine, so to speak, to the point where they could, with their own fingers, cause a computation to happen. That was a big thing. I mean, for me, you know, my whole sort of story of doing research has been using computers as tools. It always surprised me that that tool use wasn't as extensive in the world of science as it might have been. I think in this kind of ready access to information that the web has provided and that the conventions around the web have provided is also important. I think that one thing that is true today that wasn't true in the past, you know, even when I was working on my book New Kind of science in the 1990s, to know that I knew everything about, or not everything, but to know that I'd really dug to the bottom of some particular question and really mined out the history of some particular thing that was quite hard to do, that required serious library work and so on and often talking to lots of people. It's much easier today to know, does the world know anything about this? Okay, I'll search the web. I'll go look here, I'll go look there. It's much easier to get to the bottom of those things. And that's, I think, an important, important development. I think that the idea of having clear expositions. I think our math world system was a good first example of that. Wikipedia kind of based a bunch of things on math World as a kind of, let's just say what there is to say as opposed to the kind of more certainly more ornate style of textbooks that had existed in the past. That's another thing that, that's definitely helped in terms of, you know, just tell me the facts type thing. And, you know, I would say in terms of, let's see, I was, was going to this whole. Well, the, the kind of. There's been both good news and bad news about the decreasing cost of putting sort of scientific material out into the world and research material out into the world. The good news is more stuff is getting out there. The bad news is more stuff is getting out there and it drowns and things. There had to be these whole systems built that in a sense are systems that are built for the average when there's something really surprising that comes out there. The systems that are like, oh, we've got the, you know, the kind of. The physics, you know, system that is dealing with the stuff that normally happens in physics and the 50,000, 200,000, whatever papers that come out about this kind of thing. We've got the thing that handles the 200,000 papers. When the one really paradigm changing paper comes out, it's like, oh my gosh, we don't have a way to deal with this. Partly because there were 200,000 other papers at a time when science was much smaller. That was a different story. Look, in terms of other things with research, another area that is still as yet just bubbling up is automated data. Taking the fact that instead of the person writing it in a lab notebook, it's a heavily automated system. There's a lovely one based on Wolfram language that Emerald Cloud Lab has built for biology and chemistry of just automating kind of doing those experiments. And again, that revolution is just at the threshold of happening, of being able to go from something where it's kind of like the description of the experiment to the doing of the experiment and taking the data automatically and having that all be part of sort of this computational process. That's a thing that's, I would say, just emerging now. You know, I think one of the things about research and the rhythm of research is this, that that there are times when there's a kind of an ability to absorb new ideas. And there's a time when it's really hard for new ideas to emerge. And usually in these periods of time when fields are just sort of making incremental progress, they're going, okay, they didn't get stuck, but they didn't have a big burst of paradigmatic progress. In those times, it's pretty hard to insert new ideas. Takes a lot of effort. At times when fields have kind of gotten to the point where, oh my gosh, we're stuck, that's okay, you can inject a new idea. Or at times when everybody knows there are new ideas in the air, that's another time when one can inject those things. So kind of what makes research sort of happen and be energetic is often being in one of those periods for some particular area. Let's see. Yeah, Rock is commenting in 1991. There's a system called Eric which was indeed an early online system. And. Yes, good example. Let me see, Maybe one more thing here. Okay, let's try. I got two more. I think I'll look at one from rbs. Historically, what have been the most difficult problems or obstacles for us to overcome or solve in areas of science and technology? I think my answer to that is sort of the development of new paradigms, new ways of thinking about things. That tends to be the blocker. And it's Often those new ways of thinking about things define what questions one should ask. And if you don't know what questions to ask, you can't even get started if you don't know. We could have invented lots of stuff about theoretical computation long before there were computers. Doesn't require computers. It's abstract, conceptual kinds of things. But until there were computers and until we had this paradigm of thinking about computation which came from the technology of computers, mostly we just didn't get to ask those questions. So I think that this question of what's the framework in which we're thinking about things, which defines what kinds of questions we might ask, that tends to be a big blocker. And it really, until that emerges, we just can't make forward progress more so than where the question is in sight, and we just don't know how to answer it. Now, there are areas where we have that, for example, in biomedicine, there are plenty of areas, you know, can we cure cancer? Can we, you know, prevent aging? Can we do these kinds of things we would like to do? Those questions are absolutely in sight, which is. And we kind of know the definition now. Now sometimes those questions, when we peel them apart a little bit more, they. They become a little bit. It becomes a little bit less clear that the question really is in sight. You know, like, I don't know, something about aging. You might say, well, what really is aging? What is the, you know, is it telomeres dropping off? Is it oxidation or, you know, oxidative damage? Is it this? Is it that? You know, what really is it? You know, you want to cure it, but what really is it? You know, we have an intuitive sense of what it is, but if we want to, you know, we need kind of a way to think about it that allows us to define it better so that we have a chance to sort of attack it in a scientific point of view. So I think, as I say, I think that that tends to be the issue. And in technology as well, it is often a what's the right question to ask. It's like, I don't know, a social network. You know, the concept of those could