Into the Impossible With Brian Keating - Are Humans Smart Enough to Understand the Universe? (ft. Stephen Wolfram)

Starting point is 00:00:00 Your summer starts now with Memorial Day deals at the Home Depot. It's time to fire up summer cookouts with the next grill, four-burner gas grill, on special buy for only $199. And entertain all season with the Hampton Bay West Grove seven-piece outdoor dining set for only $49. This Memorial Day get low prices guaranteed at the Home Depot. While supplies, price invalid May 14th or May 27th, U.S. only exclusions apply. See Home Depot.com slash price match for details. When you need to build up your team to handle the growing chaos at work, use Indeed

Starting point is 00:00:35 sponsored jobs. It gives your job posts the boost it needs to be seen and helps reach people with the right skills, certifications, and more. Spend less time searching and more time actually interviewing candidates who check all your boxes. Listeners of this show will get a $75 sponsored job credit at Indeed.com slash podcast. That's Indeed.com slash podcast. Terms and conditions apply. Need a hiring hero? This is a job for Indeed sponsored jobs. aren't whales building rockets? They have bigger brains than we do after all. In this episode, we talk about why more brainpower doesn't always mean more understanding and how neural architecture faces physical constraints. Yes, we are locked in a kind of prison, which is the prison of what

Starting point is 00:01:16 human minds deal with. The things we care about are the things that human minds can kind of deal with. But there's a lot else out there in the computational universe in the Rulia. That is behavior that human minds can't really wrap themselves around. as even super-intelligent AIs may hit hard computational limits. In our conversation today, we explore why intelligence has a ceiling and how ideas like Wolfram's really have computational irreducibility and prane-sized scaling reveal the boundaries of thought itself. Wolfram created Mathematica, Wolfram Alpha, it's probably in your pocket right now,

Starting point is 00:01:51 in your cell phone, and he's now building a radical new theory of everything grounded in computational reality. If he's right, smarter doesn't always mean deeper. It might just mean we get stuck. Stephen Wilfrum, you've created Mathematica, you've built Wolfram Alpha. You've basically taught computers how to think. Your theory of everything, what you call the Ruliad, is considered by many and to be the front runner among computational approaches to fundamental physics.

Starting point is 00:02:17 But here's what I really want to know, Stephen. If the universe is just the entangled evolution of all possible rules and observers like us are simply slicing our way through the Ruliad from our own computational vantage point, then what makes our experience, our qualia, what it makes them feel so real, so privileged, what does it mean to have a brain? You're describing how the universe might even be thinking in a certain sense, but not in the woo-woo sense. Are we discovering the universe or are we really just bumping up against the limitations of our own computational prison? Well, let's see. I mean, the way I see it these days, the sort of the Ruliad is a representation. of everything that is computationally possible.

Starting point is 00:03:02 We, each one of us, is sampling a tiny thread of what is possible in the Rulia. And what happens in the Ruliaad. Just as we are sampling a tiny thread of what happens in physical space, you know, we're sitting on this one planet in a, you know, in a corner of a galaxy, that's one of 100 billion galaxies, you know, we are sampling a small part of the, of even the physical spatial universe, let alone. this kind of much larger computational universe that is the Ruliad. So, you know, if you're asking what makes that feel real to us, what else would it feel?

Starting point is 00:03:41 That is, if we feel anything, we will feel that it is real, so to speak. If we are, we are not, you know, and if we're asking what, so, so there's a sort of interesting question of, you know, how do we know that anything is real? what does it even mean for things to be real? What's the difference between living in a simulation of the real and living in the real, so to speak? So these are complicated questions. There's a whole bunch to say about them. But maybe we can kind of dig into questions about, well, for example, let's say the only thing that any of us are really aware of is what we are perceiving.

Starting point is 00:04:23 In other words, I have a certain feeling about what's going on. I know what's happening in my own mind to know anything about what you might think is happening. That's merely an inference. You seem a bit similar to me. So I kind of project what I feel is going on as something that I would imagine you also feel is going on. And that's how we kind of have a shared sort of objective reality. Each one of us has just the particulars of what's going on sort of inside our own mind. I mean, I've kind of often thought if you think about a computer and what does a computer think is going on in the world? Internally, to the computer, it is something very similar to the kind of thing that we think is going on in the world. It's just that we don't identify with computers.

Starting point is 00:05:13 So we don't sort of, we don't project ourselves into, it seems very alien and unnatural to us to imagine that there's a thing being perceived by the computer that is kind of like the thing that's being. perceived by us. You just jumped right into a pretty complicated area. I mean, there's a lot to untangle there. To give another thing, people, people would say, well, what, you know, could the universe be a simulation? What is what mean by that? I think what people often think they mean is there's some simulator out there who's playing a video game and we're all part of that video game. That has a certain implication. that there are many video games that could be played, and the kind of godlike figure who's playing the video game

Starting point is 00:06:00 picked this particular cartridge or whatever it is to put in, and that's the one that we're all operating within. That's kind of the idea that there is a simulator who is making arbitrary choices, and we are then the working out of those arbitrary choices. That would be kind of one version of what it means, that that's sort of the beginning, what it means to say that we are operating sort of,

Starting point is 00:06:24 simulation is there is a choice about what simulation it is. Now, in my view of how sort of things actually work, the idea of the Ruliaad, this kind of entangled limit of all possible computations, there's no choice about that. There's only one of it. If you just aggregate all possible computations, you inevitably end up with the Ruliad. So the simulator has no choice. And the question of what we actually perceive is then a story of where we are, who we are, how we observe, how we sample this really addable possibilities. And so it's like saying, well, you know, can we explain why the night sky looks the way it does? Well, that's because we're on this particular planet, this particular place, in this particular galaxy, etc.

Starting point is 00:07:14 It looks the way it does. But there's no theory that says why we ended up on this planet in this particular galaxy and so on. That's just where we are is there. And so we have this perception of what the universe is like, what the night sky looks like, and so on. And so similarly, when you say, well, how did we get the particular sampling of physics that we got? Well, it's because we are where we are in the Ruliad. It's not because there is something about the Ruliad that is determining that. It is the kind of the thing that is the case about where we are.

Starting point is 00:07:52 It's not something for which you can derive where we have to be. You can't derive that we have to be on this particular planet. Right. And even in a simulator and the matrix, ultimate matrix hypothesis, the thing that sometimes confounds people is that we imagine, as the title of your wonderful essay, which I'll link to below, what if we had bigger brains, imagining minds beyond ours? And it naturally thinks, you know, it starts off with the, just the raw facts.

Starting point is 00:08:22 of the human brain that in this three-pound supercomputer, we've got, you know, 100 billion or so neurons, which all my, you know, woo-woo friends like Deepak Chopra love to say, oh, that's the same as a number of stars and the same number of galaxies. It's just nonsense, of course. And it's coincidental, just like you said, the constellations or the, you know, the planets that we have are as well. But it seems to me that, you know, computer companies or let me say, it seems to me like AI companies like OpenAI, they can't get enough of bigger and better processes. from Nvidia, for example. And so it might lead one to believe that certainly for human-based computation, bigger is better.

Starting point is 00:09:01 And yet, and yet, you know, Einstein reputedly, reportedly, his brain was slightly smaller than normal and cats have smaller brains and sperm whales have six times larger brains. So why is bigger and not better? How does that fit into this notion? I mean, I would think the Roliad would privilege things that have more and more connections in their connectome. But it doesn't seem to scale, as I know. naively would have thought. Well, you're jumping into several very different deep kind of areas.

