#27 - Sean Carroll - The Enigma of Complexity

27:02You know, every bell curve starts out as an exponential. Sure. The fact that you're on an exponential right now means almost nothing, because extrapolation is super-duper hard. But the more general question, you know, you have to separate it into sort of metaphysical certainty kinds of questions. Are we absolutely certain of this or that? And the answer is always no. We're not certain. We can't. You know, in the world of physics, in the world of empirically understanding the world, there could always be a surprising result tomorrow that changes our minds. But then there is a slightly easier question about, given what we know about the laws of physics, can you do this, right?

27:39If you have a discussion of high-speed rail, you're not worried about the speed of light being a limit. But it is there, and you can't say, well, we're getting better and better at building high-speed rail. If we extrapolate it a million years in the future, we're going to hit the speed of light. That's just not right. Same thing is true with the entropy of the universe. It doesn't matter what configuration of stuff you have. it's still relying on free energy in the universe, and that amount is finite. So there's no way of being clever to get around it. The laws of physics are involved. So we're as certain about the heat death being a good model of what's happening in the future, in your view, as we are about probably equivalently the Big Bang as the beginning.

28:21Or how do they stack up against each other? Where would you, like, certainty level of theory is standing the test of time? That's a very good point, because, of course, we don't know. the future is hard to predict. When I'm talking about the heat death of the universe, etc., I'm presuming a certain very, very natural set of assumptions about the future evolution of the universe that might not be right. So I put much less credence on that than you will never travel backward in time or faster than the speed of light. Those are much more solid statements.

28:57So the statements I'm making are conditional on our best current theory of the future. I think it's the way to plan, but until something better comes along. So when I talk about the Big Bang theory, the speed of light, traveling backward in time, my credences are 99.999, etc. When I talk about the future heat death of the universe, maybe 90 % is a better credence. When we're talking about space-time emerging from quantum entanglement, who knows? But probably, if I'm careful, it would be less than 50%. I think it's a very promising way forward, but we just don't know. 10 %er? Yeah, that's not...

29:33I mean, as an order of magnitude guess, I would imagine, but... As poker players, you're like, that happens all the time. Exactly. Yeah, yeah. That's pretty good. That's the most hopeful thing I've heard. A long time. I'm thrilled. Then also one more thing on the heat death, though, because now let's dig into those 10%. So how should I think about, because quantum fluctuations would still occur after the heat death, right? And like, yes and no. Okay, so maybe. And the heat death lasts for a very long time. I don't know if time actually still makes sense as a way to concept at the time. but yeah can sufficient amount of quantum particles appear to form something new is a question right yeah it might not be that we can interact with it because at that point sorry you're gone it's it was the heat by definition yeah yeah so we don't know for sure i suspect no even if yes it'll be super duper far in the future so uh but this is worth saying something about we haven't talked too much directly about quantum mechanics yet.

30:37I don't know if it's going to be. We're going to get into it. Let me say one particular thing about quantum mechanics. The phrase quantum fluctuations is fraught. It's a dangerous one. It's not that it's wrong, but it gives you exactly the wrong impression because it gives you the impression that you close your eyes and you see all these things popping in and out of existence. That's exactly wrong. If you have a quantum mechanical system and you look at it and you measure it, then you will get measurement outcomes that seem to jitter around, okay? Which makes you think that when you're not looking at it, it's jittering around.

31:10But all the rules of quantum mechanics say no. There's a fundamental difference between what the system does when you're not looking at it and what it does when you measure it. So something like empty space, as far as we understand it, is stationary, is static. There's nothing happening. So there's no probability per unit time that something wild is going to occur. If you were to probe it over and over again, you would get different measurement outcomes, but there's no one around to probe it. So fluctuations are a result of us interacting with, or anything interacting with a... The appearance of fluctuations is exactly that.

31:51It makes sense that if you interact with it, you poke it, something should occur with the system. It does make sense, but for the wrong reasons. Because in classical mechanics, if quantum mechanics had not been right, it would still make sense that if you poke things, you might get different answers. But in that theory, you can poke things gently. You can poke things as gently as you want and disturb them as little as you want. And quantum mechanics is fundamentally different. Even the gentlest of pokes can really give you a very different measurement outcome than you expected in quantum mechanics.

32:25So there is something fundamentally new there. Why is that? That there is, it's always going to create a fundamental. This is the quantum measurement problem. This is what is very hard. So different people give you different answers as to the why is that question, depending on their attitude towards what quantum mechanics really is. If you're a many worlds person, such as myself, this conversation we've already had about entropy increasing, etc. is super relevant here. because what happens is you, the observer, and the system you're looking at, what you call a measurement, involves you becoming entangled with that system.

33:02So if the system is just an electron that could be spinning clockwise or counterclockwise, or according to the rules of quantum mechanics, it can be in a superposition. So it's not fluctuating back and forth. It really is both, spinning clockwise and counterclockwise. But when you and I measure it, we only ever see spinning clockwise or counterclockwise full stop. We never see that superposition. And the many-worlds story is that you have become entangled with the spin, and there is now part of the wave function of the universe saying the spin was clockwise and you saw it be clockwise. And there's a whole other part of the wave function of the universe saying the spin was counterclockwise and that's also what you saw.

33:41And these two parts are never going to communicate with each other, never going to interact, never going to influence each other. Those are the separate worlds that we talk about in many worlds. Why are you more subscribed to the many worlds interpretation as opposed to the perhaps more popular or at least more common interpretation, which is the Copenhagen, which is that when an observer interacts with a quantum system, the wave function essentially collapses. And the superposition, it like chooses a path and that's it. And the other potential literally ceases to exist. Why do you fall in the many worlds?

34:20Right. That's a great question, perfectly fair. And two things are going on. One is that the Copenhagen interpretation, which you just described very well, is not a good theory. By a good theory, I don't mean one likely to be right. It just doesn't make sense. It's not written down. It's not full and complete. There are questions you can ask that it doesn't give an answer to. When does a measurement happen? What counts as a measurement? Not to mention that it's grossly ugly, right? It's like literally saying that quantum systems behave in one way when you don't measure them, and they behave in a very different way when you do measure them, and I'm never going to tell you what a measurement is.

34:57That's not okay. That's not a viable thing to be the most fundamental theory of nature. So I'm not really even worried about Copenhagen being right. It doesn't have a chance of being right. But there are many alternatives. There's not just many worlds. There's hidden variable theories. There's epistemic theories. There's objective collapse theories. So there's a good set of perfectly viable alternatives. The reason why many worlds is my favorite is because it's the simplest, honestly. It just says that thing that you thought was true about quantum systems evolving differently when you're looking at them and not looking at them, that was a mistake.

35:37All quantum systems always evolve the same way, according to the Schrodinger equation. What you left out, the mistake you made, not you live personally, but Niels Bohr, is that you forgot that you are a quantum system, that you have a wave function, that you can become entangled with other things. So putting my computer scientist hat on, I feel like they would say that the many worlds is actually less simple from a computational standpoint, because especially if you think that the universe is in some ways doing, and a lot of people do, subscribe to the idea of a computational universe, then presumably a computational multiverse would be requiring hell of data centers.

36:18Like, that's going to be incredibly computationally expensive. Like, I mean, I don't think exponential describes it, right? It would be so absurd. So they would argue that it's actually less simple from that standpoint. And so that's, I think, why a lot of them would throw out the many worlds interpretation. Unless you are really deep into the simulation argument, there are no data centers that are running the computation of the universe. The universe computes itself, okay? And the way to measure the complexity of that computation is not to say how much data is there in the universe, but what is the algorithm?

36:51That's what counts. And many worlds has by far the simplest algorithm. It is only the Schrodinger equation. That is the only thing that ever happens. So the difficult thing to believe about many worlds is that there could be, which I think is, let me rephrase what you're getting at here. That's a lot of worlds. that's a lot of stuff going on right okay but that stuff is potentially there in any version of quantum mechanics if you believe that an electron can be in a superposition of spin up and spin down and you believe that you are made of electrons and protons and neutrons then you should be able to believe stretch your imagination you can be in a superposition of different states And indeed, the universe can be in a superposition of different states.

