Future of Science and Technology Q&A (January 5, 2024)

Hello, everyone. Welcome to another episode of Q and A about future of science and technology. I am not in my natural habitat today, so if strange things go wrong, apologies for that in advance. All right, let's see. We have a bunch of questions saved up here. One from James. What scientific breakthroughs would you like to see in 2024? I think the most exciting breakthroughs will always be the ones that we don't expect are coming. There are ones where they are sort of long anticipated and then they finally arrive, and there are ones where kind of nobody saw that coming. In terms of things that I'm personally involved in, there are. It's kind of I. We're at a. Just an amazing point because a lot of things that I've worked on for a really long time have sort of come to fruition recently, from our physics project to all the implications of the physics project, the ruliad, all these kinds of things. There are several pieces of that that I'd love to see make big progress in 2024. One of them is what we're calling infrageometry, which is kind of generalizing the ideas of how one constructs the mathematics of space, calculus, things like this, starting not from the traditional idea of continuous space that has existed since Euclid and so on, but rather from the idea of these graphs, networks, hypergraphs, making up the underlying stuff of space. And there's kind of a different development of what ends up with things like calculus and differential geometry and so on, that starts not from the idea of a manifold and Euclidean space and so on, but rather from this underlying structure of networks. That's one area where I hope to see big progress made in the course of this year. And it's kind of interesting because it's kind of a rebuilding of all of the notions of calculus people study, for example, calculus of one variable, two variables, et cetera, of the things we need to be able to construct is calculus of two and three quarter variables. It's just something for which the mathematics really doesn't exist. But it's a development that we are progressing with and I hope this year we'll make lots of progress on. Another piece of that story is things about our physics project and figuring out more about how to connect the underlying machine code of the universe, as our physics project suggests it, with the actual experimental observations that can be made. And it's a complicated road understanding sort of what implications are there for black hole mergers, what implications are there for the cosmic microwave background, and so on from our physics project, what does A dimension fluctuation look like in cosmology, astrophysics. What might we be able to say from our new view of quantum mechanics about the things that are done in quantum computing? One of the things that I think is quantum computers are conceptually interesting, a good practical way to push the development of computational hardware that isn't based on semiconductors and electronics. But I think there are sort of problems with the sort of the interpretation of, with actually the sort of the quantum flag for the things that people talk of as quantum computing. My own guess is that the things that go wrong, the decoherence, the errors, all this kind of thing, that actually there's a lot of interesting information there that may very well be very much informed by the kinds of things that we're doing in our physics project. Because our physics project goes underneath the traditional view in quantum mechanics which says, okay, and now you go from all these probabilities and quantum amplitudes and so on, and you make a measurement. And that's just one kind of discrete act. In our physics project, one's able to sort of take apart that measurement process and understand it in more detail. And I think that what we see as we compare that with experiment may well actually reveal features of what's going on that sort of allow one to see into kind of the discrete multi way structure of space and time and quantum mechanics and so on, which would be, which will be exciting. But seeing how that works is a challenge and of course it will be really neat. I don't expect that this year to actually make the connection between our models of physics and discreteness of space, time and so on, and actual experimental observations. I don't expect that to happen in a time as short as a year. I think in a year we may know sort of what there is to look for and then the actual mounting of the experiment is a non trivial matter. I mean, if we're lucky, there will be things that have already been observed long ago, but we can now just say, oh, we understand that now. So other things that I'm kind of hoping will emerge in the next year, perhaps some progress on observer theory, which I've just recently been talking about and understanding kind of how sort of what observers like us are like and how much of what we perceive in the universe is a consequence of our features as observers and being able to home in on what is the sort of overall description of what observers do when it comes to observers with human sensors or observers with measuring devices and so on. That's another place where I think progress can be made and where the leverage is very high. Because I think what seems to be the case is that just two attributes of us as observers, that we're computationally bounded, that we believe we're persistent in time, just those two attributes are sufficient to give us the big theories of 20th century physics. So as we add more attributes that are sort of characteristics that we as observers have, I suspect we will learn more about sort of the inevitability of laws of physics as we are going to perceive them. So other kinds of things. Well, gosh, there's so much from our physics project in particular, I think two areas where I'm hoping for progress. One is biology, the other is economics in biology. Sort of the big issue as far as I'm concerned is what kind of overall theory for biology can one get that is kind of is a, is a kind of a large scale computational description of what's going on? You know what, when you look at molecular biology, what is the kind of, what kind of thing is biological tissue? It's not like a liquid quite. It's something where there's this kind of bulk orchestration of molecular processes. And I'm hoping we'll make some progress on that in the next, in the next year. In very different area of economics, I've long been very curious what sort of a foundational theory of economics might look like. And this kind of whole structure of lots of events in space time interacting with each other, that seems like something that will be. That's kind of like all these events in space time are kind of like all the transactions happening in an economic system being knitted together. And then the question is what sort of the aggregate behavior? In the case of space time, we think it's things like gravity and the Einstein equations and so on. What are the analogous things in economic systems? So there's a couple of things there. There are other directions. There are a lot of old chestnuts which I think have the chance to be. I don't quite know what you do with an old chestnut. Maybe you boil it and eat it or something, I'm not sure, but resolved in some way. I mean, another sort of collection of old chestnuts are things like computational complexity theory and the P versus NP problem and so on. In our kind of view of the ruliad and this kind of structure of all possible computations that the P versus MP problem becomes essentially something which is a geometrizable problem. It's something which can really talk about the structure of rulial space. And I don't Know, I'm not sure that's quite on the agenda for the next year, but it will be really nice. I think there is the potential to make progress on that kind of old chestnut. Well, at least what is it from about 1970s, so it's a 50 year chestnut, so to speak, that I think we might be able to make progress on with the techniques that we now have. I think there are are other things that are a little bit more longer timescale than the year that having to do with kind of understandings of distributed computing progress made there, those kinds of things. Another big area in sort of things that could happen in the span of a year. LLMs arrive pretty much in their working form as ChatGPT over the last basically year or so. I think the big science question there is how do they work and why do they work? I've tried to analyze this a bit, but there's vastly more that can be done. To what extent can we find, for example, semantic laws of motion, given that you said