Starting point is 00:09:30 So, I mean, maybe we should finish on one issue about sort of the perception of reality. I mean, the thing that matters to us is what we ultimately perceive. There is, we are getting that there is a sort of, we think about the world in terms of the outside and things actually happening in the outside that are there. then being transferred through our senses, through our eyes, through our touch sensors, all those kinds of things, to the internal perceptions that we have. In other words, we're taking, and then the question is, well, you know, what if it isn't actually, you know, your eyes that are sending those signals down your optic nerve?

Starting point is 00:10:10 What if it's something that is sort of digitally generated and it has nothing to do with sort of the outside world as the outside world is? Well, so in kind of this sort of computational view of what's going on, of the Ruliad and so on, in some sense, there is no distinction between the merely computational and the actually, the real thing, so to speak. Everything in the universe is just a feature of these computational constructs. And so it's the question then is, well, what is happening when we perceive things in the universe? I mean, this is where you're forcing me through it, very high speed, through a bunch of really philosophically complicated kinds of ideas. But just to try and say something about, you know, what happens when we perceive things? What's happening is, I think, you know, you were sort of talking about how brains work, I think the key thing that brains are doing is going from all of those inputs that we have.

Starting point is 00:11:11 You know, we have millions of photoreceptors in our eyes. We have millions of touch sensors on our skin and so. on, we're taking sort of those millions of kinds of inputs where those are coming into our brain, somehow we're having a sort of model of the world based on that input that we're getting. And then the big thing that our brains do is they decide what to do next, probably roughly 10 times a second. So they're doing a huge amount of compression. They've got huge amounts of data coming in, and yet all they do with it is to say, what are we going to do next? I kind of got to realize recently that it's sort of a little disappointing in a sense.

Starting point is 00:11:47 We think of, you know, this idea of consciousness and our thread of consciousness and so on is a great achievement of, you know, kind of humans and perhaps human-like things. But I kind of think what started that all off was an incredibly mundane thing. Sometime, you know, a billion or two years ago in the history of biological evolution on Earth, which was when there started to be mobile, animal-like. things, the animal had to decide where to go next. And the animal could go to only one place. The animal can't go both left and right. The animal has to make a decision. It's going to go to the right. It's going to go to the left, whatever. And that means that the animal has to take in all that input

Starting point is 00:12:27 and then come out of that with a definite decision about what to do next. And I kind of think that incredibly mundane sort of need for a mobile thing that is sort of a biological organism is what probably drives the thing that that is terribly significant to us, which is this idea that we, we abstract this thread of conscious experience from all of this input that we get from our senses. So that seems to be, and so there's sort of a question of what, what are we doing when we do that? Well, we're, you know, the big thing is that we're taking all this detail about what happens in the world and we're just deciding we're concentrating that all down onto this sort of thread of perception that we have.

Starting point is 00:13:10 this thread of experience we have. Was that what, you know, they refer to as Galileo's error. So, you know, for you listening, or maybe those of you with that, I said, here's my friend Galileo, who, you know, really said that our job as scientist is to measure what's measurable and make measurable, as he said, what is not yet so. In other words, that was the project, the Galileo project, not your neighbor, you know, in Cambridge over there, different kind of Galileo project, as opposed to the Wolfram Physics project. In Galileo's view, mathematics was the core operating system, if you like, of all of

Starting point is 00:13:46 nature, and understanding it could reveal universal truths. We needed the sensors, but that was the job of man to build sensors to transduce information. But as you point out, and as my friend and a former guest, Jan Lacoon has pointed out, you know, the human eyes processing terabytes of data, you know, in some sense. And yet, the brain must be doing some immense filtering process, which is a simplification. And so in terms of those computation or compression, is compression really the core job of the human brain in your perspective? Well, that's an important part of it.

Starting point is 00:14:21 I mean, that's what leads us to go from kind of the complexity of the world to kind of what we perceive about the world and then how we decide what to do next. But I think I want to come back to Galileo. Yeah. So I mean, your implication there is that, Galileo thought that what was really there there, and I'm not completely sure that that's how I would interpret what Galileo said, but let's take this as a conceptual Galileo, even if it wasn't the actual Galileo-Galoree. Let's consider it to be your avatar of Galileo.

Starting point is 00:14:58 What that avatar might have said is the implication that the true there there in the universe and physics and so on, is something mathematical, and that we are sort of poking at that mathematical thing, that the real sort of way the universe works, one might say, is according to something mathematical. Now, that's a, that's a, that's a, that's a funny concept, because mathematics, we have to decide what we mean by mathematics. What mathematics probably meant to Galileo was a set of ideas that had emerged from kind of human thinking about things. You know, mathematics sort of in its, you know, it's probably emerged historically in ancient Babylon and sort of two different threads.

Starting point is 00:15:48 One was kind of accounting, doing arithmetic, counting, and the other was land surveying and geometry. And those two threads kind of developed over time, algebra developed and so on. And in Galileo's time, it was. Those are, that is what mathematics is about, things like arithmetic and geometry. And those things seem very pure. And I think Galileo might have imagined that there's all this mess associated with how the world actually seems, but at some ultimate level, the world must be just described

Starting point is 00:16:23 in terms of the mathematics that humans had invented to that time. Now, that's a, in a sense, that's a, it's a very, it's a very unhumble kind of claim, because it says we humans in the course of those couple of thousand years of developing mathematics had nailed it. We got kind of the abstract concepts that are going to be that are the core of how our universe works. I don't think that's actually right. I mean, what happened after Galileo, sort of one came to Newton who really made use of this kind of mathematical idea. He sort of introduced the notion that there were mathematical principles of natural philosophy. You could described nature in mathematical terms. Now, Newton was pretty lucky in a sense. So was Galileo,

Starting point is 00:17:10 that the particular things, that maybe it isn't luck, the particular things they studied were things where that approach works. Newton studied mechanics, studied, you know, how things move and what forces are needed to change their state of motion, things like that. Had Newton tried to study fluid mechanics instead of solid mechanics, he would have been completely stuck. When if you look at the motion of fluids, they have things like fluid turbulence and some randomness that you get from fast flowing fluids, those kinds of methods that Newton had that came from sort of the development of mathematics and algebra and what Newton did himself with calculus, they just don't tell you very useful things. That the fact that, and it's a good example of some other things, the fact that Newton was able to use mathematics, Galileo was able to use mathematics. It's peak pollination season, and my business is scaling fast. To keep the nectar flowing, I need a phone plan with top priority data speed.

Starting point is 00:18:09 That's why I chose GoogleFi Wireless. My connections stay strong even when the hive is buzzing. Plus, unlimited plans start at $35 a month. Now that's a deal that doesn't stay. Explore GoogleFi Wireless plans today. Plus taxes and government fees. GoogleFi Wireless is not subject to data traffic deprioritization during times of high network usage. mathematics to talk about things is because the things they chose to talk about were things

Starting point is 00:18:34 about which mathematics has something to say. In other words, if, and so I think that's a, it's a very common characteristic in science that you have certain methods, and then the science sort of wraps itself around the things about which those methods have something to say. Do you mind if I just interject there, because it's something I wanted to ask you for quite some time, which is the interlocking of our technology with the problems of the time. So exactly. exactly what you just said about Galileo Newton, but to apply to today's marriage of LLMs plus GPUs, which were never designed to, you know, replace, you know, human functions of chatting and so forth.