37:40That's just part of the buy-in. It doesn't get heavier and heavier. You're not carrying that. It's not new information getting created every time it branches, because that same information was contained within the superposition. That's right. It's one quantum state. It's an element of the same mathematical space. It's only obeying one equation forever. But computationally, it would still require more compute to render, I would imagine if we stick with the simulation hypothesis, a gazillion worlds versus one, even when there are some superpositions, because they're sometimes just like a probability function without having yet to kind of run a Monte Carlo simulation over each of those in superposition states.

38:22You can just keep them kind of computationally probably fairly compressed in a superposition, and only later when they interact it's like, oh, now in many worlds, Now it's like a Monte Carlo simulation over Igor interacting here and all the different outcomes. So there's two different answers I can give to this, sort of the glib one and the more relevant one. The glib one is before or after the universe starts out in a simple state. The quantum mechanical state of the universe is a vector, not in a three-dimensional space, but in a very, very, very, very large dimensional space. Hilbert space, right?

38:59Hilbert space is what it's called. the length of the vector doesn't change. All it does is it rotates, okay? That's all the universe ever does. It's a vector moving around in Hilbert space. The length doesn't change. So the reason why it seems more complicated to you is because you started it lined up in a very simple-looking state, and then it's sort of the components of it become more and more difficult to prescribe. But it's still a vector. You just chose a less useful basis to express that vector in. It didn't get any bigger, okay? As you described that the vector moves, you mean that it touches upon more higher dimensionality?

39:36Nope. I mean that it obeys the Schrodinger equation. So I guess the reason why it sounds fuzzy is because we're used to a world full of stuff. We have space, we have things, we have nice chairs in the space, and they're spread out. And so that's a way of describing the universe in terms of which, and I guess this is sort of my second answer to the question, in those terms you need a lot more information to describe the universe today than you did before, and that's exactly because the complexity is increasing. But mathematically it's no harder. It's our choice of variables to use, right? It's our choice of description that we chose to look very simple at the beginning and it becomes less and less convenient as time goes on.

40:26But people really have this feeling like that the universe is in a box and it gets heavier and harder to carry. And that's really not a good image. Coming back to emergence, I feel like it's a term that often gets dismissed by scientists, rationalists as being too associated with like woo. And to be fair, it often gets sucked into that world and used to explain things, often like quantum mechanics as well, like quantum mysticism, where people are like, oh, it's the quantum world that's doing it. And it's sort of used as this almost just like hand wavy explainer for anything that we can't quite explain.

41:07But emergence is actually a very real and important field of study. How is emergence featuring in your work at the moment? Well, glad you asked that, Liv, because I have a book coming out. Volume two of the biggest ideas in the universe. Volume one was called Space, Time and Motion, Classical Mechanics, Relativity. Volume two is called Quanta and Fields, quantum mechanics, particle physics, quantum field theory. And I'm writing volume three, the title of which will be complexity and emergence. So emergence, and it's very much, these three books are very much appetizer, main course, dessert. The middle book, the quantum field theory book is action-packed.

41:49There's a lot going on in there, but super duper nutritious if you go through it. And then you get to have fun and play in book three, where we talk about complexity and emergence. But you're absolutely right that the definitions don't line up, just like life or information or entropy and things like this. People argue over complexity, what the right definition is. There's no right definition. There's different aspects or different things that can happen that naturally get associated with emergence. One, the one that I like to focus on, is just absolutely indisputably real, which is there are conditions in which I can take a theory that is somehow more microscopic and comprehensive, like a theory of atoms or quantum mechanics or whatever, and I can throw away almost all the information contained in a state, keep the right information, and still make accurate predictions.

42:38A classic example would be the Earth going around the Sun. I don't need to know what all the atoms in the Earth are doing to tell you how it goes around the Sun. What I need to know is its center of mass location and its center of mass velocity. All I need to know, six numbers because we're in three-dimensional space. as opposed to 10 to the 50th numbers that I would need to give you all the atoms, right? That's kind of miraculous, just the fact that we can make good predictions, throwing away almost all the information about a state. That's a kind of emergence. It also happens when you just take the air in this room and you average over the locations of the molecules, etc.

43:18You get things like pressure and temperature and density, and you can make predictions. Meteorologists do that, right? So that's emergence. It's the existence of a hidden macroscopic pattern within the microscopic laws that allows you to throw away information. But of course, when you do that, the center of mass of the Earth example is kind of misleading because the kind of theory you started with is the same before and after emergence. It's a theory of points obeying Newtonian mechanics, okay? In the air example, the kind of theory you end up with is different. You started with points of molecules, you end up with a continuum of fluid, right?

43:59And that can happen in very more dramatic ways. And so a lot of people will point to emergence as what happens when you are surprised at the macroscopic behavior. Some people will even invoke emergence to mean you could not possibly have predicted that this would happen. And in both those latter cases, I kind of struggle to understand what they mean. Like, what if I'm not surprised? Did the emergence go away if I choose not to be surprised? What if I didn't know how to predict it, but then I figured out how to predict it? Did the emergence go away? These are all like human judgment words. They should not be part of your definition of emergence.

44:39It does feel, and this perhaps ties back in with the question around earlier information being conserved in the universe um in certain because if we look at emergent things that are playing out in through the evolution of life on earth it presumably if you were observing the the eukaryotic cells once they they emerged um it would have been hard to predict the development of multi-organ animals that seemed like that is an almost new information that has been created so is that not really what's going on or is it like a misnomer of the term information because i've always struggled to wrap my head around this idea of that no there was all the information that is in the universe has has always existed from the moment of the big bang it's just that it's different you just couldn't access it or something like that so what laplace's demon yeah only laplace's demon um this hypothetical thing that has perfect information or perfect access to information?

45:43Or is it truly the case that new information is being created in certain situations? If we put aside quantum mechanics for a second, that's a more subtle question. Let's get to that. Let's imagine first that we're just classical, that Newton had been right. It's still, you could absolutely describe much of what happened since the Big Bang and the development of eukaryotic organisms, et cetera, mostly in a classical framework. And there, information is conserved at the microscopic level. If you were Laplace's demon, if you had access to the exact microstate of the universe, the information is the same from moment to moment.

46:14But we're not Laplace's demon. Someone, I forget exactly what the formulation was, but someone on my Patreon for my own podcast said they want to start a drinking game to have a drink every time I say you are not Laplace's demon because you're not. And so your accessible information certainly changes over time. I mean, imagine, for example, you were playing poker. and in a Newtonian classical world, there is a predetermined fact about the matter about what your two whole cards are going to be. You can't predict it. You don't have the information. Laplace's demon knows, though. And if the deck of cards had 10 to the 100 cards in it, you would be essentially out of luck in predicting what it would be because the probability of getting away would be so small.

47:00But it's still there. So this distinction between what can be done in principle by Laplace's demon and what can be accessed accessed and used in the real world by we bounded macroscopic creatures is super duper important. And Laplace's demon would have also the ability to know that pressure will emerge as a macroscopic description of the many atoms. Well, so Laplace's demon doesn't exist. So you can make up whatever Laplace's demon you want. And I know that some people, yeah, so this is a hilarious question because there are people who will insist that Laplace's demon only knows the exact microscopic state of the universe, not how to coarse grain it, none of the emergent higher-level things.

47:41So Laplace's demon doesn't know about entropy or temperature or love or life or death, right? Of course, I can make up a version of Laplace's demon that is slightly cleverer than that. Who knows all of those things? Yeah. But let me also say the reason why quantum mechanics has to be put aside and then brought back in is, and again, it's the complexity evolving over time question. The early universe, from a quantum mechanical point of view, could very well have had its exact quantum mechanical state be super simple. But then this branching happens over time. And the branching is like, you know, the way that I like to think about it is take, you know, a perfectly circular disk.

48:22And it's very simple. There's not a lot there. But now I divide it in two with some complicated cut. Now both of the pieces are kind of complicated, even though they fit together to form a simple thing. In quantum mechanics, the universe is like that. Our observable branch of the wave function of the universe requires an enormous amount of information to specify what's going on. And there are also many, many, many other branches. And they all fit together to make something very simple, which is kind of awesome. Some people seem to believe that it's possible to infer the behavior, the sort of macroscopic behavior we're seeing of humanity and the economy and how it will progress over time by reducing it back down to the second North thermodynamics.