these things, what is the inevitable sort of progression of things that make sense that you have to say next? The LLM is finding the things that you say next that agree with kind of the training that it's had from all those things it found from the web and so on. But the question is what kind of theory Are there laws of motion analogous to the laws of motion and mechanics that say in meaning space, if you're going in this direction, you're going to have to inexorably go like this. Is there kind of a semantic grammar of language that gives you a kind of construction kit for language? But that's analogous to the kind of noun follows verb follows noun that we have in kind of the syntactic grammar of language. So sort of LLM science is an area where I think lots of interesting progress could be made, kind of really figuring out in the end, is there a narrative description of what's going on inside an LLM beyond just saying, oh, there are all these complicated layers and transformers and this and that and the other, and it was all mushed together with training and that's how it works. Is there something more narrative that can be said? Is there something more science like? More, you know, science is really the translation of what exists out in the world to things that we humans can think with, that we humans can kind of understand and have a narrative about. Is that possible in LLMs? I think the answer will be to some extent, yes. And I think that will be an important thing. And I think that's a fairly short timescale thing. I think when we understand that, it will probably tell us a lot about neuroscience. Because in neuroscience, sort of one of the biggest challenges has been this. We just don't know this intermediate level between sort of a bunch of neuron firings and the level of sort of psychologically describable thoughts. And I think that sort of intermediate level is the kind of thing that we can now get to explore with LLMs. LLMs are, I think, an emerging, really quite good model of neuroscience, not at the level of individual neurons and what they do and the details of the electrochemistry of neurons, but at some bigger level. In the same sense that a Turing machine is a model of computation. It doesn't have the details of electronics and field effect transistors and things like this, because that turns out not to matter in understanding broader questions about computation. So I think the same will be true about neuroscience versus LLMs, other areas of science where progress might get made. I mean, I mentioned things in physics and there are some old chestnuts there, like the 100 year old chestnut of what is dark matter? Why are the rotation curves of galaxies seem like there's something in galaxies that isn't the luminous matter in stars and so on. What is it? How does that work? I have some guesses from our physics project about how what I've been sort of tentatively calling space time heat might be kind of the thing that leads to the effects on space time that we right now would describe as dark matter. In that interpretation, it's not really matter in the same sense that other things are sort of made of atoms that you can pick up and so on. Other areas. Well, I mean, there's a lot of areas in biomedicine where slow progress will get made. I mean, I think the. It's interesting that, for example, an area like aging, I sort of was interested even from when I was vastly younger, 40 years ago, whatever of, you know, what is aging, how does it work? You know, will we be able to solve it, so to speak. And nobody was working on it back in those days. Gradually, aging has become a more and more popular kind of topic to study. And there's sort of slow progress in particular directions. But in terms of sort of a big theory of how that works, we don't really have that. There are sort of potential poke this and you might be able to sort of reverse the clock because we're sort of understanding there's an inexorable biological clock. And you know, for example, the kinds of things that were discovered, what, a decade ago or something. Now to go back, go from any cell, you have to back to a stem cell that's capable of differentiating into any other type of cell. The things that one can kind of use to do that, maybe those things are things that can sort of reset the clock for aging. One doesn't know yet, but there are potential breakthroughs that could happen along those lines where it's like, I mean, that would be a really cool thing if it ever happened. Sort of the elixir of eternal youth type thing. It's just like here's this thing and, and oh well, the way biology evolved and because in fact, in evolution, the fact that the old fogies die off is probably a good thing for the species. So it's something that will be selected for in biological evolution. Even if the sort of immortality might be cool for the individual. At least in the course of, in the past it's been sort of bad for the species and maybe it's even bad for the species now. Maybe if we reach human immortality, not nothing new will ever happen in the world. Because everybody will say, well this is how I've been doing it for 6,000 years. And you know, really I'm happy doing that. I don't want to change anything. It's always a complicated thing. You know, the fact that it's often this when something is sort of successful and lasts a long time and so on, and everybody thinks that's a big success, it sort of builds inertia that makes it more difficult for something new and different to happen. So it's always a, a complicated multiple edged sword. And certainly I'm sure it has been for biological evolution. But that's obviously a dramatic breakthrough that I don't think we can know when it will happen or whether it's even possible. It could be that it's like an operating system for a computer which gradually develops more and more cruft and eventually it crashes and you have to restart your computer. And maybe that's sort of an inexorable feature of systems like that and like us in biology, so to speak. There are other things. Like cryonics is one thing that I could imagine. At some moment it will just be solved. Just like mammalian cloning was at one moment just solved. That some kind of, oh, if you do this and you tweak things in this way and you give the electric shock at just the right moment, then you can clone mammals. I could imagine the same kind of thing in terms of cryonics being able to kind of hit the pause button for human life type thing. But let's see other areas where there might be dramatic progress. I mean, I think. There are always things that are sort of out there, like molecular scale computing, which is very similar to molecular biology and possibly related to all these questions about theories of bulk orchestration and so on, but possibly a little different. It's like, how do you make that universal constructor, that was a big topic in the 1980s, really died off of, can you make a thing that will just start assembling things out of atoms? And, you know, we can do that with an atomic force microscope. We can pick up a few atoms and put them in definite places. But that's not a thing that scales up very well. Can you make something where you just feed it the plans and it just goes off and it spends, you know, a few hours and it's just gradually, much like protein synthesis happens in biology, it's just steadily making a thing out of carbon atoms, for example. Eventually it will be figured out how to do that, and that will probably be. And once you have a universal constructor like that, the universal constructor can clearly construct more universal constructors. And it's kind of off and running. And that kind of notion of doing that is something people have talked about since the 1940s. And as I say, it got a bit popular in the early 1980s and then died off. Kind of absorbed into some combination of chemistry and material science, which went off in rather different directions. What other things might be coming? Well, you know, there's always the unexpected. I mean, if we look at different areas of whether it's in medicine and, oh, did you know that if you just do this, you solve this big problem, for example, to an area, well, like physics, where there's a question of why do things work in this particular way? Or we can finally find this theory that underlies that and explains how things go on. I'm trying to think what other areas? If people want to ask about them, I'm happy to talk about them. I think there are areas that will show sort of slow engineering progress, like a lot of AI kinds of things. I think another area that is sort of a question is sort of will AI magically solve science? I think the answer is