Starting point is 00:19:12 But they're very good at it. They're very good at linear algebra. It's not very sophisticated mathematics at some level. Your book on chat GPT explains it very thoroughly. We talked about that last year. But are we locking ourselves into a new type of prison, you know, Sam Altman's prison? I don't know. Some prisons where these things are so good at doing this one type of very abstract and very important type of reasoning that we might not be able to solve problems like, for example, come up with new laws of nature, ab initio, you know, unify the laws of quantum mechanics with the laws of gravity, etc.

Starting point is 00:19:47 I know I know that you can do that. But in terms of the conception of a theory of everything, are we perhaps because of our success, we are victims and we are entering a new prison? of GPUs plus LLMs. Okay, so there's several different directions there. So one thing I thought you were going to say is, you know, in the time of Galileo and Newton, they had algebra and things like this, and they fashioned their theory to be one where that kind of approach would work. I thought you were going to say,

Starting point is 00:20:18 and here we are today with computers, and you have a computational way of thinking about the universe, and isn't that sort of as limited as the way that Gallow, Laleo and Newton had of thinking about things based on the kind of intellectual technology of their time. Okay, so answer that question. And then I want to talk about the LLMs. Yeah. But that was, so, you know, that's an interesting question.

Starting point is 00:20:41 I've certainly thought about it. I think the thing that for me is kind of the suggestion that we are not fooling ourselves in that way, is the fact that with the things we're doing in computation, we have in some sense reached, the end of what is abstractly possible in the following sense. One might have said, well, you've got algebra, you've got this, you've got that, you've got these different methods, and imagine you're making machines to do algebra, machines to do geometry and so on. The machines you need to do these different things are different machines. What got discovered about 100 years ago was this idea of universal computation. The idea that you could have one machine that could be programmed

Starting point is 00:21:23 to do all these different kinds of things. That's the idea that made software possible, That's the idea that made everything we do in practice with computers possible. But it's also an idea that tells you there's a bottom to what's going on. It's not the case that you keep on saying, oh, I need something sort of more fundamental. It's like you reach the thing that is the universal thing that can do everything that anything can do. So I think that the idea that there's, in a sense, you know, I talk about things in terms of computation because that is a metaphor of our times. I could equally well talk about things in terms of just rules that are followed by a system, which is something a little bit less familiar than sort of what we experience with computers

Starting point is 00:22:09 and the way that computers run and so on. But I think that, you know, what we're seeing is it seems to be the case that, you know, what we know is that at some abstract sense, we have kind of reached the end of what, we've reached kind of the lowest level of what can be talked about in terms of rules, in terms of computation and so on. So I feel much more confident that, you know, we can be talking about things at a truly fundamental level rather than we are just talking about things in the form that we can talk about them in the, you know, third decade of the 21st century and so on, and that at some

Starting point is 00:22:47 time in the future, when we'll have a completely different picture of what's possible. With respect to LLMs, I think the third. thing to realize is that our brains operate in certain ways. Those ways in which our brains operate determine the things that we care about in the natural world. That is, the way I imagine, with the Ruliad, for example, there's a lot of stuff going on in the Ruliad. But yet, our particular sensory systems, our particular ways that our brains work, we concentrate on only certain things. So in a sort of extreme case, we could say, well, you know, I'm sitting in this room, it's got a bunch of air in it, that's a bunch of gas molecules. There's zillions of

Starting point is 00:23:27 molecules bouncing around. But the only thing I notice is, you know, if I, if I wade my hand, I can kind of feel that there's air flowing around it. That's the only thing I notice. I'm not not paying attention to every individual molecule and so on. That is, that's a thing that for, with my kind of sensory system and my way of thinking about things, I only get to sort of talk about these very large-scale fluid motions and things like that. So at some level, the things we care about talking about in physics and science are things that are relevant to our sensory experience. If we were different from the way we are, you know, even, you know, you talk about, I don't know, a dog or something like this that has a very good sense of smell different from ours, the things,

Starting point is 00:24:12 you know, dog physics would no doubt be different from human physics. There's a bunch of things that dogs will be very concerned about that we barely notice and more extremely with other kinds of critters and so on. So the first thing to say is I think the science that we care about is science that somehow relates to our way of sensing things going on in the world. Now, to this question of whether there are certain kinds of things we can figure out, what is science ultimately doing? Science is going from the natural world over here. What is science trying to achieve? It's trying to go from the natural world, and it's trying to essentially have a translation, have a bridge between what actually happens in the natural world and the kind of narratives that we can kind of tell

Starting point is 00:24:58 ourselves in our finite minds. So there's all this stuff going on in nature, but science is about kind of how we can think about what's going on, how we can take all the stuff that's going on in nature and stuff some sort of filtered version of that into our finite minds and develop some narrative that allows us to make it kind of predictable and understandable what's happening in nature. So we're not getting all of nature. We're just getting this tiny little piece of nature. And I think then the question is, and that's, so then you can say, well, well, how do the LLMs relate to this? The LLMs are kind of built in our image. LLMs, the fundamental operation of neural nets, which is what LLMs are based on, is kind of a cartoon version of what happens.

Starting point is 00:25:46 happens in brain. We didn't know until very recently that that cartoon version was good enough to be able to do these impressive human-like things. Turns out that it is. And what we can expect the LLNs to do are many of the kinds of things that we do. Now, there are lots of things that we know can be done that we don't do. Like, pretty much nobody can run code in their minds. If you say, I've got this program, what does it do? It does a complicated thing. There's no way. that a human can sort of run that code in their mind. By the way, and LLM can't do that either. Right, right.

Starting point is 00:26:23 You make the case for the working memory. You make a distinction between working memory and processing speed, and even, as you say, even LLMs can't do that. They stumble on things like how many R's are in the word strawberry, right? Well, but they're doing the same kind of thing that we're doing, which is broad but shallow computation. The thing that is sort of a coincidence of history perhaps is that, you know, After Newton and Galileo and all those folk, we developed this kind of very formalized way of

Starting point is 00:26:52 thinking about the world that eventually led us to computation and computers and eventually led to this idea that we could actually do sort of a whole tower of formal operation and do that and build that up so that we could do these kind of irreducible computations that we can do with a machine that we can't really do with our brains. And so there's sort of these two different branches of how you can approach things. You can approach things by sort of just what you can think through with your own mind, and you can approach things by actually doing computations, doing kind of big towers of computation. And we can readily see that there are things that we can get to with big towers of computation. I've spent a large part of my life doing those

Starting point is 00:27:35 kinds of things. There are things we can get to with those big towers of computation that human minds just don't get to on their own. And that kind of tells us, yes, we are locked in a kind of prison, which is the prison of what human minds deal with. Those are the things that we have this sort of story of the fact that the things we care about are the things that human minds can kind of deal with. But there's a lot else out there in the computational universe in the Ruliad that is sort of behavior that human minds.