49:15and it feels to me like that level of reductionism is a bit hubristic. Would you agree? I'm super hubristic. I mean, come on. I study the universe. I mean, what can I tell you? I think that the laws of physics that govern the universe that we actually experience in our everyday lives are pretty well understood. I think that the macroscopic behavior of the universe, up to and including politics and society, must be compatible with those underlying laws. So you have to distinguish sort of dependence from usefulness in some way. In principle, everything that's happening in society or biological evolution or whatever follows from the laws of physics and the initial conditions of the universe.

50:05In practice, that's a dumb way to think about society, right? It's much easier, you know, if you want to make a cup of tea and someone says, okay, start with the standard model of particle physics. That's just of no help, much less complicated things, right? There might, however, be some things we can get out of thinking as physicists. You know, there's a whole field called econophysics, where they try to use principles from physics to understand economics. I'm working on a long-term project. It's going to be years before it comes out, but it'll be a book called The Physics of Democracy, where we use ideas from physics about phase transitions and emergence and complexity to understand society and voting and emergent choices.

50:49Like there's no better, more vivid example of emergence than taking the preferences of individuals and grouping them together to get the preferences of a society, right? And I don't think that we've been, you know, we've been kind of slapdash about choosing how to do that. The interesting new thing is that we're both, there is this emergent thing, But we're kind of shaping it through the forms of government, the voting system that we have, the ways we communicate and things like that. So there's this feedback loop between it's not like the air in the room where there's one right way to coarse grain to get an emergent description.

51:25We're sort of choosing how to coarse grain ourselves. And I think we should be doing that in a self-aware kind of way. Right. Yeah, because historically all of our, well, I mean, there have been some instances such as the founding fathers who decided we want to start a new country and we're actually going to think about this from first principles and design this system in an intelligent way. And it, you know, I'm not saying America is perfect, but it's worked out pretty well. Much better than it should have, honestly. Well, but that, yeah, that speaks for something, you know, there's clear value in thinking about designing the rules of the game as opposed to letting them just emerge naturally from or evolve from previous, you know, especially the nature, which is often read in tooth and claw.

52:11So are there particular areas of systems design that you're focusing on? Well, I'm still very much in the research stage here, but both voting theory and game theory are very important, right? There's different actors with different interests, and we're trying to figure out how best to accommodate them. And it's not just a descriptive question, it's a prescriptive question. You know, whose values count? How much do they count? But I do think that at least we know more about voting theory and game theory than the founding fathers did, right? So in principle, we could do better. Now, in practice, the reason why I say that America sort of did much better than it should have is those people there who were there in the 18th century designing the Constitution, etc., were both pretty darn smart.

53:03And for the most part, acted in good faith. There were, you know, the people from the South wanted certain aspects of things. People from the North wanted other aspects of things. I saw Hamilton. It was great. You could actually see that play out in real life. Fine. But overall, like, they were reading their Locke and Rousseau and Plato and Aristotle and thinking hard. You know, the Federalist Papers were a really in-depth examination of some of the very difficult questions that would arise in politics. I doubt we could reproduce that now. I don't, the last thing I want to see now is a constitutional convention, because it's the right people would not be there.

53:43A good constitutional convention right now for the United States could vastly improve our system. The chances that that's what we would get are infinitesimally tiny, I think. Is that because power maximizers would be attracted to it, which tends to be... Less good faith actors would be chosen. And also, you know, the country is old now. It's settled in its ways. There's a lot of historical baggage that we have to put up with, much less of a feeling of experimentation. It's a more complex system. It's a more complex system. Certain norms are entrenched and so forth. So, yeah, I mean, I would not want to see who...

54:25I mean, look at who we vote for. Come on. Right. Well, more specifically, look at the options that are provided by the system. I think that's the more interesting. I mean, yes, OK, the voting thing, but I, you know, I not to get too heavily political, but I'm not thrilled about either of the likely options that's going to be presented. Now, I can't vote. I'm a immigrant, you know, so it's like I just have no involvement anyway. But still, it's like, how is it that a system that is in many ways been so successful is, is it just that beyond a certain level of complexity, the inertia becomes so big that it tends to just put forward the two things that are most likely to continue maintaining its existing structure?

55:05I don't think it's inertia, but I do think that there are, you know, I think in some sense, the system is almost, the country is almost too big to be governed in a fair way. It's so difficult for people to figure out new things. They just stick to the old. Like you mean that the oh Biden again Trump again oh Clinton new Clinton like we're okay just just just stick with something simple it's like the inertia the the desire for simplicity no no it's more it's it's it's that well at this point you know it feels like the the America is almost split into these two strong power structures you've got Republicans and Democrats and what are the ways for that to continue sustaining itself is to put forth the two candidates who, just looking back at history, they've both successfully been voted in before.

55:56So let's just put them in again. I think logic might not be the right word to talk about this, because on the one hand, you know, we have primaries. People voted for Trump and Biden, right? They had alternatives. So why did they vote? And, you know, the answer is complicated. I don't want to oversimplify it. Part of it is not paying attention, being low information voters. Part of it is money, getting out the word for different things. Part of it is the media system choosing to highlight certain things and not others. There's just a million things that go into it and full employment for people who study this stuff for a living.

56:31But don't oversimplify it. This is the only thing, only huge mistake you could make. How does game theory, or actually more specifically decision theory, change? Or would it change in the many worlds framework? work? So I think it wouldn't, but this is a very good question. I did talk a little bit on my podcast on Mindscape with Laura Buchak, who is a very elite philosopher of decision theory. And her angle is that standard formulations of rational choice theory, etc., don't allow for risk aversion in particular, risk aversion, to be taken into consideration. So standard decision theory will tell you that if there's truly two choices, one is between an option with utility one and the other is 50 % utility zero or two, you should be indifferent between them.

57:31If they're truly measures of utility, obviously there's marginal values, et cetera, but okay, if they're truly utility. And Lara's point of view is that I can absolutely invent a set of axioms for rational choice that says, no, I don't want to risk getting zero. And it would be completely consistent, right? You know, it's not like you reach some contradiction or anything like that. And she says that's, she's not saying you have to do that, but that it's just as rational as you have to be, okay? So if that's true, so I'm giving you a long-winded backwards answer, but in the straightforward way of doing decision theory or rational choice theory, many worlds has no impact.

58:12In other words, it's not any different than just random numbers, just expectation values for probabilities. No different than the Copenhagen interpretation. If you say there's a 50-50 chance of being in one world with a lot of value and another world without, your decision would be no different than if there was just a certain 50 % chance that one world would come about, as opposed to two worlds with 100 % probability that both worlds exist. But if you were super risk averse, that might change that. And honestly, the answer is I don't know. So do you think there's any low hanging fruit or solutions to these multipolar coordination problems?

58:54Because so many of the issues that we're now facing are a result of it getting increasingly hard to coordinate in many ways, particularly with information, the ecosystem is breaking down and so on. So are there any sort of schools of thought that you're going down or areas that you think hold promise? I think I'm more in the stage of diagnosing the problems than in solving them, I have to confess. I mean, the problem, as you mentioned, is very much there. What I think is underappreciated, at least temporarily from my current point of view, is just, again, how big the system is. we're more connected than ever before.

59:33So not only is the world literally bigger, but the parts of the world we interact with are bigger, right? Because we have giant corporations, we have the internet, we can talk to people, we have fewer and fewer mainstream news outlets to get, no local newspapers anymore, right? And so that gives us a feeling of helplessness because we can't affect what is affecting us. And yet we can see it all.

1:00:29into other communities and just not getting them and not understanding them, etc. But they have to bump into them in a way they didn't used to have to, right? And so I don't know how to do better at this. In fact, I kind of worry that it's just going to get worse. How does competition manifest within academia and physics? That's a great question. And I will not try to wimp out just by saying it's complicated, but it is, right? You know, I often, one of the things that strikes me is we are not very well trained as a society to disagree with each other. I think we're better trained in academia because, you know, people accuse, people on the internet are just like, well, you're just afraid of criticism or whatever, like, no, I'm afraid of idiotic criticism, which is what I get on the internet all the time.

1:01:21I guess a professional physicist and philosopher, people disagreeing with me is just my day job, right? Like it happens all the time. But it's not like, oh, you're an idiot. It's, you know, the substantive disagreement. It very often gets, not very often, but it will get heated and personal and annoying sometimes. But much more often, I have people on my podcast who I deeply disagree with. We have a great conversation about what it is we disagree with, right? And I'm just surrounded by people I disagree with. And that capacity, even though it's only, it's sort of mediocre level in academia, it's almost non-existent outside academia, at least in the general social media landscape.