no. I think that there are sort of. What does that even mean? Will it be the case that there is sort of a transformation in the way that certain kinds of science are done? I don't think so. I think that a lot of what we see in terms of just search Some big space of possibilities that's much more of a traditional computational problem than it is a more AI ish problem. Things like read the literature, see whether there's something that you can pull together from all these different little threads and shreds in the literature that gives you some definite thing. We've already been able to do a surprising amount of that just searching literature. Even before the web one could search, there were online database systems that would search the abstracts of all scientific papers, things like this. So that's been a sort of a slow, a slow progression. I don't expect to see something where it's like particularly dramatic. There will be things that are, yes, it's very useful. One can do this and that. But I don't think it will be kind of a wow, breakthrough kind of thing. I think there are probably, well, there are things where, I mean, I think the real question is what are the old chestnuts and which ones might be cracked, whatever one does in, in short order. And I think, you know, two of the big methods that have emerged, and that's always what lets one crack old problems as new methods. Two methods. One is the LLM story. The other is the story of things that have emerged from our physics project. Those are two sort of big methodological things, I think, that have emerged in the last few years, and they certainly have. Neither of them has been fully worked through. The low hanging fruit still has to be picked and on a timescale of order a year. I think in both those cases there's a bunch of scientific low hanging fruit to be picked. Let's see, Question here about graphene from Giovanni. You know, graphene is one of the strange things you can make just out of carbon atoms. You know, people used to think, oh, the only things you can make out of carbon were, you know, pure carbon were diamond and graphite. And then buckyballs were discovered. These kind of clusters of 60 carbon atoms and more. The remarkable thing about buckyballs is they've always been there. They've been in soot forever. They just hadn't been identified. And it's sort of an interesting, like many science discoveries, it's kind of a thing where it was hiding in plain sight for a long time. Graphene is kind of the tubular version of a buckyball. And yes, there are, you know what's great about it, it has high electrical conductivity. It is something from which you can make electric circuits. People have slowly started to make transistor like things out of it. I am not on a sort of A month by month basis or even year by year basis. I don't know what the latest developments have been there. My impression has been that it's going to be a slow progression of adding more sort of engineering possibilities to what can be done with graphene. I mean, people talk about, yes, it can be a very strong material. You could make a, you know, a space elevator out of it maybe, possibly you could make electric circuits that have all sorts of special properties out of it. These are things which are. My impression is they're in the slow engineering development, improve the technology type phase. They weren't blocked in any way, but they also weren't. Weren't like, you know, they're not at the stage where major breakthrough is needed. They're more at the stage of engineering development. I would say another area and same situation is probably fusion power where it's like, can you build a tokamak? Can you figure out, can you build this magnetic confinement mechanism for the, for the 10 million degrees centigrade plasma that has energetic enough particles that they'll undergo fusion? Can you contain the thing? Even if you can contain it with some kind of big magnetic field bottle, if it starts showing fusion and starts generating lots of high energy neutrons, for example, even the things that generate the magnetic field, the magnets, will they just be destroyed by the neutrons? And is it kind of game over at that point? These are engineering efforts that really are sort of the, you know, typically what happens methodological advance, rapid progress. Then sometimes there are things which just need a lot of engineering grinding over the course of decades to take them from. Yes, it's in principle possible to. It's routinely doable. Gosh. Aaron asks about bio vehicles with synthetic adipose tissue. I do not know about that. I think that the possibility of sort of a breakthrough in electrochemistry of, you know, lithium ion was sort of a breakthrough long ago. But the possibility of a different chemistry that will be a breakthrough is, I think there's nothing to say that's impossible. There's nothing to say it's impossible to find a substance which will provide photovoltaic performance that's vastly higher than silicon does now. And these are things where sort of searching for materials, it's like the quintessential Edison searching for the filament, which ended up being, in his case, carbon for the light bulb. You know, when people have tried for decades sort of just search molecular space for things that are good for, you can name them, good photovoltaics, good high temperature superconductors, potentially Things for batteries, although those have been a little bit less systematically searched for. And the big one, drugs searching for molecules which have particular biological effects. All of these efforts have been sort of partially successful. People are always saying I'm going to use AI to just win every battle there. It's never happened. I don't think it will happen as such. I think it's just a slow, steady progress with occasional wow, we found something unexpected. I mean, in my own efforts in sort of studying the computational universe of possible programs, there's this occasional moment where you find something totally unexpected and where it's a major piece of progress. And I think that could certainly happen in the sort of material world of molecules and so on. Although it hasn't been a thing that's been sort of big dramatic success. There are now, well, there have been for many years a whole sort of rotating series of companies that have grown up saying now we're going to be able to do rational drug design, combinatorial chemistry, sort of AI based chemical discovery, molecular discovery, you know, incremental progress, but not drama at this point. Let's see. And in terms of, you know, different technologies for batteries, I don't know, could happen. You know, what always is the case is that you get a dramatic new idea for some technology and then there's an incumbent technology. And the incumbent technology has had a lot of engineering development, 50 years, let's say, of engineering development and often it's actually pretty rare that the new entrant is immediately better than the old one. I don't, oh, when transistors came in. Vacuum tubes have been developed for many years, many decades and I don't think there was an immediate oh yeah, transistors everywhere. There were places where transistors made an inroad and then microprocessors got developed and so on. But you know, when you've got something where there's been a big long sort of period of, of detailed development that usually initially wins out. You might see that the transistors are on a much more rapid slope and they're going to take over, but it still doesn't happen immediately. And you know, it's always a big surprise when something happens like ChatGPT where that was sort of so much better that it sort of was an immediate thing. I would like to say that things like Wolf Malpha have, have served in the same way of being sort of. There was, there was. What was before it in those directions was very weak and then it was able to deliver something surprisingly strong in, in 2009 in that case. But you know, that's, I would say a lot of the time, things that have had extensive engineering development, the new entrant doesn't immediately win big. And you really have to sort of watch the curve to understand that it's the thing that's going to win. I think that's a. You see that for example, with our computational language, which is now 35 years old and although greatly developed over that period of time, but you can kind of see that that gradually takes over different areas and different kinds of things one thinks about as compared to for example, low level programming languages. But that's a slow, long, complicated thing to sort of relative to the development that's been done on low level programming