have been invented, that the raw material for them could have existed quite a long time before they actually became a thing. But it was kind of the idea that that will be a thing that was worth having was an important step. That was sort of a paradigmatic step. And, you know, this is the conceptualization of what is worth doing is an important piece. And that, I think, tends to be the biggest blocker you know, it's. It's. Now, again, there are counterexamples. You know, one could have said, okay, a good example with the whole ChatGPT thing. It could have said, yes, we want a system that can resolve, can pass the Turing test. It can generate natural language. It can have conversations. People have been kind of science fictionally talking about that for a very long time. I think the precise definition of what should it do? It didn't quite have that. It was all a bit vague while it's kind of Turing test like things. But we don't really know how it would work. And I know I, for example, had not really conceptualized that a text completion system, when it worked well enough, would be able to pass the Turing test, basically. And that was a thing that, again, if one had known this sort of precise definition of given a prompt, can you continue? If one had really conceptualized that, one probably could have solved this problem, maybe. But that was a case where curiously, the technological methodology advanced ahead of the conceptual side of things, ahead of the kind of. And this is how it will work and why it will be useful. And we can now. We're now kind of. It was a case where sort of a surprise, sort of scientific technological discovery happened that then caused one to think about it. I mean, I suppose there are plenty of others where there were sort of surprises. I don't think anybody expected X rays, for example, back around the late. The end of the 1800s and so on. And that led to a whole level of other kinds of things being studied and so on. So occasionally these things happen and kind of, you know, the surprise discovery gets made sort of independent of paradigm. All right, one more. Well, okay. Pedro is asking about unintended consequences of revolutions. What lessons from the Industrial Revolution have we learned that we could use for the AI revolution? Oh, boy. You guys don't ask simple questions. What happened in the Industrial Revolution? Well, there were things. The concept that you could have machinery and production lines and so on, eventually that would take over from the individuals doing things for themselves, the centralization of production. I think that that was a. I mean, already there had been plenty of kind of, you know, you have the particular tradesmen doing this. It wasn't. Everybody wasn't a blacksmith. You went to the specialized blacksmith to have the horseshoes put on or whatever, as opposed to just everybody as their own blacksmith. But then this idea that one could kind of go from sort of individual human effort for this machinery and so on, I suppose what were some consequences of that? I think some of the unintended consequences were a different way of thinking about things. The concept that you could make things industrial, that you could have kind of this thing which became a production line. I think that led to a bunch of thinking about education, about political systems, about these kinds of things. I think that was quite unexpected. I don't think anybody could have expected that this idea of, let us make kind of the organization of countries as organized as a production line was a thing that would come out from this kind of machinery that was made in the Industrial Revolution. I think another thing that the whole structure of society and the notion of kind of the centralization of kind of industry and so on, that's another thing. I mean, there had been a centralization of governments with monarchies and things like that. The centralization of the industrial, of the process of manufacture with kind of the quotes, captains of industry type thing. That was a thing that was sort of new at that time and I guess was a. I'm not sure anybody really expected that. I think also the kind of whole structure of there are towns, there's, there's, you know, you can, you can have this whole sort of centralization of labor, of people kind of coming to the site of the industrial stuff as opposed to everybody has their own, you know, everybody has their own equipment out on their own farm type thing. That's another thing, sort of the centralization of that. So a good question. What does this, you know, what can we learn in the AI world? Well, okay, I hear a few things, I think so. One of the things I think is going to emerge is the concentration on specialization in education. This kind of idea that you should sort of the best thing to learn is the specialized tower of knowledge in some particular area. My guess is that it will become clearer that that isn't a good idea. Just as back at the, at the time of sort of the Industrial Revolution, the exact opposite happened. You know, it was the case you were operating your farm and you kind of had to do everything and then. Well, actually, no, I was just working on a production line and just had to do this one thing as a very specialized thing. My guess is that will reverse and that the value put on general thinking and general knowledge will increase greatly. So that's perhaps one unintended consequence. I think that the question of sort of who has access to, to the most powerful tools, so to speak, what has happened in the Industrial Revolution and following that, all the things that happen with the sort of growth of computation and kind of the technology is that the people who really quotes understand what's going on. It becomes a smaller and smaller group of people. And yes, it is. I mean, in a sense, it's kind of ironic because at some level, you know, with something like Chat GPT, nobody understands what's going on. You know, I wrote this book about what's Chat GPT doing and why does it work. And arguably I, you know, I have probably a better understanding than the average of kind of how it works. But I wouldn't say, I quote, understand how it works. It's a, you know, I know the mechanics of what it's doing, and I have some ideas about the sort of bigger picture of what it's doing, but it isn't the case that it's like, you know, I kind of can be an engineer who's adjusting this one valve inside the thing, and I know what it's going to do as a result of that. So in a sense, we've gotten to the point where for a while it was like only a few people understood the machinery to like, nobody really understands the machinery, and we're all in the same boat. And that, in a sense, makes it a thing that is. Is kind of a. I think that's an interesting moment because it really flattens out the sort of the access to this kind of, the. The ability to really leverage