1:02:02So there's 100 % competition. There's very limited resources. I tell my graduate students, at a typical good physics department graduate school, 99 % of the people there want to become tenured professors of physics one day. 25 % will. that is going to make life tough right and i tell them i don't want to hide it so i'm like this is what you're in for i don't feel bad about letting people into grad school because getting a phd in physics leads to good things otherwise but it's going to be competitive and it's not necessarily you know competitive with your nearest neighbor because you're competing with people elsewhere and things like that but another thing is this all just anecdotes that adds up to a picture but you know students sometimes say you know I would like to ask more questions and seminars but I worry that the other professors are judging me and your instant instinct is to go don't worry they're not judging you but then you think about it and you're like yeah they're judging you all the time and then you have to say like they're judging you for asking questions and also not asking questions they're judging you in the seminars and in the hallways and ranking people and saying, like, who's good, who's not so good, is 100 % always happening in academia.

1:03:18It's a little bit more meritocratic than when it happens in Hollywood or Washington, D.C. or whatever, or the business world, but it's absolutely happening. And it's the single worst aspect of academia. I love academia. I'm part of it. I'm old and established enough. I don't need to worry about that that much anymore. I know there's people who think super highly of me. There's people who think super lowly of me. I've learned to deal with that and trying to get along with my work. But when you're a student or you don't have your job yet, it's just ever present. How much of a problem is the publishing pressure?

1:03:55Because obviously so much of science, the wider metagame is you need everyone's competing for funding or for tenure. How do you do that? You publish more papers in the right prestigious journals. The journals have all this control because they're the ones who have this, they've got the, somehow they've gotten the monopoly on prestige. And it creates this kind of stuck system. At least it appears that way from the outside. Is that actually the case? Your own little multipolar trap, basically. Well, I don't, I mean, I was with you up until the end about the stuck system. I think that academia is actually pretty successful at generating new ideas.

1:04:34And I mean, there are bandwagons, absolutely. but the bandwagons come and go, right? That's kind of what you want. But what I would point to is just the fact that there is hyper-specialization so that I can't productively judge the work being done by people in my department downstairs, right? Because it's too different. It looks pretty good to me. And that's why people start relying on these simplistic metrics. How many papers have you written? How many times did it get cited? Where did it get cited? Like what you want ideally is read the papers and judge whether they're good. But who has time for that or even the ability to do that?

1:05:13Right. And so that is like there's a sort of cheapening of the way in which we evaluate people on the basis of that. And honestly, it's harder for smaller, less prestigious places because they're going to have fewer faculty. They're going to be less able to judge whether the work is intrinsically good or not. so they're more likely to promote and hire people just because of quantity rather than quality. I would encourage them not to do that. I would actually try to engage with the substantive output that people have. But the other thing to mention, which I say very sincerely, is for the most part, I found that people within academia try really hard to be fair in judging other academics.

1:06:01I've been on grant review committees. I'm on the committee that gives out Guggenheim fellowships. And the blood, sweat, and tears that these people do to read proposals and really try to sincerely judge them. I've been on grant review committees where super famous people have applied and nobodies have applied. And the nobody's got the money because the super famous people phoned it in in their grant proposal. And everyone was very comfortable in saying, sorry, you didn't deserve it this time. But there's a million problems. One problem is just that we spend way too much time begging for money in academia, especially in science.

1:06:40We have a lot more money in science than in the humanities. But as a result of that, some huge fraction of effort by academics, especially experimenters, observers, and things like that who need telescope time or equipment money or whatever, they just generate grant proposals, right, rather than generating science. And that's no way for the system to run. I mean, the simplest thing would be to say, you know, you get a grant and the grant lasts for three years and every year you have to say what your progress is. Forget that. Have the grant last for nine years and every three years you say what your progress is.

1:07:13So people can actually spend their time doing science. But then that's a loss of control from the grant granting agencies, et cetera. yeah that said it does seem odd to me the current as i understand it to be the state of scientific journals and where they get their reviewers that are unpaid as i understand for the most part oh yeah then you submit after a lot of work um you don't get paid for the submission by the journal either so you produce the content they check the content unpaid and then even you need to pay actually to then read someone one of the other reviewers content later so it's not like you're getting into the system at least if i once submitted i get to now be part of it seems a bit like a racket from the outside there's definitely a racket aspect to it yeah is that your that's what i meant before or we meant with is that the kind of academic flavor of the coordination problem that it's hard to break out of it well i think that uh there's a couple things going on one is that, you know, capitalism exists, and if people find a way to scrape money off of many other people, they're going to do it.

1:08:19And publishers like El Civier or whatever absolutely do this. And they're a more respectable one. There's a lot of just predatory publishers. They own most of the non-Springer Nature journals, right? Right. Things like that. And they charge exorbitant fees. And it's completely out of proportion to the value that they add in, which is why normally I tend to publish in places like the Physical Review that are published by the American Physical Society, not by a commercial publisher, right? Furthermore, you can always just put your paper on the archive and people can read it for free. Sometimes I don't even publish my papers.

1:08:57It's too much work. My perception is that archive got a bit of a boost over the last few years. I see many more links now when I'm looking up someone's paper for them to also exist on Archive. I don't know. For me, I trace it back to kind of the COVID times when it seems like we changed a little bit the speed of publishing because we needed new information and people were just posting on Archive prior to peer review or with like minimal peer review. Is there such a trend? It's very, I'm not going to even guess because it's a very complicated system and there's microclimates within the system. In the fields that I work in, traditionally of high energy physics, astrophysics, and so forth, we invented the archive.

1:09:40And we were early adopters of it. And the idea of putting out your paper before it got peer-reviewed and published is just second nature. Of course, everyone does that. Lots of senior people in my field don't publish in journals at all. They only put it on the archive. Because they have tenure. What do they need? Whatever. Other fields are radically different than that. and there are, you know, ways to gain the system vis-a-vis prestige, etc. And, you know, some people, if you want to publish in Nature or Science or whatever, then maybe they don't want you to put on the archive before you do it.

1:10:16That just depends on details. And so there's a whole field of study of science studies that tries to think about this, and I kind of don't worry about it because it doesn't bother me. I don't bring in millions of dollars because I don't have any equipment or very few people my group is a few students and a postdoc or whatever some whiteboards and some brains I don't need to concern myself with that so I'm not an expert but the only thing I would say is it's very very hard to get the full picture exactly because there's so many different styles I've often just said why doesn't everyone just make their papers publicly accessible.

1:10:56It's just completely insane to me that there is publicly funded research that is not publicly available for free. But it's because it's a trap. That's the thing. If they don't publish in the fancy journals that charge a lot of money, then they don't get the prestige. That's right. It's a prestige trap. But also it's a coordination problem. If all the academics said, let's start a journal where we can just publish it and make that, then it would change tomorrow. Has anyone tried? Yeah. Right, and that's the thing. And then some will defect because that's... People have tried, but they've tried in the fields that were already amenable to it, right?

1:11:32So it's like the Journal of High Energy Physics, Journal of Astrophysics and Cosmology. But these fields were really not in that trap already. So it's more like the bio fields where you might actually be making a lot of money or, you know, computer technology kind of things. I'm free from the danger of making lots of money from my research. People don't want to trap me. really. If you could redesign physics academia, what would you change? I would make it easier to be interdisciplinary. I have this vision of at least a couple of universities that just don't have departments, that just try to hire the best people, offer classes, let students design their own curricula in consultation with a faculty mentor or whatever.

1:12:18But I've just, it's always is very, very depressing to me when I'm on a search committee or whatever. We're searching for a biophysicist and someone says, well, are they really a physicist? Or an economic historian? Are they an economist? Are they a historian? Like, who cares? You know, like I'm a physicist philosopher and it was very, very difficult for me to get a job allowing me to do what I want to do. Even if, you know, if someone had done a job search for a physicist philosopher, I would have been right near the top, right? Maybe not the top, but I would be a viable candidate. But no one does that because you either do physics or philosophy.