Starting point is 00:28:09 can't really wrap themselves around. I mean, the kind of story of my, my kind of day job life is building our computational language, Orton Language, and what is the point of that language? The point of that language is to make a bridge between the way humans think about things and what is computationally possible. So, in other words, there's much that is computationally possible that we humans just look at it and say, that looks alien. I don't know what's going on. We're, you know, In order to kind of address those kinds of things, the science of the Galileos and Newton's doesn't really do that. That kind of science and the mathematics that it's associated with doesn't get there. That's, you know, 20 years ago, I wrote this book that had a title, which at the time people were like, how can you say that?

Starting point is 00:29:01 The title was a new kind of science. And the reason I called it that is because that's what it was about. In other words, there had been a kind of science that people have been doing for 300 years, which was based on this kind of idea of using mathematics. And this is a kind of science that is different. It's based on sort of building these kind of formal towers of computation that go very far beyond what sort of unaided human minds can deal with. In the opposite end of the spectrum, we talked about supermassive.

Starting point is 00:29:35 We can talk about planetary-sized brains, but what about the opposite side? People talk about Boltzmann brains. And in the context of the Rulia, where all possible computations exist, do Boltzmann brains represent inevitable observers that will come into qualification as we would consider an observer, so to speak? Or do they fail to qualify as observers? Because they lack sustained computational and reducibility. Well, there aren't very many. When you want your brain to spontaneously form, from a bunch of random molecules bouncing around, that's a pretty rare thing to happen. It is a very interesting scientific question,

Starting point is 00:30:14 which I don't feel that I've answered yet, the extent to which observers like us are inevitable in the Ruliard. It is, you know, at the level of talking about Boltzmann brains and sort of all things are possible, yes, occasionally you'll just randomly get a thing that's a bit like us, but that's not, I think, enough to explain what we actually observe. And actually, it's interesting that the thinking about sort of the emergence of brain-like things leads you much more into biology than I had expected.

Starting point is 00:30:46 In other words, I'd always imagine, I think it becomes important that, for example, things like self-replication occur. Because one of the things to realize is if there was only one mind, things will be very different. The fact that we operate the way we do is partly a consequence of the fact that there are lots of minds that we're communicating with, lots of similar minds that we're communicating with. As an example, the fact that we sort of believe in objective... You said this place was steps from the water. We just haven't found the steps yet.

Starting point is 00:31:17 How much did we save? Enough. Enough to get lost. Or you could book a stay with Hilton. Welcome to your oceanfront room. Just steps from the water. The Hilton sale is on now. Book on Hilton.com or the Hilton app

Starting point is 00:31:33 and save up to 20% to get the stay. you expected. When you want savings, not surprises. It matters where you stay. Hilton, for the stay. Reality. The fact that we, as a consequence of the fact that we all kind of agree about, in some sense, about what's happening, because there are lots of us who are sort of similar enough that we kind of come to the same conclusion. If there was only one of us, it were very hard to, it's not even clear what we would mean by objective reality. It's if there's only one of us, that one of us is experiencing things, but they don't get to sort of say, well, this is the way everybody will experience this. This is the way it is. It's just all they knows is the way it is to them.

Starting point is 00:32:15 So sorry, you just made me think of Pascal, you know, and the cogito ergo sum. It's sort of, oh, Descartes, sorry, sorry, you just made me think of, I'm going to edit that out. You just made me think of Descartes, you know, cogito ergo sum in that, you know, the, the, the refutation perhaps of a single master simulator or maybe a god, you know, could be found in what you just said, which is, which is that these, it's sort of the interactions. Am I, am I overstating the case? I think the issue, so there's a bunch of different issues. I mean, one is, is there, all we know for sure, and I agree with Descartes on this, all we know for sure is what we internally experience. But we can make many inferences, guesses about what's going on from the fact

Starting point is 00:33:04 that we assume that the other people we see are like us. It is interesting. We've been thrust into this situation of having alien intelligences among us with the AIs. And the question of whether we perceive in the AI something similar enough to us that we imagine that it has the same kinds of experiences we have is an interesting one. I mean, if all you're doing is you're in a chat interface or something and you're talking to the thing and you might be talking to a human in a chat interface too, it's, you know, it very quickly comes to the point where you probably will have a theory of what's going on in that thing you're talking to that says, this is, this is like me. It must have the same kinds of inner experiences that I have, even though I can never know that for

Starting point is 00:33:50 sure. Right, the theory of mind. I never know that for sure. All I know for sure, is what's happening inside me, so to speak. Now, you know, there's a, there's a lot to say about kind of, well, here's an example of something, kind of, the fact that we kind of agree on objective experience is a consequence in many ways of the fact that we are sort of a, our minds, we're sort of a flock of minds that are in some sense very close together relative to the vast mass of what happens in the Ruliad. So in the Ruliad, many, many different kinds of things can happen. But it's because we're all sort of going through, we're all sort of existing in the Ruliad very nearby that we kind of agree about

Starting point is 00:34:33 what's going on. We agree that, I don't know, I'm looking and seeing whether the moon is out today. But we would agree about what the state of the moon is because, okay, we're a few thousand miles apart, but it's sort of close enough that there's a moon in the sky. We're not, you know, on opposite sides of the galaxy where the sky is completely different. So in other words, we agree on objective reality because we are, in a sense, close together in the Rulia ad, and that's something where it's important that there are, there's sort of many different entities that can have that communication to be able to form that sort of objective conclusion. Now, I think you're asking about sort of the God versus the humans and so on.

Starting point is 00:35:23 I kind of feel like a lot of kind of the, to me, things like the existence of the universe, which is something that you might say, well, you know, the fact that the universe exists, you might associate with there has to be some prime mover, some God that's making the universe exist. In my way of thinking about things, you don't need that. The Ruliad is something that is a formally inevitable thing. It's just like saying one plus one equals two. You don't have to have an actual one rock, another rock, and put them together and make two rocks. One plus one equals two is an abstract thing to talk about. The Ruliad is also an abstract thing to talk about. So the existence of the Rulia is not something that you have to debate. What is non-trivial is that there are

Starting point is 00:36:07 observers like us within the Ruriyadh. And that is something which potentially is amenable to sort of scientific investigation. In other words, how inevitable are observers like us? How common are observers like us? If you ask for observers like us but a bit different from us, how far away? What is the nearest thing that is kind of somewhat like us but not us? That's kind of an extraterrestrial intelligence or across the Rulia, intelligence, alien intelligence kind of question.

Starting point is 00:36:39 But I think, so, you know, I think in some sense, the sort of, The existence of the universe is an inevitability. What is not so obvious is that we are here to perceive it, and that the universe as we perceive it is something that anything is perceiving, so to speak. It could be that the universe is just doing its thing, and nobody is there to kind of sense it in the way that we do and to conclude that there are laws of physics of the kind that we conclude there are. Interesting.

Starting point is 00:37:13 So if the Riliod contains all possible computations, including every brain state of a conscious observer, is free will sort of the subjective, you know, effect of these reducible branches or is an illusion as many of my guests, Stephen, I get very frustrated. I've never met a person who acts like they don't have free will. And yet so many of my guests from Sopulski to Dawkins to Dennett and some and to Sam Harris, in particular, they claim free will is a complete illusion. Sabina Hassenfeld or another one. Where does free will come in to the ruleyad in our brain states or is it illusory? So it is a very interesting thing that we assume about ourselves that we have free will. Even the doing of science requires an assumption of free will.