languages. And particularly in that case the kind of, the sort of. The incumbent knowledge about those kinds of things and beliefs about those kinds of things. Let's see, I was asking, do I think the traditional engineering disciplines will be relegated to some kind of LLM? You know, will they be automated with LLMs? You know, the story of many things I've tried to do is automate what used to require lots of human effort. Whether that's calculating integrals in math, whether that's solving lots of kinds of computational problems. Automation is one of the big drivers of progress. And yes, there'll be lots, there'll be more of that. Will it be dramatic and sudden and now we can suddenly solve all engineering just by pushing an AI button? I don't think so. I think that what will happen is that the sort of the things will just gradually rise. It's kind of a rising tide of possibilities. And what becomes possible to design, becomes possible to do gets ever more complicated and that will allow one to be ever more ambitious. I mean, as computer aided design, Computer aided engineering have come in, the complexity of the devices that one can build has vastly increased. The actual challenge of building them remains sort of a fixed challenge of what people can figure out how to do and how to define and so on. But what actually gets built is much bigger because of that layer of automation. Let's see, unsupervision asks, do you think we'll see mass producible room temperature superconductors in the next decade? I do not know. Nobody's been able to make a room temperature superconductor. Is it theoretically impossible? Absolutely not. There's nothing that says it's theoretically impossible. What material might make it possible? We don't know. It's kind of every so often somebody will come up with some exotic material that Involves pieces of lanthanum or pieces of whatever else. I forget what was. There was a recent one. What was that one? Oh, I don't remember. It used a rare earth metal of some kind that was a bit exotic and one hadn't heard much of. Wasn't homeion, but it was something like that. And there is a lot of chemical space, molecular space, that is unexplored and that's where the opportunities exist to find sort of surprising things like robust room temperature superconductors. If there were robust room temperature superconductors, there's a vast range of technology that becomes possible. But we don't have them and we, you know, we, we don't know how to make those. So Dougie is asking. It's been suggested AI will replace coders and programmers. Do you think AI might also replace many physical and chemical experiments? Interesting question. I was our longtime kind of partners at Emerald Cloud Lab who've been building this big Wolfram language based sort of lab in the cloud that sort of robotically does experiments that you specify in computational language. That's an interesting question, whether you actually need to do the experiments. The answer right now is a definitive and resounding yes. The fact that you can do the experiments, and this is something that I think will be sort of this concept of cloud labs and so on will I think be quite a practical breakthrough in doing biology, chemistry and so on. Experiments that instead of sort of everybody has to build up their lab and spend a year building a new lab and so on, you can just use the cloud lab and just type in the specification of what you want to have happen. And the thing is already set up. It's kind of like cloud computing. You don't have to build your own servers, but building your own servers is a lot, lot, lot easier than building your own biology or chemistry lab. And so I think that is a practical matter of what is possible to do in biology and chemistry and so on. Research that is potentially a major driver of progress. It also has another feature which is you sort of, you have a born reproducible experiment. It's like, here's the script, just run it again. It better do the same thing. It also gives you kind of a formal description language. The fact that such a script exists and can run those physical experiments gives you a formal description language for experiments, which is very useful in kind of understanding. Well, I take this experiment, I tweak it a bit, I take this experiment, I rerun it, et cetera. The question of whether you actually need to run the experiments in the end, right now, the answer is very much yes. Interesting question. Can you take many runs of those experiments and kind of machine learn how the experiments work? The answer is probably somewhat. How far that will get, I'm not sure. I think it will be hard to find surprises that way. I think what machine learning is good at doing is kind of, you know, fitting things the way that they sort of naturally go together. Which means if there's a big sort of poke out surprise, you don't tend to see it. You do see that if you do things like exhaustively search all possibilities, then one of those possibilities ends up being the thing you never thought what it was going to do and it does something really surprising. But machine learning tends to be a thing that kind of smoothly interpolates between what's already known and it doesn't notice these sort of big spikes that poke out. So a question here from PRAB about zero knowledge proofs. So I do think that's an area where there probably will be progress made. I mean in cryptography was a rather obscure thing back in, even the beginning of the 1980s was sort of, oh yeah, it's this thing that gets done, but it isn't really a thing that's sort of in the public consciousness very much. And then as the web developed and things like this, cryptography became important and big. But public key cryptography had come on the scene. The ability to do things like make digital signatures, the ability to publish a cryptographic key that allows people to send you data, but doesn't allows only you to decode the data. And then there's a vast number of protocols that have been developed based on that, like the HTTPs protocol for validating websites or the protocols for digital money. All these kinds of things. These are all protocols that became possible because of public key cryptography. Now one of the big questions is to what extent can you take data, encrypt the data so nobody can find the whole thing but sort of poke at the data and ask specific questions and get those questions answered without having to decode all the data. So for example, let's say you have a bunch of medical test results for lots of different people and you're doing a study where you just want to know, does everybody who you know gets some, oh, I don't know, does everybody who gets flu have a, you know, is there a higher probability of them getting this other thing sometime later or whatever else? I don't know what. And that might exist in all these medical records, but it's for Privacy reasons. One's not going to reveal everything that's in these medical records, but if you could have all the medical records encrypted, but just be able to poke at them and say, just give me this one fact across 10,000 different medical records, I don't want to know anybody's individual results. I just want to know the aggregate distribution of whether people got sick or whatever else it is. And can one do that? Is there a way of doing that? A common buzzword is fully homomorphic encryption of whether you can kind of whether you can exchange the operation of encryption and the operation of computing something, whether you can compute sort of through the encryption, so to speak, without having to decrypt first, make a plain version, then do the computation. So there's been slow progress of this. It is theoretically, you know, there are theoretical results. There are a few, you know, there's buzz terms for this. There's a little bit of work that's been done, particularly in the blockchain space on zero knowledge, ZK snarks and things like this. There's just a whole collection of mathematically quite sophisticated things. You can find nice Wolfram language implementations of a bunch of these things. I think there are sort of two directions. One is making those algorithms more practical, and the other is figuring out protocols that can make use of these things. I mean, it's worth remembering that with public key cryptography, one might have said, oh, this is never going to go anywhere. It's too difficult to encrypt megabytes of data with public key cryptography. But that turns out not to be necessary. All you need to do is deal with the sort of distribution of keys using public key cryptography. And once you've got the key, you can use a more traditional cipher to do