1:12:56You've recently done a solo podcast about your updated beliefs around the advent of super intelligent AI or highly intelligent AIs. Great, obviously, that you're putting out, hey, I had a certain set of beliefs and I've updated them in a way. Yeah, can you talk to that first a little bit and maybe about the, not journey, but like the point that made you change your opinion? Well, I've changed my opinion in various different aspects. I mean, I still believe that I don't ever, well, I want not to use the words intelligent, much less super intelligent when talking about AIs, because I do believe that what they're doing, the current models, of course, the future could be very different.

1:13:40I'm not saying what is allowed or disallowed by the laws of physics, but looking at what is actually going on, I think it's just a very natural human mistake to treat them as the same kind of reasoning creatures that we are. They are something, they're doing something, and they are absolutely optimized to sound like they're doing the same things that you and I are, but they are clearly not doing the same things that you and I are. I agree with that, but them doing something else does not imply that they are not intelligent in another way. I agree that we are like anthropomorphizing them. But still, yeah.

1:14:20And then we're going to the definition of intelligence. But is the chess engine more intelligent at chess or smarter at chess? It's like, well, from one you could argue maybe no. But in the end, the thing that matters is it produces better results in chess, right? So if we take the intelligence away, like I agree that we anthropomorphize it. But yeah. I can see it producing very valuable results. And you do too. Absolutely, yeah. Yeah. But that's the reason why I don't overuse those words, because until we had these AIs, the only other actually intelligent creatures we knew were ourselves. So it's almost irresistible if we attach that word to something else to attach all the connotations of that word.

1:15:05I mean, it's not just AI. I don't like, do you guys know the balloon analogy for the expanding universe? People try to explain the expanding universe by saying, imagine I have a little balloon, I put dots on it, and I blow it up. And all the dots move away from each other because the balloon is getting bigger. And they say the expanding universe is like that. And I hate this analogy. And why? It's because, well, a person who doesn't already know the answer says, well, okay, so what's inside the universe? Like, what is it expanding into? And you have to say, no, no, no, I didn't mean that. I mean, that's just part of the analogy that doesn't go along, right?

1:15:38Some aspects should be taken seriously and not others. It's super hard for people to actually do that. I would rather just not give them the misimpression in the first place. So I don't have the right vocabulary, but I would rather try to accurately capture what the AIs are doing rather than using phrases like superintelligence. Cool. Yeah, I agree with the phrasing, but I wonder if as you've updated after listening to Jeffrey West, and it takes time, as you correctly point out, sometimes you hear something. It's not like it immediately changes, but rather one sits there and digests, etc. So I wonder if we'll get to another degree of update potentially through this conversation now.

1:16:18So the thing that I see differently is you talked about that, well, it's anthropomorphic to assume that they are intelligent in the same way or agentic. Agreed with that. But neither of that implies that it won't be more capable than us across certain domains. My phone is way more capable of multiplication than I am. So, I mean, that's impossible to disagree with what you just said. Yes. And then I agree that the idea of artificial general intelligence, if it implies that, hey, the AI is, there are various definitions for it because, again, it's a broad term. But one that I like is it is at least as capable as humans across all economically relevant tasks.

1:17:03And it's like, will not necessarily this current architecture of AI, but even a different one do that? And I think your point is relevant there. It's like, no, it's a different mind. So it might never actually cover even specifically every single economically relevant task we do because it's a different structure of mind. But maybe that doesn't even matter is how I feel about it. Exactly. I completely agree. The thing that matters to me is it will be more capable at a number of them, probably. Also very capable at a number of ones we don't even consider ourselves intelligent on and many we don't know about.

1:17:42And through that capability, we will integrate it into our economy and into our just like various aspects of life very, very deeply. And that's where I think that also the point about it's not necessarily agentic stops mattering as much because at the point when one relinquishes a lot of the control which is what you do to a more capable system or person right um yeah does it still matter whether it's agentic or not you're kind of like kind of yeah ultimately it doesn't have a goal but you gave it so many steps to do that you can't monitor that uh it can just drift off in some way. You know, I don't know whether I'm hitting my head against a brick wall because people don't agree with me, but I would argue that, I mean, I agree with all the substance of what you just said.

1:18:37I think AI and other technologies that we might not attach that label to will get much better than human beings at many, many things. Maybe not all the things, but so many things. And we will be absolutely turning over a lot of work to these computers and algorithms, etc. I nevertheless just think we make ourselves dumber by applying anthropomorphic language to them, whether it's general intelligence or agents or values or any one of these things. I think that we're papering over the very real differences and therefore not addressing the more realistic worries. There's good parts and bad parts of all these things, right?

1:19:22And by forcing them into a box that was created by biological evolution for entirely different purposes, we're going to mischaracterize them. And so I want to take them seriously as what they are, which is a little bit different. The thing, though, that people often hear when one says, well, are they intelligent? And one is talking more about the different structure of mind. What many people hear, I believe, is, oh, they will never even reach our capability. So the thing you're though saying is you can very well imagine it surpassing human results on across many many domains do you think that we will be able to control what it does well well i don't know i mean do we control the atomic bomb uh we control whether we turn it on or off not the people who got hit by it i mean right the control is diffused we control the internet yeah exactly right no can we turn off the internet yeah and i think that's that's an example people often are like oh it's not they they view it i i see ai becoming increasingly more like a sort of distributed more agentic form of the internet in that it will become just so ingrained into our economy and our society um that that i mean that might be a great thing but it also it might speed up the things that were that are already misaligned in our system further that's a big concern I have like right now our economy I mean has been great for our it's like we're here in this wonderful house with this all this technology and so on love that at the same time the we are externalizing so many costs we're starting to approach I'm not saying we're at but we are if trends continue we're approaching some kind of planetary boundaries in terms of resource extraction and if AI just sort of speeds up that process even faster then that existing this alignment is going to increase and so that's like an area that i'm personally very concerned about uh does that resonate 100 that's that's the dark side of efficiency that i was mentioning before right because efficiency as a you know concept all by itself sounds good but if it's efficiency of extraction it could be dangerous you're going in the wrong direction the last thing you want to do is go even faster.

1:21:38You can be efficient and really good at doing something bad. And that's why it kind of rubs me the wrong way when people worry about existential risks from AI. I'm not against worrying about them, but I think that the right strategy for ameliorating them is to really worry about the short-term obvious risks, right? And we'll get better at dealing with AI and sort of incorporating it in safer ways if we don't extrapolate out to the extreme, but we just, you know, like, what are the like, the less sexy, but really right in your face kind of things that could go wrong right away? Because there's a lot of them.

1:22:16And, you know, but it's, it's not an either or, that's the thing. I think a lot of people, I don't understand how this, this concept has come up of where it's like, if you worry about existential risk, you aren't worrying about present day risk, because the people I know who are worried about existential risk are also very worried about the real term things as well, because they're like, look, we've already got misalignments happening. They're already doing things we don't want them to do in many situations. So to me, they're all examples of the same kind of malaise, essentially. And by the way, I want to also throw in, people worry about it, and it's such a big thing, but it's because obviously the benefits are going to be so massive and great of it that ameliorating those risks means being able to capture the benefits, which is the goal.

1:23:08That would be fantastic. But yeah, I guess it's just a quirk, I think, almost of the media in the fact, because it is this sexy, dramatic sounding thing. What's the biggest story? Oh, Terminator, Extinction. That gets sort of the media attention. So people feel like that's the thing where all the funding is going. But from what I use, I mean, as it is, like all areas of AI risk are drastically underfunded compared to the amount of funding that's going into just like speeding everything up and the progress. People start very idealistic and okay we're gonna you know work on saving the world and then someone comes along and says okay but if you do this other thing that is not really going to save the world but I'll give you a billion dollars.

1:23:53Right. It's very hard I don't know what I would do. I'll put a little bit aside yeah. I wonder, actually, as we discussed anthropomorphism of AIs and actually of the concepts of intelligence and other things, I want to point towards consciousness. And as I understand the history of the thought of the different theories of mind, I'm personally more closer to the functionalist or functional theories, like attention schema and global workspace theory, etc. So attention schema theory I like because it's probably a bit more understandable. But as I do have a physical body, I have a representation of my physical body in my mind, which is taking up a separate space.

1:24:37And that's the body schema. And the attention schema theory is that this thing is equivalent to how we have attention, what we're putting it on. Obviously, we're processing much more data, but not everything receives our attention. And that we equivalently have an attention schema, as we have the body schema, which is the brain's representation of our attention. And that this representation of the attention is not equal to attention, it's slightly different, which is why consciousness and attention in that theory are so connected. And experientially, they're very connected, but they are slightly different.