Starting point is 00:38:01 The idea that we can do experiments, the idea that we can pick the experiment we do and come out with a conclusion, requires an assumption of free will. We don't have free will. No experiment, we can't do an arbitrary experiment. We are determined in what experiments we do and we're forced to do just those experiments. So we kind of assume, and we in our thinking about science, that we have some kind of free will. So how does that work? Well, I think I've sort of had an understanding of this for a long time now.

Starting point is 00:38:27 And it's all related to this concept of computational irreducibility that I've been talking about for 40 years now. So the starting point there is you might imagine that if you write down particular rules, let's say a particular program, that you would always be able to foresee what that program will do. And in some sense, you can, because you can just run the program step by step and see what it does. But the thing that has been the kind of the conceit of science, since your friends, Galileo and Newton, has been, okay, if we know the rules, we solve the problem. We can just jump ahead and see everything about what will happen. That's what one who'd assumed was the case.

Starting point is 00:39:08 That's what mathematical science tends to encourage what. on to do. That's kind of what, when people talk about predictability is sort of a key aspect of science, that's what people are focusing on. But it isn't always the case. What can happen is that you have an irreducible computation where you know every step that you can follow, but to work out what will happen after, let's say, a billion steps. You basically have to just run those billion steps and see what happens. There's no way to jump ahead. It's all related to these questions about, I talk about this notion of computational equivalence, the notion that if it is the case that these things, even these very simple rules that one might write down, are no less computationally capable

Starting point is 00:39:52 than our minds or our computers or our mathematics, then there's no way that we'll be able to win out over those simple programs. Those simple programs will be, we'll be stuck just keeping up with those programs as they run. So how does this relate to free will? Well, the, The point is, even if you have deterministic underlying rules, you can't know what's going to happen, except by running those rules and seeing what happens. So if you were looking at some preacher, you know, the moth that's, you know, trying to bash itself at the window, trying to get to the light, that doesn't look like it has free will. It looks like it's just following some very simple program.

Starting point is 00:40:31 It just keeps on doing the same thing. But the thing that I think is where you start thinking about free will is you can't work out what the thing will do any more efficiently than just by watching it do it in and seeing what happens. In other words, it's not the case that you can say, no, you don't have free will. I can tell you don't have free will because I know what you're going to do. What's happening is you can't know what the thing is going to do. The thing is just going to do what it does and you are merely a passenger watching what it does. And so that's the sense in which, you know, you can say, well, looked at from the outside, you can, you could say, well, I know what it's going to do. You can, you could say,

Starting point is 00:41:09 well, I know what it's going to do because I know it's rules. Right. Like the weather. Like the weather. We know the rules of weather, but we can't predict, you know, because it's a complex system, correct? I mean, we can't predict, you know, mere moments ahead if you took it the entirety of the system, even though we know basic principles of climate.

Starting point is 00:41:27 Is that a fair comparison to? That's a complicated, that's a complicated case that has a whole bunch of other tentacles associated with it. But the basic idea is that for something to be not. have free will. It is because it is, it is, it's, we know what it's going to do. It is not, it is not acting freely in the sense that it is doing things that are just intrinsic to it. It's something where we can, from the outside, we can say, we know what it's going to do. So I think, you know, I think this really is, is the story of kind of our perception of free will. The universe is doing

Starting point is 00:42:06 what it does, we are part of that universe. You can't jump ahead and say, we know what we're going to do. We're just doing what we do. And so that's why we are, we have both the perception and in a sense the reality of free will. There is no way to know what, what we're going to do other than by just running us and seeing what we do. I think that's, that's the, and you know, this becomes a very practical issue for, you know, for AIs, for example, do AIs have free will? In other words, is it the case that, and, okay, so that's important if you want to sort of make sure the AIs do the right things. You have to say, well, I'm going to put these, I'm going to constrain the AI. I'm going to set the AI up so that it can only do the right things. The problem is if the AI has free will,

Starting point is 00:42:54 then you can know, you can never say for sure based on sort of what you put into the AI, what it's going to do. It will have, it will be doing things which are unexpected and surprising. And the fact is that if you have an AI that is capable of doing arbitrary computation, if you're not constraining it to be sort of this, this impoverished version of computation, then it becomes inevitable that the AI in this same sense has free will. It is, it is doing irreducible computation where you can't know what it's going to do except by running it and seeing what happens. So that's a, it's kind of a, I think it's a, it's an important kind of almost societal issue for the future is, do we have AIs that can do these computationally sophisticated things that are probably pretty useful to us, but they will

Starting point is 00:43:42 sometimes do things that are unexpected and surprising and not what we want? Or do we say, no, we want our AIs to be constrained to only work in ways that we can readily understand and predict, in which case the AI won't be able to do as much for us. So that's a, so, so, a very practical manifestation of computational reducibility. I mean, it's sort of remarkable to me. I invented this idea 40 years ago, and I invented it as a way of understanding sort of what was possible to do in science and what wasn't. And it's ended up getting so many tentacles. I mean, like the idea that's central to blockchain of, you know, proof of work in Bitcoin and so on, is a computational irreducibility idea. I never would have imagined that this idea that was

Starting point is 00:44:24 something associated with kind of a limitation of science and a theoretical understanding of limitation of science would in not too many decades get used to kind of be the proof of value that was burning huge amounts of energy and so on in all sorts of efforts of things like Bitcoin mining and such. It's sort of interesting to me that these very conceptual sort of almost philosophical things that one comes up with in science can end up being sort of becoming very, very practical and real. And so this question of free will that you raised, which might seem like a philosophical question, it's a very much something that is a very practical question when it comes

Starting point is 00:45:07 to AIs. I mean, there are all kinds of issues that boil down to questions about, you know, just how much free will do we want to give the AIs, so to speak. Right. And will we turn them off or will they cause them pain or will they cause us pain? and so on. These are ethical questions that many people wouldn't have dreamed of. I just, again, find it interesting that Sam Harris does believe that AI can have free will, but he doesn't believe humans can have free will. And to the extent that we model, as we as you said earlier, we're sort of looking at these as extensions of what we are familiar with, which kind of for me makes me want to ask the question of, you know, are these AIs training us? We've heard, you know, you'll hear stories in the news and provocative headlines and stuff of AIs, you know, convincing people, you know, to leave their wives or whatever and not, you know,

Starting point is 00:46:04 things that I don't find very important. But these questions of, you know, we are prompting them and we seem to think that we're in complete control. But is it not possible, Stephen, that they could be prompting us in a certain sense? I mean, will, is there a threshold? We think about neural networks. What is that threat? What is it based on hardware, software, both?

Starting point is 00:46:24 Well, I mean, look, we humans can be kind of lazy. You know, it's kind of like there was a time when you would read maps to try and figure out where you would go in your car. Most of us, and I started doing that very early on, much to the amusement of my children at the time, just because I'm just going to follow what the GPS says. I'm not going to think about it. And, you know, we can imagine that more and more, there's sort of an auto suggest for life, so to speak. Yeah. And, you know, people just, well, do what the auto suggest says. That's the sense in which the AIs take over.