the encryption or decryption. But the protocol exists to know how to do this key exchange using public key cryptography and achieve certain sort of objectives in terms of the way that secrecy is maintained and so on. And I think that's an area where there will be progress in actual progress. There might be some breakthroughs. There will be a higher. I was just trying to remember, in the last probably few months, I've seen at least a few products that rely on zero knowledge systems, and they're not yet practical out there in the field. I think those particular products. But the fact that there are products starting to be developed that rely on these ideas I think is interesting. Juan is asking, given some of our greatest accomplishments as species have been when we mimic nature, how important do you think biomimetics is going forward? You know, there are meta lessons of biology, like natural selection. We do natural selection on our technology, as happens in biology, so to speak. We learn things at a sort of meta level, then we learn things at a much more detailed level, like, oh, can we create molecules that have this particular feature? Because look, biology does something like that, do we end up using. There was a push for a while to do sort of DNA based computing, and that's really panned out for a whole variety of reasons. But that's kind of like take what biology has already figured out about DNA and leverage that to build a kind of technology stack on top of it. I think some of these cases where people say, oh, we're going to imitate the eye of the horseshoe crab or something to do this or that thing, those tend to be more in the nature of sort of inspiration about how things could work than a detailed, let's go measure horseshoe crabs to see what we should do. Have we mined everything we can mine in biology? Definitely not. There's a huge amount in molecular biology where we don't understand the mechanisms yet and where when we do understand them, we have the chance to kind of make our own versions of them. When it comes to large scale biology and kind of, I don't know how we might make robot arms based on the, you know, the bones and muscles and tendons and so on of our sort of biological construction, I would say we're a decent distance of the way through mining that knowledge. But there's plenty of biological knowledge we just don't have that we haven't been able to mine yet. Kerbal asks, can you see a time when the discovery of new mathematical theorems and axioms will be generated from AIs? It's an interesting question. Looked at it a whole bunch. The real thing is it's pretty easy to generate infinite numbers of mathematical theorems given a certain set of axioms. You just keep applying them. You generate this big entailment cone, one might call it, of things entailed from those axioms. The real question is of all these theorems, most of them, when you look at them, you just kind of roll your eyes, you say, I don't know why I'd care about that. So there's this question of what constitutes a theorem. One should care about an interesting theorem. Can one, for example, machine learn what we humans have traditionally thought of as interesting? Can we pick out the interesting theorems in some automated way? I think that's not hopeless. We've tried to do it a bunch and we haven't been very successful at it. I don't think that's abjectly hopeless. There is a sense in which interestingness is determined by history and in which kind of something is interesting. If it was studied historically and led to this, and led to that and has a certain context that people understand, that's a way in which that's sort of a generator of interestingness, which in a sense it has to be lived by the human mathematicians in order for it to be interesting in that kind of direction. It isn't good enough to say, oh, the AIs discovered this. The human mathematicians are just going to shrug and say we don't care. Because in order for us to care, in order for us to quote, understand this in a way that sort of merges with the things we already know, we have to pull it into our actual experience, so to speak. So I think this question of whether machine, whether an AI can sort of pick out attributes that it's noticed from the few million theorems that humans have published in the literature of mathematics, it's noticed. Humans tend to like this, let me make it work like that. It's kind of like image generation systems where some of them have been trained to be make pictures that are attractive to humans and yes, they can do that to some extent and you know, make theorems that are attractive and interesting to humans. Now, when it comes to axioms, that's a much more complicated issue because axioms are sort of what you assume is true and you're then going to build up from. But the question is, how do you pick axioms? Not very many have been conventionally picked in mathematics. So things like the axioms of Pyrna, arithmetic, the axioms of set theory, the axioms of geometry. They're a small list of these axioms that have been picked for the progression of mathematics as it has existed to today. When I wrote my book A New Kind of Science, I decided to put on, actually, it ended up being two pages, all the axioms of mathematics as currently practiced, including actually some other quite obscure ones. They all fit in those two pages, admittedly in a rather small font, but. So the question then is, okay, well, so are we going to have new axioms? Well, usually the way those axioms were derived was here are some things we think are true, either intuitively think are true or just we think are true. You know, a point has no extent essentially an axiom of Euclid. The, you know, two parallel lines never meet an axiom of Euclid that turned out not to be true about physical space. But these are things where you go from sort of your common experience or the things that you think of being. Wanting to be true, then you make up axioms, then you say, well, what else is entailed by those things that I've assumed are true? So I think it's very difficult to say, you know, except insofar as one sort of capturing what humans have typically thought of as true, then that's. That. That's kind of the. The best we can think of to do to sort of go back. There's sort of the idea of reverse mathematics. Given that you have a particular theorem that you want to be true, what are the minimal axioms that give you that theorem? Well, in a sense, that's what we've been practicing in coming up with the axioms that we've already come up with, which then get encrusted with huge amounts of history and huge numbers of things built on top of them. I mean, it's kind of like in an economic system, you invent a thing like Bitcoin, for example, and is it worth anything? Well, as soon as you build enough stuff in the world, this whole network of consequences and entailments and so on, then it becomes somehow a real thing with real value. And the same is perhaps true with mathematical axioms that I just invent an axiom. It's like, okay, fine, I don't know. You've got a circle operator in there, and it does this and that. You kind of shrug your shoulders. It's kind of like, so what? But then if it turns out that on top of that has built this whole collection of kind of thinking about computing or mathematics or whatever else, then that becomes something that has more of a sort of real intrinsic value to it. Aaron asks, when Betelgeuse explodes, will humans be okay? Betelgeuse is a red giant star fairly close to us that will have a lifetime of a few million years, and eventually it will turn into a supernova and explode. I think the answer to that, I think if there was a. I think the answer is yes. I think we'll get to see this very bright spot in the sky that'll last for at least a few days. I think that the particles that will come will stream out from the supernova. I'm not sure we know exactly how many such particles there are. We have probably some information. I remember reading something about this recently, and I think, you know, given that we. Given the Earth's atmosphere and Given that there aren't so many particles that the atmosphere gets destroyed, which would be bad news for us, I think we get to be okay. But that is a detailed question. I think some of what is to be said about that. When you see a supernova explosion, you certainly see things that have been. I actually don't know what the density of high energy particles from a supernova explosion would be. And are there lots of. For example, are there what actually happens? They'll probably be high energy protons, high energy electrons. There are high energy protons from the Sun. That's the