1:25:19So consciousness is like the flashlight of attention looking at the attention schema that is yourself. That would be more a global workspace where it's like you put the flashlight you put on something that you're putting attention on. And through that, it gets more capabilities to interact with the rest of your brain, basically, and can now access more memories and all sorts. Michael Graziano recently wrote a paper where he wanted to combine all of those functionalist theories and to say, hey, we're using different language to describe kind of the same thing. Let's work together and then we will stand on like a better footing to defend these theories as like a good description of what's going on.

1:26:04Whenever I see some famous philosopher, neuroscientist, academic say there are these two opposing theories, but I can actually figure out how they're saying the same thing. what always happens is that one of the theories says oh yes you're exactly right and the other one's like no no we didn't mean that there's a resistance to be yeah it's a recent paper as well so i don't i haven't seen the backlash from higher order and the global workspace theory yet but yeah the as i understand the history of those like functionalist theories coming up it's happened at the same time as we developed computers as well and someone proposed that maybe it was only seeing the kind of, and computers are very functionally designed, so that it was designing them that allowed us potentially to understand our mind in a new way, in a better way.

1:26:57So it's not, and people describe similar things where neuroscience not always has been informing AI, but sometimes actually AI development has also now recently informed neuroscience. So I wonder if, equivalently to anthropomorphism, there is like computer morphism or something you see the inverse of that also and maybe we are not that dissimilar um in some way maybe there are like some structural things that are just efficient way to combine information right like as we discussed uh compression algorithms for example probably something that will exist in some form with any um anything that tries to be intelligent Yeah, I think that, I mean, again, these are complicated issues, but you're certainly right that human beings, as soon as they discover a new technology, say, well, maybe we're like that.

1:27:49You know, back in the, when we discovered simple engines, suddenly people were machines, right? And then, you know, we discovered steam engines, and okay, people are thermodynamic. We discovered computers, and then people are computers. And there's something right in all of these things, right? They're not crazy. And I think that it's a little bit difficult, but we have to keep in mind two things. One is there are certain universal principles that you would not be surprised if came up in both these artificially constructed things and in ourselves, right? Levers. Levers are all over the place, right?

1:28:24Not really that surprising. The form of the eye is not that different from the form of a camera, okay? but there is a difference because one is designed and one just sort of came up through biological evolution and that's plus or minus you know like the artificially designed things tend to be much more fragile than the things that are intelligently designed because we have a purpose we try to optimize them for that purpose they're much less general purpose things like a human being is very inefficient for lifting heavy objects compared to a forklift but we're much better at other things right and you know when you when you envision robots and things like that they're always these metallic things but the thing about being metallic is it breaks and the human being can break their leg and it self-repairs and that's because they're made of little tiny pieces that can coordinate this is a definition of the difference between complexity and complicated it kind of is yeah yeah a complex system has this ability to dynamically evolve well an emergence Exactly.

1:29:24And that's, I think, what makes something alive is that it can emerge and things, whereas a forklift or even a very complicated machine is very complicated, but it is not alive because it will always do the same thing that it is designed to do. It's not going to evolve into something else. And I think that's where AI is interesting and AI, particularly in its physical instantiation in robotics. If we have now, if we're now developing a machine that is able to sort of self edit its own code, make self adjustments. Now we're merging the two and we're blurring the distinction between the complicated and the complex, which is very exciting.

1:30:07It's also kind of scary because it's like, well, where does this go? yes all i can say is yes um the i do think you know i try to occasionally tell this to my physicists computer science friends but i do think that you can't ignore the biological side of progress that is going on right now because biology has had a four billion year head start and it's figured some things out and you know francis arnold uh nobel prize winner from caltech who does synthetic biology you know she just sort of shakes her head at everyone inventing nanomachines and things like that. She goes like, it takes them so long and then they reinvent the cell.

1:30:42Every time they reinvent biology. And I do think that the future form of technology will be more hybrid, right? More cyborg-y, more in between biological and technological. And that is a space that hasn't been explored nearly as much as it could have been. So that's the nature of phase transitions is that there are different forms that will take over. So they're hard to predict. It seems like there are then, yeah, relevantly true pattern that you find while looking at one thing potentially in the other thing, right? Like, we're always trying to explain the human through the thing we're working with.

1:31:21And as you said, like, sometimes we pick up some kernel that is relevantly helpful, actually, for us to understand ourselves better. You've developed this or furthered an idea around naturalism, which you, that poetic naturalism, and that made just a lot of sense to me when I heard you describe it. Maybe do so just briefly. Sure. Poetic naturalism, the motto is there's only one world, the natural world, the world that is discovered and described by science. But there's many ways of talking about the natural world, and that's where emergence comes in. But the slightly longer description says there's not only many ways of talking about the world, but there are different ways that are useful for different purposes.

1:32:05So if you're being purely descriptive about the world, if you're just being a scientist, then the natural world itself kind of tells you what ways to talk about it, right? There's the Laplace's demon way, where you're maximally comprehensive and microscopic. There's the various emergent ways, where you have tables and chairs, planets and people and puppies. But then there are judgmental ways, normative ways, aesthetic ways of talking about the world, And those are not fixed by the phenomena. You can disagree with someone about the aesthetic value of a painting, and there's no experiment you can do to figure out who's right and who's wrong.

1:32:40And poetic naturalism says that's okay. Just recognize it, right? And it goes further than paintings. I would go all the way to morality, something that we make up, something that is not fixed by the data, something you can't do experiments to distinguish between different theories. But it's still there. It's still legit. It's still important to talk about. less real no not at all and it's still you you discussed that uh for example free will is a in that framework something that as long as it's useful to explain certain actions then use it and it's uh relevantly real and that's a question that i often wondered previously of like okay so what is this line in the sand of where something is real like ontologically accurate as a description about the world and where is it just a good model?

1:33:29And it seems like the consistent thing to say is, well, we just, there is no, like, I don't know what the ultimate reality ontologically accurate description is. So everything is kind of a model. That's right. And then the notion of real disappears. So then you start using it actually again, you redefine it and use it for other things. So I think you've discovered what is called structural realism. I don't know if you're a fan of structural realism already, but this is an existing perspective in the philosophy of science, one that I actually really, really like. And so the idea is the following.

1:33:57When you go from Newtonian mechanics to general relativity, so in Newtonian gravity, you say what the world is is an absolute three-dimensional space, absolute time, some particles that are instantaneously interacting with each other through an inverse square law of gravity. General relativity says, no, that's not the world. The world is a four-dimensional space-time, and gravity is a field on that space-time which obeys some differential equations. Like, all of the words are different. But the predictions are almost the same, right? Except for, like, a little bit about Mercury, etc. You get the same predictions at the end of the day for what you're observing.

1:34:39So, and there might be another level underneath. We don't know yet, right? It might be some quantum mechanical immersion space-time. So there's a worry about the what is real question. Like, how do we know what is real if we keep throwing away all of the ingredients in our ontology and replacing them with something else. So the structural realist would say the fact that the empirical predictions agree in a certain regime reflects the fact that even though you're changing the ontology, you're keeping some structure. You're keeping some relationship between the different parts of the ontology that is actually not going to get thrown away, right?

1:35:13We're not going to invent a new theory that we say is better for which Newton's inverse square law doesn't work anymore, right? So we We don't need to know the stuff. It would be hubristic to say we know the fundamental stuff of reality, but we know some aspects of the behavior and the patterns and the structure within reality. Similarly, in poetic naturalism, you're also not assuming that we know the stuff. And then basically, though, the question of what's real and what's not kind of shifts the line in the sand, though, a little bit as well. So where does it go? So quantum mechanics is somewhat more real potentially than Newtonian mechanics in terms of it is more useful in more situations or something like that, I could probably say.

1:36:01But Newtonian mechanics is also quite useful. Maybe it's also somewhat real. Then things like free will are also useful. They're also real in some sense. Where do you put the line in the sand? is a girl that finds her relationship with a unicorn she imagines. Is that real in a relevant sense to you as well? Yeah, I want to, in all super-duper honesty, say I don't care. I want to know what helps you get through the world. What has causal efficacy? What makes good predictions? What can you share and convey information with? I want to count the unicorn as not existing because I think it doesn't really have those.