Starting point is 00:46:56 It's not that the AIs are rounding us all up with autonomous weapons or something. It's more just that the humans get kind of lazy and they just follow what the AIs tell them to do. Now, it's sort of an interesting question. What then do the AIs lead the humans to do? What's happened with sort of the current round of AI is that we've trained those AIs from a trillion in words of what we humans have written, so to speak. And what the AIs have got out of that is kind of the average of what humans have said. And so there's a great tendency to say, well, we're kind of going to, everything's going to be average. Everything's going to be,

Starting point is 00:47:38 you know, the AI is going to say, the AI is going to say a definite thing. The AI is going to say things that are kind of in some sense the average of what we humans say. Now, we can certainly imagine setting things up so that the AI is trying to not what's done right now. but one can imagine something where the AI is trying to be kind of an awkward AI that is trying to say the things that are like the unexpected, controversial kind of corners of what we humans have said. Not the way that things are set up right now, but it's something one could sort of imagine, but the fact is that sort of the tendency is sort of to be circling around sort of the average that the AI has learned from all the things we produced. Now, it's worth saying that

Starting point is 00:48:22 The way you, one way you break out of that is you start doing computations. What comes from just sort of averaging all the things we humans have written is something that is just sort of statically there. When we compute things, it is easy for us to compute things that have never been computed before, to go out into the Ruliad, basically, and find things that we can, we can just sort of pick at random, what direction we go, and we'll go to places that have never been visited before. Is there an optimum way to, is there an optimum way to call,

Starting point is 00:48:52 colonize the Rulian, Stephen? Well, what we've been doing in a sense as we develop sort of in intellectual history, I see as being a progressive colonization of the Ruliat. I mean, that is you can think about any mind. Studies and play. Come together on a Windows 11 PC. And for a limited time, college students get the best of both worlds. Get the unreal college deal, everything you need, to study and play with select Windows 11 PCs. Eligible students get a year of Microsoft 365 premium, a year of Xbox GamePass Ultimate with a custom color Xbox wireless controller. Learn more at Windows.com slash student offer. While supplies last, ends June 30th, terms at AKA.m.m.m.

Starting point is 00:49:33 With any sort of paradigm for thinking about things is lives at a point in the Ruliaad, a place in the Ruliaad. As we sort of expand our domain of thinking, as we get more paradigms for thinking about things, we're colonizing the Rurliad, much like we get different points of view about the universe by sending spacecraft out further and further to kind of explore what different points of view on the universe. In Rulial space, the development of paradigms is kind of the successive expansion of the Rulia. Now, you know, what directions do we go? Well, that's what we societally tend to choose. We say, we're going to colonize in this direction. We're going to develop our paradigms in this direction. These are the kinds of things that we're going to understand. These are the kinds of things

Starting point is 00:50:17 that are going to turn into words in our languages, where those words in our languages are things where we can exchange those words between us, and we all sort of have collective agreement about what those words mean. So in a sense, there is a sort of a big expansion that we can make in the Ruriad to many things that seem completely alien to us.

Starting point is 00:50:36 With our finite minds, we tend to go only in particular paths out into the Ruliard, where we are exploring things that we can kind of still think about in our minds. And that's, you know, there are many contingencies, there are many possibilities of how sort of human intellectual history could have developed. We've picked particular ones. Now, it's a fair question, are there better ones versus worse ones to pick? I don't know the answer to that fully.

Starting point is 00:51:05 I've studied that a bit in the case of mathematics. When we develop mathematics, there are maybe three or four million theorems that human mathematicians have written down in the history of human mathematics, but they're an infinite number of possible theorems of mathematics. And there's a question of how do we explore those infinite number of possible theorems of mathematics, which by the way, are we can think of as being elements of the Ruliad. So how do we, you know, what are the paths we could follow? Are there paths that are somehow more, more productive than other paths? Well, in the end, I think it depends a lot on what end point you're looking for. So in other words, there are things where we can say, this is the direction we could go, but the place we end up

Starting point is 00:51:44 with is something that is very unfamiliar to us humans. Perhaps in the case of more physics-related things, perhaps is something that does not map well onto the biological senses that we happen to have. So you speak about rural particles, which to me, you know, is inescapable to not think about the other way that you can colonize the Rulia quickly is to have, you know, massless particles, right? So you can either shorten the distances or you can speed up the rocket ships or the are the particle. And first, can you please explain Rulial particles and what those are in the context of the communication that we are engaging with and so forth? But also,

Starting point is 00:52:24 are there more efficient ways, particulate forms of Rulio particles that have massless properties that travel at some ultimate speed limit? It does C apply to thinking? That's a good and interesting question and complicated. So first thing to say is I think our perception of the universe critically depends on the fact that we're not massless. In other words, if we were photons, massless particles, time would not elapse for us. And our perception of the photons that you collect, cruelly collect in your detector, those photons, the last thing they knew was when they were emitted at recombination time, you know, 100,000 years after the beginning of the universe.

Starting point is 00:53:07 And the next thing they know is splat. They ran into, you know, Brian. detector of those photons. There was no, they had no experience between those two things. They led to their minds very short lives from the beginning of the universe to smashing into your detector. But so it is, our perception of the universe critically depends on the fact that we are not massless particles. We have an experience of time that, you know, our experience of the universe is an experience

Starting point is 00:53:41 of the progression of time. I mean, the progression of time we can think of as being kind of the universe doing computation. We're part of the doing of that computation. And it's for us to experience that we can't operate like massless particles. Now, in terms of kind of this idea of rural particles, that's the kind of the point there is you're at different places in physical space. How do you communicate across physical space? Well, you need a you need, even the possibility of communication requires the idea of motion. It has to be the case that a thing can go from one place in the universe to another and still be the same thing. This is something people worried about in antiquity, how that works. By the time of, for example,

Starting point is 00:54:29 Galileo, Galileo sort of just said, well, you know, motion is what it is and then tried to analyze how motion works. The fact that motion is possible is not obvious. Even in traditional physics, if you're close enough to a space-time singularity, any material object, you know, your spacecraft or whatever else, will be distorted beyond recognition when you're close to that spacetime singularity. But most of the time, we can think of pure motion as being a thing. Now, the question is, well, what is ultimately the carrier of pure motion? What is the thing that we can sort of move around the universe without changing? And the answer to that is basically particles, like electrons and photons and things like that. Those are the

Starting point is 00:55:11 in a sense, the elementary carriers of pure motion, the things that move around the universe without changing. In our models, electrons are made of atoms of space, and the atoms of space of which an electron is made will change as the electron moves. It's like if you have an eddy in water or something, the molecules that make up that eddy will change as the eddy moves, but yet the eddy preserves its identity. So that's the sense in which sort of particles are things that preserve their identity under motion. And so then the question is, well, we have in rural space, we have all these minds in different places in rural space. And the question is, what is it that can be transported from one mind to another? So at a very practical level, you know, in my mind, there are sort of neuron firings

Starting point is 00:55:56 going on, and I'm thinking of some concept, I'm thinking of a cat or something like this, and that corresponds to some pattern of neuron firings in my mind. In your mind, the way that you represent the concept of a cat, where we are a completely different set of neuron firing. So the question is, how do we sort of get, what is it that we can move from my mind to yours that will communicate this concept of a cat that has to be packaged up and unpacked at the other end in different forms? And the thing that this is, we're kind of moving from one mind to another across rural space. We're trying to move something from one mind to another.