dominant component of cosmic rays that hit the Earth. They make auroras, things like this. When the high energy particles are sort of trapped by the magnetic field of the Earth and interact with the molecules in the upper atmosphere. I don't know how much more intense kind of the effectively cosmic ray flux from the supernova at the distance of Betelgeuse is relative to what we encounter all the time from the Sun. Is it like a big solar flare? Is it like a small solar flare? Or is it like something we've never seen before that will cause much more trouble? I actually don't know. Good question. Weasel asks, do you think smart textiles and computing fabrics will take off or be viable? Would I wear a sweater to hear instead of a hearing aid? You know, there are always these things which people sort of say, hey, I got this cool thing from the future. Look, it's a smart coat or something. They're things which are constantly from the future and eventually arrive. Most things constantly from the future eventually do arrive and sometimes in a form slightly different than they were originally imagined. But the idea of smart clothing where sort of it does things, I don't know, the billboard T shirt, so to speak, that's continually changing what it displays and so on. The sort of the thing where don't wear a smartwatch, just, you know, your shirt sleeve will have. That will be sort of able to do the health monitoring and so on that that was able to do. My guess is eventually those things will take off. Eventually there'll be electronics that is sufficiently low power, it can be powered from body heat and things like this. Eventually there'll be sort of. I mean, I've seen lots of samples of various kinds of flexible electronics and this and that and the other. So my guess is that the general principle is what was having to be made in very well controlled electronics will be able to be made in many different media in the future. What the big use cases will be, I'm not quite sure and yeah, it's an interesting question. Let's see. Wyatt comments. Things like math, geometry, especially tessellations, have patterns that are universally implicit and can be interpreted to be interesting by their own existence and not just by humanity viewing them. Well, it's pretty hard for us to ascend, sort of transcendental our own history. I think of the what's interesting question, I think that I've sort of lived that because I've studied a lot what's possible in the kind of computational universe of possible programs and so on. And much of what one sees there is kind of lots of interesting complicated forms where we ask is this significant? Is it not significant? It's hard to say. It's a thing one discovers. Just like you go out in nature, you find some weird creature living under some rock somewhere and it has some complicated mechanism going on in it. Is this interesting or is it just the weird feature of the such and such beetle? Hard to know. And I think it's very difficult to know. What we're going to care about in the future is not necessarily what we care about today. An example I've given many times is nested fractal patterns which were created, well, they were created by nature, but they were created by humans, by mosaic layers in the beginning of the 1200s and nobody cared for nearly 800 years until the kind of whole context of fractals and nesting and so on emerged basically in the 1970s. And then people are like, oh, that's a fractal pattern, that's cool. So fractal patterns are to us today, you know, a Sierpinski pattern or something today that's kind of self evident to us. In the 1200s people, you know, a particular person, I think his name was Cosmo Cosmati invented this and knew about it and probably thought it was really cool and nobody else cared. And I think that is the what causes people to care about something is a complicated issue that is very historically related. And I think you really can't say abstractly things like tessellation patterns are going to be cool to everybody. I would say that, you know, features of numbers you could say are cool to everybody. But you know, for example, the Pythagoreans were as enthusiastic about perfect numbers as they were about about primes. As it turns out, primes have been much more important in the history of mathematics as it has actually developed than perfect numbers have been. But I don't think one could say that as an abstract matter. But I mean this whole question about what is interesting the theory of interestingness. And is the nested pattern, is this particular geometrical pattern actually interesting question? I don't know. Do chimpanzees find sort of tiling patterns engaging and interesting? Do they, can they readily predict how the pattern will continue? My guess is yes. If you feed them nested patterns, can they readily predict how those work? I'm not so sure about that. And then what about other kinds of animals? You know, can you. I don't know how you do the experiment with dogs or something, but I bet there is a way to do it of kind of, you know, the. Do they expect the tiling to go on this way? Do they find the tiling preferable to the kind of random arrangement and so on? I don't think, I think that's a. You know, those are difficult to answer questions. It's very difficult. I've been very interested in this question about how do we transcend our own history and our own detailed features to see what is sort of intellectually possible beyond that, how to do. Let's see. Juan asks, is there a way we can use Brownian motion of molecular scale as a type of fingerprint for nanosensors to create things that are piracy proof? Yeah, I mean there's a whole movement of so called physically unique functions, puffs which are kind of. When you make a semiconductor device, the details of where the molecules were moving around right at the moment when that silicon piece of silicon crystal was formed can matter. And so you can end up with these things where you have a physically unique object. And sometimes you can even end up with a pair of physically unique objects. Like for example, our fingerprints. They kind of agree on the two sides because they were formed together, so to speak. And so you can have similar kinds of things where you form something together. I mean this was a thing used. Gosh, it's not new technology. This was used in, back in certainly the middle ages and possibly before people would take a thing, a stick or something else, they'd break it and one person would get one half, one person would get the other half. Do they match? You know, that was kind of. That's kind of like this physical unique functions idea. And that is a thing that seems to have been slowly gaining in popularity. And I think there may be some sort of significant pieces of technology that you've built that way. It's kind of a physical source of burnt in randomness. Now the question of Brownian motion, which is continually happening. I mean the one thing when something gets a solid thing gets formed that has a particular shape or particular electrical properties. It's another thing when one's actually using the continual randomness of things like bounding motion, molecular motion, and so on. Yes, you can use that to generate. In fact, it's a routine thing. In most microprocessors, there is an instruction that just says, pick up some noise from the environment and make a bit based on that, pick up some thermal noise from a transistor junction type thing and use that to generate a bit. So in a sense, as you try and generate those bits very quickly, you. You have to have sort of. The heat has to diffuse into the place where you're generating those bits for them to be sort of independent. But yes, the idea that you can kind of have a randomness antenna picking up heat from the environment to get your local randomness, that's a thing. And that's even a thing in existing technology. I think as protocols get developed, there will need to be more, you know, whether it's physically unique functions or something else that will drive the need for other kinds of randomness generation and so on. Let's see, Wish is asking, why are the axioms of mathematics necessarily the ones that are effective at describing things we see? Well, because we made mathematics as a way of understanding for ourselves things we are seeing, we didn't make mathematics for the things that are computationally possible. For example, I've studied quite a bit this question of what's computationally possible and what we can then say about that, but that's what's computationally possible is much broader than what we have identified and thought worth capturing in our mathematics. Let's see, Raoul is asking, what