this kind of technology. You know, another, another kind of thing is some of the questions about sort of the qualitative understanding of how the world works. There's been a long period of time when it's all machinery, it's all mechanical, it's all clockwork. It's all. The clock is assembled this way, and so the hands are going to move in this particular way. We can open up the clock and we can see what's going on inside. That's been the tradition from the 1600s to today. It wasn't the tradition before that people didn't think you could open up the donkey and see how it worked inside and expect to make predictions about what the donkey would do. That is a thing of modern times that was very much amplified by the Industrial revolution and the technology that came from that. That is a thing that, as we work more on computation and we understand ideas like computational irreducibility, the notion that just because, you know the rules by which something operates, it doesn't tell you. You can kind of immediately say what's going to happen. You have to just follow those rules and see what happens. Well, the fact that there is this phenomenon of computational irreducibility and the fact that it Says there are computations where you just have to follow the computation and see what happens. And that's the value of doing the computation. You can't just jump ahead and say, I know what the answer is. There's this kind of trade off, which I think will be an increasingly discussed thing between computational irreducibility and controllability. That is, if you just have to wait and see what happens, you don't get to say, and I know what's going to happen. It's going to always do the right thing. You have to kind of let go a bit of that if you want to have the advantages of computation, so to speak, and the advantages of the automation that you get from that. And I think that's going to be a sort of a major debate back and forth. You know, to what extent do we want to rein in the computers and make them only do the things that we can readily understand and have a narrative about? And to what extent do we want them to do great things, perhaps sort of behind the scenes that achieve these particular human objectives, even though there are some things that will go on that we do not understand. And that's going to be, I think that will be a big, kind of a big debate of to what extent do we want to have less wonderful things happen, but more predictable things happen versus letting more wonderful things happen with the recognition that there could be things that go wrong that we don't expect. And I think what we're seeing, I think there will be, well, other things that will happen. I mean, some of the things that this idea that was an industrial revolution idea that you can kind of production line education, that you can put everybody in the same class and they learn the same thing, just like everybody's going down the production line and doing, doing the same thing. Well, that will no longer be a thing. We can certainly expect that there will be a much more personalized version of education and a much different set of things that we're trying to teach in education. And I think there'll be sort of a post AI revolution version of that that is paradigmatically different from the post industrial revolution pre AI version of that. So that's another thing I would expect. I think that this whole question of, well, a lot of questions about, oh, for example, question about sort of the government of the world and how that works and the extent to which sort of what we can change and the way that works as a result of sort of AI Another set of interesting questions. I'm trying to think what's an obvious Unintended consequence of the same kind that was the sort of stratification of society and the emergence of cities that happened from the Industrial revolution. I mean, I suppose some kind of utopian thought which may not happen this way is that there will be something of a reversal of those kinds of things in, As a result of, of, of kind of the AI revolution. I'm not sure if that's really how it will work out. I mean, there may be unintended things. This is the life of computational irreducibility. There may be unintended things. I'm certainly not seeing about what. What will happen. And I think the. Yeah, I mean, yeah, very interesting question. Well, I should probably. Wrap up here. I. I see from G flat. Do you think it's fundamentally possible for science as we know it to hit a wall and slowly degenerate? Every individual area of science with every particular paradigm gets mined out, and it needs a new thing and possibly a new kind of science to make further progress. That's been the lesson of the history of science, the history of ideas in general. I think that this issue of will we ever reach the end where we've discovered everything that can be discovered, invented everything that can be invented? The answer is no. In fact, computational irreducibility guarantees that in a computational universe, so to speak, you'll never reach the end of what can be discovered, what can be invented. There's always more to discover. There's always more sort of places where you can find that little piece of computational reducibility that lets you do a thing that you couldn't do before. The one question, which I think is an important question is, do we eventually say, that's enough. We've got to the point where we have everything we need. We're living happily ever after. We just don't need to know anymore. And at different times in human history, some traditions, some cultures have gotten to the point where they say, we're good now. We've got what we need. We're in a stable state. We'll just let things keep running now. In fact, computational irreducibility means that will always, eventually get blown up. Something unexpected will always, eventually happen. That will always lead one to be pushed in a different direction. But I think that the way that the real issue is, do we decide that we're good now? We don't have to push further forward? And that could happen. There could be periods of time where it has happened in human history where there are periods of time where people sort of say, we're good now. And that can last for in the past at least hundreds of years. Whether that's still true, I'm not sure, although human nature does have a certain constancy to it. And until sort of there are major changes in human nature, which I think will be associated with things like effective immortality and things like this, we're stuck with human nature as it is. And so there will be some things that I don't think will really change their rhythm and character. All right, I better go back to my day job, but thank you all for lot of very interesting questions and comments, and I look forward to doing this again in. Well, next week and after. Thanks a lot and bye for now.