Starting point is 00:56:33 The thing we try to move that is sort of packaged up is what we imagine, what we think of as concepts. concepts, where a concept is something which is sort of the packaging of all those neuron firings into a robust thing that we will often describe with a word in human language and that then can be unpacked by another mind. So if you're asking, I mean, to me, it's sort of a remarkable analogy between things like particles like electrons and so on and the notion of concepts that are transportable from one mind to another. But now you're asking me to go further than that and to talk about what would it mean if there were sort of things concepts that were like massless particles. And I suppose in a sense, what one would imagine, okay, so this is a, it's like

Starting point is 00:57:25 what happens to that poor photon that is emitted and then a moment later for it, it smashes into your detector. It's, I think, the typical sort of particle that exists in the universe, which has mass. What is mass? Well, mass, I think, okay, so in the traditional modern theories of physics and the standard model and so on, what gives particles mass is their continual interaction with the condensate of the Higgs field. So the kind of thing, which is always, I've always thought it was a bit of a pluge is the idea that particles get mass because throughout the universe, there's this field that exists. And as a particle moves, it's constantly interacting with that field. It's constantly being kind of kicked by the presence of the ether view of it. That's the ether, almost the ether.

Starting point is 00:58:17 Well, it is, yes, but we have an ether again. It's a hink's condensate. Okay? So in our models, it works differently from that because space in the, in sort of physics, as it has been worked on the last 100 years or so, there's the idea that space is just this background thing. And so to have something that particles have to sort of put effort into to getting through, you have to introduce something like the Higgs field where you're saying, and the particles keep on interacting with this background field. The vacuum in usual physics is just like, well, the particles just go through the vacuum.

Starting point is 00:58:53 Nothing much happens. Well, in our models, the vacuum is space has to be made. space is not something that just exists as a background. Space is constructed out of sort of the aggregate effect of all of these sort of processes that are going on in this network that we ultimately think of as being space. And so when we think about a particle sort of moving, what we're thinking about is that particle is being recreated out of different atoms of space at every successive moment of time. And so this idea that when we have a massive particle, we're thinking about that as being part of the sort of the recreation of the particle as it goes through space.

Starting point is 00:59:38 So an analogy. If you have a piece of glass, for example, and you shine a light through it, it is not the case that the photons just go straight through the piece of glass. Instead, what's happening is the photons are being absorbed by atoms in the glass. and then a moment later, another photon is being re-emitted that goes on and goes on going through the glass. That's why light goes one and a half times slower in glass than it does in a vacuum is because what's happening is the photons keep on getting absorbed, then there's a bit of a delay, then they get re-emitted again. And that's kind of the process that's going on.

Starting point is 01:00:15 And you can think of that as being sort of very qualitatively similar to what we imagine is happening for massive particles in the actual structure of space in our universe. But I'm not answering your question of what. So I think my best sort of immediate analogy there would be that would be to say that I think, and I don't know whether the analogy can be stretched this far, but I think that sort of the massless concepts are the ones where there is sort of inevitably no difference of interpretation between the emitting mind and the receiving mind. I'm not sure if that's right, but that's what I would. That would be, so it's kind of like when you have a

Starting point is 01:00:58 massive particle, things happen to it on its way from the emitter to the receiver. Whereas when you have a massless particle, things aren't happening to it. That's why no time has passed for it. There's nothing happened to it on its way from the, from the emitter to the receiver. But I'm not sure. It's a good question. And I, you know, it's always one of the things that's always difficult. I'm sort of reminded of my long ago friend Dick Feynman, who was always very big on the kind of intuitive explanations of things. And that's what people heard from him is the intuitive explanation of everything. But behind the scenes, he was a really good calculator. Yamava Resort and Casino at San Manuel is California's number one entertainment destination for today's superstars. Catch the Jonas Brothers

Starting point is 01:01:45 return to the Yamava Theater stage on April 30th. The powerful Volta. of Demi Lovato on May 17th and the signature Southern Country Rock of Eric Church on July 19th. Tickets on sale now at Yamava Theater.com, only at Yamava Resort and Casino, celebrating its 40th anniversary. You in? Must be 21 to enter. And so he would do all these calculations and work out, you know, this is how it had to work. And then he thought, oh, the calculations are easy. Nobody wants to see that stuff. I'm just going to tell them this intuitive explanation. But he knew the intuitive explanation was right, because he did this calculation underneath.

Starting point is 01:02:23 That's right. Sometimes you can kind of jump from tree to tree, so to speak, with pure intuitive explanations. It's complicated. Sometimes I think when one gets good at it, you can do a lot of tree to tree jumping, so to speak. In some sense, there's also a certain grounding. I mean, for me, for Dick Feynman,

Starting point is 01:02:47 that grounding was always mathematical calculations. done by hand or pieces of paper. For me, it's computer experiments. I mean, that's kind of the way that I ground my thinking about things, is, you know, you do these experiments, and one of the things about these experiments, which is related to the whole computational irreducibility story and so on, is that very, very often those experiments reveal things

Starting point is 01:03:10 that are not what one's intuition expected would happen. In other words, there's a lot of surprises out there in the computational universe. You know, that's a thing that I'm continually, you know, it's a continual sort of humbling experience that sort of practically every week, I'll have a thing that I'm studying, and it's like, I'm sure it's going to do this. But I wouldn't have bothered to study it unless I had some idea of what it was going to do. But then when I actually run it, actually see the results of the experiments, it's, no, actually, it managed to do this other thing in a very clever way that I never imagined.

Starting point is 01:03:43 It's like the free will of the equations. I mean, that's masterful. So, Stephen, I have a couple minutes left. But I want to ask you something. Forgive me if it's too pedantic. But I kind of think of the Ruliat as sort of the ultimate Amazon web service, you know, cloud provider or something that's running these computational processes. And each instance, you know, node is a physical rule and so forth. But if I were to analogize, make that analogy that, you know, you could almost think of it as all these clients, you know, in this distributed cloud that, you know, is abstract in some sense.

Starting point is 01:04:18 what would you, given what you just said about surprise and the experiments, if you could, you know, force to really add to do something or rent out one very dense portion of it to solve one physical problem, one problem in math or physics, what would be the problem that you'd most like to direct the computational firepower, the biggest computational firepower in the universe, perhaps the whole universe, what would you direct it to? What is the most important problem to you right now? Well, that's a hypothetical that's hard to sort of imagine executing on. I mean, I think that the, what's the most important problem? I mean, it's, again, there are things I'd like to know the answer to.

Starting point is 01:04:58 It's, there are plenty of things, you know, we humans have a certain sort of human experience in existence, and it's, you know, it'd be fun, you know, if we talk about what do we want to solve? Like, human immortality, that would be a fine thing to solve. Yeah. Now, I've actually sort of looked at that a bit from a computational point of view, what's actually involved? If you, you know, what is the foundational thing that's happening in life, in biology and so on? It's again, a big computational irreducibility story. And unfortunately, it makes things look pretty difficult because it's like when you're, if you're thinking about sort of medical kinds of things, you say, okay, we've got the system. It was sort of a piece of computational irreducibility that was adapted to the things. that we humans do and so on. And now let's say you perturb it, you poke it in some way. Well, it does these unpredictable things. And then you say, well, what's the fundamental problem of medicine? It's to, after the thing got poked and started doing the things you didn't expect it to do,