do things like dreams and higher states of consciousness talked about in Eastern philosophy tell us about ourselves as observers? Probably quite a bit. I mean, I think it's pretty interesting and people have been writing to me quite a bit about this, that kind of some of the things that we really are finding from a science point of view in the ruliad and observer theory and so on really are converging in many ways with things that people have talked about in Eastern philosophy and other kinds of areas like that. You know, I think it's a general principle that sometimes a little bit hard for people to swallow that by the time something has been talked about enough and it's gotten enough traction in the world, there's probably something to it. It may not be. It may be a thing that is, you know, the something is not what you thought it was. And when you interpret it in your way, the Thing that's important to you is different from the thing that was important before. But I think by the time there's as much talk as there is about kind of Eastern philosophy, you know, spiritual type type things and so on, there's something to it. It's not just, oh, it's all nonsense. It just got crushed by science. Ignore it. It's like, you know, and gradually we're starting to see things which people have said in, in ways that sounded very unscientific from a Western science point of view. We're finally being able to make this bridge from Western science to many of these kinds of discussions. And I think some of them, somebody like me, for example, is somewhat challenged in the fact that I don't have good sort of intuition about, let's say, Eastern philosophy and traditions around that. Or, you know, I'm not the sort of alternate states of consciousness. You know, I'm a person who, to give a too much personal information, doesn't remember dreams very much. So I don't, you know, I don't have a big intuition base from those kinds of things. So it's, you know, most of my, I have intuition based from things like studying the computational universe, which sadly not as many people as should have done have done. That's given me all kinds of intuition that's led me into all kinds of interesting things, including these areas that sort of connect with kind of Eastern philosophy and so on. But it's challenging to, you know, there's a big corpus of knowledge, but it's not kind of traditional western scientific style knowledge, a different kind of knowledge. And how one sort of merges that with the scientific is, you know, it's a challenge. But I think, I think we've reached a point in the science where that linkage is possible. Although part of that linkage has to come from people who have sort of an intrinsic intuition for those kinds of things. And you know, I know whenever I enter some new area, it's kind of always challenging to get the intrinsic intuition of that area. What's plausible, what's not. I mean, a great example of this is if you are used to physics and then you think about biomedicine, in physics we get nailed down answers. It's just like this is how it works, it's this theory. There are these rules, there's this equation, we can solve it, boom, it works. In biomedicine, things are always complicated. It's always, oh, there's always another effect. There's always a little thing, oh, in 0.1% of the population. This will happen because this particular genomic thing is different or whatever else. There's this infinite tale of sort of complexity of how things work, which is not the way it is in physics, for example. So those intuitions are different and if you try and apply one to the other, you end up getting confused. And so that's sort of a challenging thing to get to the point where you have that kind of intuition. Foreign. Let's see. Stefan asks, would it be easy to have, I think AI remaster old movies, real ones and cartoons so we can watch all the old gems and high end graphics? You know, people are already doing that. I haven't seen. This will be a fun thing. I haven't seen cartoon to live action. I'm sure that will come. I mean, that seems quite doable. You know, there's always the question of when you do kind of the AI rendering at the first look, it's like, wow, that's amazing. When you look in more detail, it's often a little bit flat, a little bit banal, a little bit kind of reversion to the mean kind of thing. I don't know if that will continue or whether it's, whether it's a question of sort of being able to make it more ergonomic for humans to pull it out in a slightly more creative direction. You know, does it, you know, the, the, if you're kind of taking, I don't know, pick your cartoon series, I'm going to show my age by picking, I don't know, Tom and Jerry. It was a cat and a mouse. You know, is there a live action version of that? Well, there might be some, you know, if you have a little creative flair, the mouse might have a tuft of some different colored hair or something that wasn't. That would be part of the sort of creative flair, but wouldn't be even visible in the original kind of cartoon version. But yeah, I think that's a thing we can fully expect. Let's see, Cosmic is commenting. Interesting is defined by coolness threshold. Yeah. Okay, well then we have the question of what's cool and how much is that societal and how much do you need to know how many social media likes you get to know whether it's cool and how much can you know that it's, you know, intrinsically know it's cool. My guess is it's, it's. You have to know the likes now. You know, a thing people have long been interested in is can you Predict, given this YouTube video, will it go viral? Or not. I've seen people try and use actually our technology, stack a bunch to make predictions along those lines. You know, should we invest in these YouTube videos because they're going to go viral? Should we? That's one version of it. Another version of it is what attributes, you know, if we weren't going to make a topic for a viral video or a meme or whatever else, you know, what, what attributes should that have? So it has the best chance of going viral. And yes, there's a certain amount of sort of analysis that AIs can give about those kinds of things. And you know, it's all complicated because as soon as things are sort of routinely AI assisted, it kind of raises the, the level for everything and means that for it to be interesting, you kind of like, oh yeah, I've seen lots of those before. If it's interesting, it's going to be the next level up, so to speak. Prutz comments. Since the scientific paradigm is a major cause for the Enlightenment, can we expect computational paradigm, multicomputational paradigm to kick off a socio philosophical paradigm of comparable importance? I think yes. I think that the kind of the idea that there was science, that there were intellectual things you could just figure out about the world, that was sort of the story of the Enlightenment and that it wasn't all just what was written in a book from a couple thousand years earlier. It was whether Aristotle or the Bible. It was kind of the concept that new knowledge was derivable, that science could say things. And then into the middle of that came things like Copernicanism, where the question of whether the Earth goes around the sun or the sun goes around the Earth, it's not an everyday issue. People don't care about that on an everyday basis. But what's important about it was the fact that that got established that the science gives you an answer to something which common sense does not give you. Common sense says the Earth is standing still. Common sense doesn't say the Earth is moving at a high speed around the sun. But yet the science said that and the science in some sense was right. And I think that was kind of a sort of an important kind of thing that cemented this idea that kind of you could invent new knowledge and the new knowledge would be meaningful. And so I think in as we see kind of computation, multicomputation coming in as a paradigm for thinking about things, yes, there's sort of a huge amount that could be different. I mean, for example, I think one of the things that's been a Post industrial revolution phenomenon is the Am I expected to understand this question? We expect to understand technology at some level at least. We expect, we're surprised, we're disappointed if we don't know why it works, how it works, because we built technology that mostly is understandable. Yet what we're learning from computational irreducibility, for example, is that if we want to take advantage of computation, that won't always be what's possible, that we will have to deal with technology that we fundamentally don't understand. We only know