1:36:41Like if the people who are not the girl don't believe in it, their lives are not affected in any way, right? You can't use the knowledge that she sees the unicorn there to make predictions about things that are not related to her, etc. So what are the pieces of various causally efficacious pictures of the world at different levels? The chair is real, I would say, because when you say, please sit on the chair, I instantly conjure up various chair properties that are pretty reliable. It's this thing I can sit on. I know approximately the size and its purpose and things like that, right? Right. So but but if you get into the nitty gritty, which like my philosopher colleagues love to do, I kind of lose interest.

1:37:26I think this is a effective get through the day kind of thing. And I'm happy just to get through the day, not to pretend to draw bright, sharp lines where they don't exist. I mean, it sounds to me like poetic naturalism is actually allowing for a superposition of theories to exist. well that's really what it is it's like the quantum mechanics the chair does not stop existing when we learn that it's made of molecules right i think that's just very very beautiful one of the people who's at uh santa fe is ted chiang the science fiction author so we were talking about almost this question you know he was raising the question if you saw something in the world that was contrary to all of what you thought were your knowledge of the laws of physics etc but it was really very vivid, maybe you first say, oh, I'm hallucinating, or I'm being tricked.

1:38:21But what does it take? When do you cross the threshold? When do you truly throw out your worldview and replace it with something new versus just saying it's a hallucination? And I think that most people, I think Ted was trying to make this point, most people are kind of reluctant to think that they're seeing a hallucination, even if it violates everything. They're more likely to throw away everything they knew than to deny the evidence of their immediate senses maybe maybe too much because our senses are very trickable maybe that's a function of being around uh scientists i kind of feel like yeah maybe i wonder whether in the general populace that's the inverse occurs well also have you well the theory would be like the stronger your world models are the more less you the more it takes for your senses to kind of uh overpower how you make up reality in the end but you're right the the early galaxies are a perfect example of this thing that it has to fit in i talk about this in the big picture a lot that we don't judge phenomena isolated right one by one we we always ask how they fit into the bigger picture um the galaxies you say you see some galaxies earlier than you expected, that is just, it's hard to convey to non-cosmologists.

1:39:40That's not going to cause scientists to doubt the Big Bang model, because the other pieces of evidence unrelated to that particular phenomenon are just so overwhelming. We might work hard to change our theories of early galaxy formation. That makes perfect sense. But yeah, and there's many, many examples of this. When we discovered the accelerating universe, like I said, the reason why that was so easy to accept is because you didn't have to change your whole world model. You just had to like fit things together in a different way. It helped. And it's kind of why Karl Popper is not right about the philosophy of science because, you know, science does not precede one falsification after another.

1:40:22Sometimes things are apparently, like, you get an experiment, oh look, neutrinos are moving faster than the speed of light. And everyone's like, yeah i doubt it and then oh a couple months past we didn't plug in the cable okay that's a much more easy to swallow i mean it's really that's the question of just like are you a good bayesian or not and are your priors well but but but specifically your priors are interconnected right so it's not it's yes you should be a good bayesian but the set of propositions for which you have priors are not isolated from each other and sometimes the better theory is one that it has some weird effects here and the alternative theory has weird effects there.

1:40:58It's like which version is actually the less confusing one for many other areas. People have often wondered about the fundamental discreteness versus smoothness continuity of the universe. Sometimes people will tell you that quantum mechanics points towards fundamental discreteness, which is just wrong. That is just incorrect. Other times people will say, well, Stephen Wolfram has a theory where it's just graphs. You know, that's true. He does have a theory, but it's not nearly as successful as the other theories we have. So in the battle of discreteness versus continuity, right now all of the victories belong to continuity, okay?

1:41:34So the reason why I wrote that paper called Completely Discretized Finite Quantum Mechanics or something like that was just like a point of a little loophole that says, okay, here is how I think you could make a discrete universe without completely overthrowing the laws of physics as we currently understand them. So I think it's possible, but I'm not necessarily advocating for that position. I'm just letting the world know that you don't have to throw the baby out with the bathwater if that's what you want. What is it that makes a great physicist? I'm a pluralist at the end of the day. I think that there's different ways to be a great physicist, as there's different ways to be a great basketball player or poker player or podcaster or whatever.

1:42:11I'll give you an example. Richard Feynman and Murray Gell-Mann, who were both great physicists, not exactly the same age, but pretty close. They were contemporaries at Caltech. I used to sit at Richard Feynman's old desk at Caltech. That was my claim to fame. Completely orthogonal personality types and completely orthogonal styles as physicists. Feynman was, I mean, both of them were amazingly smart, and they could do calculations, and they made enormous contributions. Feynman, and people don't quite realize this. Feynman almost never proposed a new law of physics. His thing was taking the existing structures and understanding them better than anybody else.

1:42:56Maybe reformulating them to be even better. He didn't invent quantum mechanics, but he invented the path integral formulation of quantum mechanics, right? He didn't invent quantum field theory or renormalization, but he drew those diagrams and he made it much clearer and accessible to the masses, etc., etc. etc. Gelman was like every Tuesday he would invent a new model of particles and so forth, symmetries, and he predicted new particles and they were found and so forth. Feynman was self-consciously a shaper of his own image. He worked really hard to be thought of in a certain way. And that way was like almost self-consciously aw shucks, right?

1:43:36Like, I'm not so smart. I'm just smarter than everybody else, right? You know, he had the thick Brooklyn accent. That thick Brooklyn accent sort of came and went, depending on who he was talking to, right? Gell-Mont, on the other hand, super uptight, you know, dressed like Barry Natalie, was the kind of person who would correct you on the pronunciation of your own name, because he knew the etymology of it, right? And he wanted everyone to think that he was the most sophisticated person in the world. Feynman wanted everyone to think he was the least sophisticated person in the world. They both wanted everyone to think they were the smartest person in the world, right?

1:44:08But they were both great at doing physics. You know, Paul Dirac and Albert Einstein could not be different. Galileo and Newton were very, very different from each other. You know, aside from being smart, working hard, being creative, I don't know what makes a great physicist. I think that doing great physics is the answer. There's this famous picture of the 1927 conference where so many of the... The Solvay conference. Yeah, greats were all at the same time. And it seemed like, looking back, it seems in a way like it's the golden age maybe of physics. But maybe it was a particular great time to make discoveries.

1:44:52What do you think? Is it that we have those great physicists today and it's just so much harder? nature is not going to have any principle of handing out great scientific discoveries to be made on a regular interval. The first half of the 20th century will go down in history as being absolutely unique in the history of physics. And we're close enough to that that we kind of still feel bad about it not being quite as cool. But we invented relativity, we invented general relativity and special relativity, quantum mechanics, quantum field theory, the Big Bang, the expanding universe, nuclear physics, radioactivity, much of modern quantum field theory, etc., particle physics, all in the first half of the 20th century.

1:45:34The way that I say it is we discovered a whole new set of things in the first half of the 20th century, so much so that the second half of the 20th century in physics was basically just tidying up the first half, right? Like, okay, so what particles are there? How do we get the infinities away, right? How do we break symmetries? But it was all in the framework that was set up. And then, you know, Now the 21st century is here, and we're kind of like, yeah, progress is not quite as fast as it used to be. It's not our fault. You know, there is something in science where there are low-hanging fruit to be picked.

1:46:08And those 50 years in the first half of the 20th century, we finally built a ladder tall enough to pick the juiciest fruits. And there could very well be a revolution tomorrow that makes it just as interesting. but there's nothing in the data that are forcing us to it, right? I mean, that's when the revolution happens, is that when the data don't quite let you become sanguine. You know, people tell stories about how in the last years of the 1800s, people thought, oh, physics is almost done. It's not really true. Some people thought that, and honestly, they had good reasons to think it. But the super smart people were like, you know, there's a couple of loose threads that are going to unravel the whole picture.

1:46:55And then it happened. Is it related to sort of the ease of experimentation? Because it seems like now, in order to do experimental verification of fundamental physics, you need ever larger particle accelerators. And it's like asymptotically more expensive, it feels like at least. I think it's partly that, but it's worse because there's two things going on. One is, as we understand more and more things, yes, to go to the extremes, to the places where you don't understand something becomes more and more expensive. Bigger telescopes, bigger products, whatever. But the other thing is, the thing that is kind of surprising and we didn't have any reason to expect, the theories that we already have extrapolate really well.