Starting point is 01:06:00 can you poke it again and get it back on track? And the whole sort of computational reducibility story tells you just how hard it is to do that. It's, again, it's a little bit related to the whole free will question also. If you want to have a rich life, so to speak, that does many things, it's hard to say, let's have something that will kind of persist forever, because there are things, I mean, we can easily make cell cultures that are kind of tumor cells or something that will persist forever, but we don't think that's a great way to live, so to speak. For us humans, the most obvious sort of big thing would be kind of how do we take, and it's not obvious how that is even sort of conceivable, you know, how do we take the kind of experience that

Starting point is 01:06:46 we have and if you like what's going on, which somebody like me tends to like, it's kind of you like your life, you're living, you kind of want to make that go on as long as possible. And so that, I mean, that's a, that's an example of a kind of where you could, but I, you know, as I said, I think it is a fundamentally computationally difficult problem. I don't think it's kind of these things where it's like, I mean, if you look at sort of what we are constructed out of, we are sort of the sole example of a successful molecular scale computational system. Life is a kind of molecular scale computational system. And it's hard to kind of the things that we do that are our current sort of approaches to medicine

Starting point is 01:07:33 and things like that tend to be these very coarse things where you say, just send in that molecule everywhere and hope that it binds to the right things and so on. I mean, we have within ourselves, we have things like the immune system that's a bit more sophisticated at going in and sort of tweaking what's going on. But we're still, we're pretty far away from that. But it's not even clear sort of at a very almost philosophical level if you sort of achieve immortality, but you achieve it. Again, it's kind of like when people say, can we adapt AI to do some particular thing,

Starting point is 01:08:07 You say, well, you adapt AI, you know, the classic is to make as many paper cups as possible, and then it turns the whole, you know, surface of the planet into a paper cup factory or whatever. And, you know, the same thing. You want the biological thing that kind of can continue forever. As I say, it's easy to get that, but it's not a good life type thing. That's right. So, I mean, that's that will be an example. I mean, I think that the, you know, there are things, it's always complicated.

Starting point is 01:08:35 When you ask for things that, let's say you say, well, I'd really like to know the fundamental theory of the universe, which I've worked on a lot. But that's a complicated thing. What do you really mean by that? With the Ruliad, we have in a sense, what I'm pretty sure is the fundamental theory. But making the connection between that and what we humans are kind of what we humans can talk about, can perceive and something. on, that's a different thing, which seems like it's not quite find the fundamental theory. I mean, in terms of things that I'm sort of chasing right now, you know, I'm very interested in sort of experimental implications of the sort of fundamental theory of physics that we have. I've been having fun using LLMs as my kind of collaborators in trying to study that.

Starting point is 01:09:28 You know, there's a million papers that have been published about physics. And the question is, are there effects that have already been known, that have been known for 30 years, 50 years, whatever, which when correctly interpreted, one realizes, oh, that's actually a sign of this thing that our models predict. You know, the cautionary tale is Brownian motion, which was, you know, people wondered throughout the 19th century, do molecules exist? And nobody knew that the fact is, in 1827, Robert Brown, a botanist, had observed. These little pollen grains getting kicked discreetly by water molecules. We didn't know that's what was going on, but they just observed this is a weird effect. It took until 1900, basically, for people to say that thing that was discovered 70 years ago actually shows that molecules exist. So one of the challenges right now is, is there something that's sitting out there in the existing literature of physics?

Starting point is 01:10:23 You don't even have to spend $100 million to do a new experiment. Is there something that's already there where you just say, wait a minute, if interpreted people, correctly, that's a sign. And it might be an experiment where people say, that was a really weird experiment. It produced a result we don't even believe because it's so weird. We don't have a theory to sort of back it up. But that's one very practical little application that I've been interested in for all the months. Very nice. Well, I do want to wrap up on that note, but I do want to say just one final question, which is, again, this is wonderful article. your writing is every bit, the equal of your mathematics, which is the highest compliment, I think

Starting point is 01:11:06 I can give to a guest. It's beautiful. It's poetic. There's so many just incredibly stimulating aspects of this article that will link to in the show notes down below. But the final question is how you end it. You know, you talk about kind of a Rulietz-spanning mind that can take care of all possible computations, but as with the mind of a human, as I get older, maybe I can do one thing in working memory, as you discuss, not even three or four, but if it could have sort of this incorporation of all possible computations, but would it not lose coherency of what we experience, sort of where we started off? And the question is, can we know if AI systems are really minds like ours, or are they really truly, as you maybe are hinting at, a type of alien that we

Starting point is 01:11:59 have on Earth, that we maybe created ourselves? But are they minds like ours? Are they something completely different? Well, it's easy to make an AI mind that's not like ours. You know, lots of things I've spent time doing of sort of exploring the computational universe, sort of running these arbitrary programs and so on. That makes, that does computation. It does kind of in some sense, mind-like stuff, but it's very alien to us. So it's easy to make minds that are very alien. It's easy to, the challenge is to make, and that's sort of one of the achievements of modern AI,

Starting point is 01:12:41 is to make computational minds that are aligned enough with us, that we think what they're doing is worth doing, so to speak. And that's kind of, you're asking what, sort of, what happens as we make those, you know, those AI minds kind of broader and broader? Do they, you know, and the answer is it's easy to make a broad AI mind and it's easy to make an AI mind that doesn't have the kind of coherent existence that we have. It's, you know, it is the case that the neural nets that we're making these days are built in our image and do kind of human mind-like thing. But as I say, it's easy to get something which behaves, more like nature behaves.

Starting point is 01:13:27 Nature doesn't have this kind of, this sort of filtering down to this kind of thread of consciousness of the kind of that we have. Nature just does all these different things. And that's an easy thing for us to be doing computationally. And we'll look at that and we'll say, well, that kind of looks like nature. It doesn't really look like a human mind like thing. Stephen, this has expanded my mind. My mom told me she named me Brian so that sometimes people would confuse me with brain. And it does occur quite frequently.

Starting point is 01:14:00 But you are always wrinkling the brain in new and interesting ways. I just can't wait to see what comes next, whether it's a book or another article. We'll link to your blog. I really just appreciate you spending so much of your time with me, especially so late at night. You were meant to be an astronomer. I think you missed your calling. You have this propensity to stay up all night. Stephen, I really appreciate you.

Starting point is 01:14:22 Thank you so much. And do let us know. We haven't met up in about eight years by my reckoning. So maybe we'll get to me. I feel like I've met you because I see you so much on video. We met at a museum of math and so far. I know we met in person. Yeah.

Starting point is 01:14:37 My test is always, I know roughly how tall this person is. If the answer is yes, then we met in person. If the answer is I haven't the slightest idea, then we probably never met in person. But we definitely met. And I look forward to meeting again. Me too, Stephen. Thank you so much.

Starting point is 01:14:52 If this episode blew your mind, go watch my last interview from 2024 with Stephen, where we unpacked the really out from the ground up. We uncovered the reality of time. And he even helped me understand the fundamental basics of how he views the second law of thermodynamics. You won't see physics or thought the same way ever again. Click here. And don't forget to subscribe. Ambition comes in all shapes and sizes.

Starting point is 01:15:17 At First Citizens Bank, we roll with your goals because we're built for what you're building. Fit for your ambition for Citizens Bank. Some follow the noise. Bloomberg follows the money, whether it's the funds fueling AI or crypto's trillion-dollar swings. There's a money side to every story. Get the money side of the story. Subscribe now at Bloomberg.com.

Into the Impossible With Brian Keating - Are Humans Smart Enough to Understand the Universe? (ft. Stephen Wolfram)

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.