certain attributes that it has. But it's a very interesting question. What can be given the computational paradigm, what will be the big takeaways from it? I think one of them is there's stuff you can understand, there's stuff you can't understand, there's stuff you can regulate, there's stuff you can't regulate. There's the concept of the science based, controlled culture, society, et cetera, et cetera, et cetera. We kind of learn from the computational paradigm that's just not going to work. It's just not a thing that's possible. It's only possible if you constrain yourself in this very narrow way, which a lot of existing science and engineering have done and have been able to build a lot from that. But if we really want to take advantage of computation as a power, so to speak, we, we don't get to be able to always predict what will happen. You know, I think that that's so, you know, knowing what the consequences of computation, computational irreducibility and so on will be. I have to say, when I came up with kind of the idea of computational irreducibility in the early 1980s, the idea that that notion would turn into things like proof of work for blockchain had no idea of that, no idea that there would be this sort of thing that is basically computational irreducibility would be the thing where people are burning the electric power output of a small country, or at least have done in the past to mine Bitcoin by sort of grinding their way into computational irreducibility, that was not something from the scientific idea that was at all readily foreseeable. I think the very notion of kind of the sort of the modern conceit that we can figure out everything which was the, a couple hundred years later, a few hundred years later, consequence of kind of the success of the scientific paradigm that led to many changes in our way of thinking about life and everything. And I think the realization now that computational irreducibility means we can't know everything leads to kind of a little bit of a pivoting back to something where it's more like things just happen. Now with the whole sort of multi computation story and the way in which we kind of extract narrative theories like the fundamental theories of physics from this computationally irreducible underlying stuff, it's kind of knowing what's knowable, knowing realizing that there are things that are knowable, but there's a lot that isn't. And being able to identify what's knowable and what's not is probably an important direction for the future. I mean, you know, it could be the case that in biomedicine there are things that we want to know and they're computationally irreducible and we can't know them. It could be, but it could be that if we sort of architect certain kinds of things differently, like say in economics, it could be that there are things in economics where we actually want computational irreducibility. We want things to be unpredictable. If the stock market was completely predictable, many aspects of it wouldn't work. It would be kind of. It would be. There would be. If it was all predictable, everything would sort of move together in a very brittle way. The fact that there's this randomness in the market is what leads to kind of a robustness of that is that lets it be in some sense stable. But what is it there that can be identified as a form of regularity? For example, the idea that there's a notion of price as opposed to just every transaction has to be a barter transaction that is a consequence of some kind of higher level description that you can give. So I think that's a few thoughts there. But it's an interesting question and I actually probably deserves more thinking on my part. Although if you're interested in sort of anticipating the future, this is a particularly far reach in terms of going from kind of a conceptual idea like computation, multi computation and so on, to what its consequences will be. I don't think Alan Turing, 1936 inventing Turing machines. The idea that word processors and so on will be a popular use of computers was far, far away from what could have been imagined. And part of the reason there was that in Alan Turing's day, even in the days of early computers, it's like, computers are expensive and what can we do with them? Well, we have to have something that's really worth doing, worth using a computer on. As I think I've commented several times before, one of the things that, that I think will be true in the future is that sort of everything will be made of computers, that every piece of material will be in some sense programmable, just like many materials today are synthetic in the sense that you didn't mine them out of the ground, they were made somehow and in chemical reactions that we controlled or whatever else. And similarly, I think things will be sort of readily everything will be programmable, whether it's the textiles that got mentioned earlier or. Or other things. And I think when computation becomes that cheap, there'll be a whole range of other types of uses for it that we haven't anticipated yet. M Comments if someone had invented calculus in the Stone Age, it probably would not have been used for anything. Do you think there are ideas that may be rediscovered because they have a use in the future? Absolutely. I think there are things where, you know, there are ideas that get. Sometimes the ideas are not really even recognizable. You know, you could try to argue, I don't know whether this argument will really work, that the invention of the idea of an immortal soul from however long ago in kind of theology and other areas, that that idea is kind of an idea that's like formal mathematics, formal computation and so on, that the concept that there could be a thing that is sort of thinking like that is purely abstract, you could say that was that idea. Now, there wasn't a direct link between that idea and the actual emergence of sort of the modern notion of computation. And really you have to go back from the modern notion of computation to read backwards into that kind of connection. So unfortunately, you know, ideas before their time and without context are sometimes very, very hard to. They don't, you know, it ends up that at best you go back and you look and say, oh, it's cute, that was invented there. And we see that in a bunch of science fiction movies of like, oh, gosh, that's cute, you know. Now this, you know, video conferencing works like it did in 2001. Actually, it doesn't quite. But, you know, we can kind of read back to these things in terms of ideas that. I mean, there have been sort of these persistent ideas about, you know, artificial life or about being able to do, you know, read human thoughts from a distance, these kinds of things. There have been persistent ideas like that, but they don't really actualize. And perhaps eventually they will be possible. But I don't know how much the connection between the original pure idea and the actualization will really be. I mean, it's a question of whether I think there are other things in mathematics There are plenty of areas where people studied them and figured out results and nobody cared at the time. That happened in lots of combinatorics, discrete mathematics, number theory in the 19th century, early part of the 20th century. And for example, in number theory, there were things that sort of emerged as important in cryptography, things that were emerged as connected to kind of analytical calculations, algebra, geometry and so on, which only got context later, even though they were discovered as theorems earlier. So, yeah, it certainly has happened in mathematics, where mathematics has this nice feature that once you have the theorem, it's there and it can be stated sort of without context, so to speak. Although I have to say, when I say that's there, if you change the axioms out from underneath, if you decide, well, actually, Euclid's fifth postulate, the parallel lines never meet. You say, we're not going to do geometry that way, then a lot of the things that might have gotten proved in that tower of geometry was okay in that case, but it might not be in other cases, but all kind of go by the wayside and you have to kind of introduce something new and different. Let's see. Well, I should probably. I see other questions here, but I also see a reminder that I got to go back to some day job activities here. Well, thanks for lots of really interesting questions and comments and they've gotten me to think about some things. I hope this has been interesting to you folk and gets you maybe to think about some things a little bit differently and I should sign off and go to my other things. So thanks for joining me and bye for now.