1:47:41So not only do you have to go further, but you go further and you don't find anything because your theories already work. That didn't used to be the case. And, you know, think of it this way. All these wonderful discoveries of, you know, pions or cosmic rays or whatever, radioactivity, even like radio waves and things like that, these were discovered at some point. We didn't always know they were there. But they were there. Like, they were just invisible to us, okay? and now we found all the stuff that's there, like lying around, except for arguably dark matter, which is different, but the dark matter doesn't bump into us.

1:48:17Otherwise, we would have discovered it already. It's hiding. It's there. Yeah, so that low-hanging fruit has been picked and our theories are really good and that's no reason to stop or give up or et cetera, et cetera, but it's just a reason to not be too hard on ourselves. We're trying our best. And some of us individually, like me, are changing our research focuses because it's hard to make progress in those areas. Do you think the frequency of such big discoveries is basically just going to keep going down? And do you think there is some limit to it eventually? I think that flashlight of attention moves.

1:48:57I think that fundamental physics was ripe for revolutionary discovery in the first half of the 20th century. and we did it still living in the ramifications of that now it is harder so the pace we again the people and the experiments are just as good or better than they ever were uh it's just harder to make something that is truly revolutionary like i said it could change tomorrow but i wouldn't place a lot of credence on that um but complexity biology computer science uh there's economics for that matter. There's plenty of other places to make hugely important earth-shattering discoveries. So the rate of discoveries total is going to go up.

1:49:40I'm pretty comfortable in predicting, but the core theory of particle physics and gravity is going to remain for a while. Is physics still a clear win-win, given how hard it is? Because obviously it benefits the world if we have greater understanding of how physics works. But given what we hear about how life can be in academia, how competitive it is, and so on. The personal cost that comes with dedicating your life to this. Is it, in your experience, been a win-win? And do you think it still can continue to be that for new people coming into the field? Yeah, you know, it's a win-win for the winners.

1:50:19So I'm not, it's not fair to ask me in some sense. But let's point out two things. One is, most physics is not fundamental physics. Most physics is doing great. Kinesis matter physics, plasma physics, astrophysics, biophysics, like many areas of physics are still growing like gangbusters. Within the specific field of, you know, taking the standard model of particle physics plus general relativity and trying to improve on it, that's a really tough game to play right now. And I think, you know, the students are not dummies. Like they see that, right? And so, but there's also a time lag. Like you grow up reading books written 20 years before the research you do right so i grew up reading books that were still in the heyday of discovering new particles and particle theory and things in cosmology and and so forth and so like those are the things i wanted to do and i think i've done a little bit in them but it's just hard to do a really um earth shattering thing um but but we'll have to see i you know i think that i i don't ever if if someone is dedicated to understanding this i i want them to know how hard the route is going to be uh tough road to hoe i think is the correct expression um but if there's nevertheless into it i want to encourage them as much as i possibly can i also want to make sure that if they're thinking about that they're making an educated choice that they know there's other exciting things out there right um and if they do that and still at the end of the day like nope i want to get the grand unified theory or the theory of everything or whatever i know it's going to be tough um i know that the evidence the experimental input is not guiding us very much but it's my thing i want to do then good i mean there's no other reason to ever become an intellectual or scholar than that i am curious about your stacking of like theories that maybe i or live would know of where you think, okay, here I put the most nines in my ranking.

1:52:24And these ones actually are currently still accepted, but I would be very unsurprised if in 20 years, maybe even they're already going to be put by the wayside or explained better through something else. So I'm curious about some of the tops and some of the bottoms. Great. Yeah. That's a great set of questions. You know, I do think, as a structural realist, I do think that we can overthrow the underlying ontology for theory while keeping its predictions and structures. So the standard model of particle physics plus general relativity, what is called the core theory in the big picture and elsewhere, is not going away.

1:52:59That's many nines there. It's not complete. So it will be added to 100%. And it might even be tweaked around the edges because it's based on quantum mechanics itself. and I'm open to the possibility that quantum mechanics is not exactly right, but I would put low credence on that. I think that quantum mechanics, quantum field theory, within the energy and scale regimes where it's supposed to apply, that's going to keep applying. Those are S-tier theories. That's right. That's right. The general framework of the Big Bang model from one second after the Big Bang to today is going to stay there. That's not going away.

1:53:39The general framework of relativity is not going away. The fact that there, the claim that there is dark matter in the universe, I would put at well above 90%, but not 100%. People try to replace it. I've tried to replace it. It's a game people play. But the amount of evidence that is accumulated in favor of dark matter is overwhelming. In fact, I will give advice to listeners. If anyone says, I have a theory that will replace dark matter, and the next thing they start talking about are the rotation curves of spiral galaxies, you know you don't have to listen to them. Because that was the best evidence for dark matter 40 years ago.

1:54:19But now you have to talk about the cosmic microwave background, large-scale structure, and things like that. So if they're not talking about that, they're not really worth taking seriously. See, dark matter, 98%, 99%. Dark energy, well, there's only two. The fact that the universe is accelerating is many nines. That's been established. Is it dark energy, or is it modified gravity? it's almost certainly some kind of energy. I've tried to modify gravity again, you know, and it just makes things worse. But it's still, that's relatively new, relatively young. Let's put it at 97 % for dark energy. Is the dark energy strictly constant, Einstein's cosmological constant, or is it slowly changing over time?

1:54:59Again, even just 20 years ago, I would have put a lot of credence on changing with time, but the theories have not panned out and the data are still happy with the constant. So I'm going to go at least 93 % credence that it is a cosmological constant. And at the bottom end of the spectrum? Yeah, so I'm working my way down to more speculative things. So cosmic inflation is a very, very popular idea for what happened in the first fraction of a second in the history of the universe. I think it is actually quite speculative. I think that it works quite well. It's very attractive for various reasons.

1:55:38It also has problems. We don't have any alternatives. As a good Bayesian, our propositions to distribute credences over are inflation versus something we haven't invented yet. There's another 2 % for actual bad theories that people have invented. But I would give inflation like 50%. Many of my colleagues would be at the 90 % level, but I'd be 50%. What would you give to string theory as currently understood? As I understand, it might not be testable anywhere. We are struggling with finding a way to ever make testable. Yeah, string theory is a very interesting case. I think it is certainly the leading candidate for quantum gravity right now.

1:56:14That might not be saying a lot. Maybe we just haven't thought of a better candidate yet. But the other reason which makes it hard to attach accredence to is that the theory keeps changing as we learn more about it, right? So it might be that ultimately it was the correct idea, but it manifests in a completely different way than we expected. so I'm just very reluctant I would be between 50 % and 90 % for me and string theory and again I know colleagues who put it at 99 % what I'm noticing as well is that the theories that you described to be in the S tier category are also the ones that you're actually writing about to a significant degree which is very fortunate because I think one of the also potential win-wins actually for people in learning physics is that you will be less confused about like everyday things so like learning the things that are very likely here to stay allows you to have world models that are gonna be worth their salt for a longer amount of time than something else that you're trying to build it on no that's absolutely true i will say that one of my more idiosyncratic credences is in many worlds where i would put it at more than 95 percent um and few people would do that but i think that just compared to the alternatives, that's where I would put it.

1:57:34Particularly in, you know, and I've written a whole book about many worlds, right? Something deeply hidden. But the new books I'm writing, the biggest ideas in the universe books, the whole point of those is to write things that will still be true 500 years from now. So I'm sticking to things that are at 99 % and above, almost exclusively. I'll hint at things. It's going to be harder to do in the complexity book because that's less well understood, but I'm going to try my best to stick to things that are not going to go away. Thank you. My pleasure. Thank you for having me in the Purple Palace.

1:58:06So there we go. Thank you so much for tuning in and huge thanks to Sean for this conversation and my occasional co-host Igor. Do check out the notes below for further reading, including links to some of Sean's more recent papers, his research, also some of his books. I highly, highly recommend his book, The Bigger Picture. It's just one of the most beautiful and comprehensive explorations of the intersection of physics and philosophy. Also check out his podcast, Mindscape. I was actually one of his early guests on it years ago, and it was an amazing conversation. Unsurprisingly, he is a fantastic interviewer.

1:58:43So check that out. It's over on YouTube, again, linked below. Thank you, and I'll see you next week.

1:58:56Thank you.