2026-10-12 Schweizer Monat The Only Thing That Has Ever Improved Anything
YouTube Schweizer Monat EN’s channel
Duration: 01:21:08
Edited for readability. Verbal fillers, repetitions, and abandoned starts have been removed without changing the substance of the conversation.
The opening preview, a clip from the section at 33:02, has been removed.
Topic guide
Timestamps link to the corresponding point in the recording.
- 01:06 - Introduction
- 01:46 - What makes a good explanation
- 05:52 - A theory’s content is what it forbids
- 06:34 - Anything physics allows is achievable
- 15:00 - Why confidence and belief are the wrong goals
- 28:35 - Will we ever reach the final truth?
- 33:02 - Death is a failure of engineering
- 35:46 - Is “the beginning of infinity” itself provisional?
- 38:20 - Quantum gravity needs a new concept, not new math
- 42:28 - Is there only one game in town?
- 49:02 - Why knowledge didn’t explode sooner
- 54:13 - What makes someone a person
- 59:29 - Is creativity just randomness at the edge of knowledge?
- 73:04 - Is the Enlightenment fit?
- 76:39 - Demography: North vs South Korea
Transcript
Alex Buxeda
David, it’s a pleasure having you with me today in this conversation. I’ve been a fan of your work for many years now, and the Enlightenment is, I think, the most important project that has ever happened. It’s the one that reaches the farthest. It creates the best explanations. So if anyone hasn’t read The Beginning of Infinity, they are missing out on something very serious, and they really should. I’m going to attempt to make this conversation an appetizer for what The Beginning of Infinity really is, because that book is serious stuff, and I would recommend it to people watching.
David Deutsch
And I concur.
Alex Buxeda
So, David, what is the difference between a good and a bad explanation?
David Deutsch
Well, a bad explanation has elements in it, or is entirely made of elements, which either do not contribute to the account of the world that the explanation is giving, or which could easily be replaced if countervailing evidence were produced. Whereas a good explanation is hard to vary in that sense. As Wheeler once said, although it’s hard to arrive at, once we know it we say to ourselves, how could it have been otherwise? Or, how could it be otherwise?
Alex Buxeda
Yes. As you take a step towards a lower-entropy state, the path backwards looks obvious. It couldn’t have been otherwise. But it’s only in retrospect that it looks that way.
David Deutsch
Yes. By the way, I don’t agree with your association of entropy with the growth of knowledge. Entropy is a physics concept that doesn’t map well onto the everyday concept of disorder. Disorder, chaos, anarchy: these human concepts are only vaguely related to physical entropy.
In fact, it would be more accurate to say that entropy is our friend, because the production of entropy is associated with the production of knowledge. The production of knowledge is error correction, and error correction is a process of reducing many possible errors to one true state. A many-to-one mapping is an irreversible logical process, and therefore, physically, it is an entropy-producing process. So that is a good thing. But it’s not true, therefore, that a rise in entropy is necessarily bad. If we were to freeze ourselves to absolute zero, we wouldn’t create any knowledge, yet our entropy would cease to increase. At the beginning of this universe there was the lowest entropy ever, and it’s only in the transition from low entropy to high entropy that the beautiful patterns of life can emerge. So it’s increasing entropy that allows for all of this.
Alex Buxeda
Yes. The reason I mentioned entropy is that if you have the right, good explanation, exactly as we were talking about before, you can tap into higher-density energy. You can remove things that were inefficiencies. You stop creating heat, and you put all your energy into work. That’s the sense in which the increase in energy density decreases entropy locally, not globally. That’s an important point; the second law of thermodynamics is not going to be disproven in this conversation. I haven’t yet been able to merge these two sides consiliently. On one side, better explanations lead to a reduction in entropy locally. Is that something you would buy?
David Deutsch
Yes, although you’re mixing the absolute value of the entropy with the rate of change of entropy. We would like to create entropy as fast as possible. We reduce our own entropy, but not as fast as possible or as much as possible; as I said, that would involve freezing ourselves. So the connection between entropy and bad stuff, or between a lack of entropy and good stuff, is only approximate, and it’s not a good description overall of what is happening when we increase knowledge.
Alex Buxeda
Okay, that seriously updated my priors on my usage of the word, and I will take that into account.
Do you think that for a statement to have any meaning, it has to say what it is not? So for me to say “gravity goes down”, I have to say what is not down.
David Deutsch
Yes, if I understand you correctly. This is Popper’s maxim that the content of a theory is in what it forbids. So you have to specify what it forbids. They don’t necessarily have to be theories of physics. They can be theories of mathematics, or morality, and so on. But in every case they have to be specific. They have to specifically forbid something before they can be, well, I’m not sure about meaningful, but before they can be rational, they have to forbid something.
Alex Buxeda
I would love to hear your thoughts on parochialism, and on whether something being a resource or not is a parochial view of it. From your work: as long as it doesn’t violate the laws of physics, everything can be converted into any other thing. And if it’s preferable for us to have a flourishing intergalactic civilization, we could create that out of almost empty space, a cube the size of our solar system, from which we could transmute the hydrogen ions if we had the right knowledge. So is something not being a resource a parochial notion?
David Deutsch
Yes, it is. I have to slightly differ from what you said. We can do anything that’s permitted by the laws of physics. So creating a galactic civilization, for example, or an intergalactic civilization, would require solving the problems of coordinating over those massive distances, which may or may not be possible according to the laws of physics. But if it is possible according to the laws of physics, then doing it is only a matter of knowledge.
Alex Buxeda
Yes. It depends on your definition of coordination, and whether it would need something faster than the speed of light for the communication to happen.
David Deutsch
Yeah. Yeah. I’m assuming that there is nothing faster than the speed of light. Again, if we find that there is, then our conception of what the laws of physics are would have to change. But the fact that we can do anything that is allowed by the actual laws of physics doesn’t change.
Alex Buxeda
I tend to see everything as a feedback loop, and I think science is the same thing. It’s a prediction made that matches or doesn’t match the environment. Then you get that feedback, you update the priors, and based on that you make the next prediction, if the prediction matches the environment. And the substrate is itself another prediction, in evolutionary terms. My skin has its color because it’s a bet on the radiation it will get from the sun at this specific latitude. Sorry for the rambling here, but I don’t have a more eloquent way to explain this. These are embeddings built up over 200,000 years, over a billion years since life started, and the priors I have now, in a conversation with you, are all embedded priors that can be updated through a feedback loop. Do you think a feedback loop maps properly onto science?
David Deutsch
Yes. Science can be considered a feedback loop, and the growth of knowledge in general can be considered a feedback loop. But it’s not necessarily a feedback loop from the physical environment. It can be, and with biological evolution and so on it was. But now we can investigate things that aren’t only the physical world. We can investigate pure mathematics, and morality especially, and so on, which use the same epistemological logic but don’t necessarily involve the physical world. Again, they can, but they don’t necessarily. So it’s feedback from criticism: from whether an idea meets the problem for which it was proposed. And that can be a much more general thing than whether it predicts correctly.
Alex Buxeda
Yes. A couple of months ago I wrote a review of Steven Pinker’s book When Everyone Knows That Everyone Knows. In it he talks about Hans Christian Andersen’s famous story about the kid shouting that the emperor is naked. I wanted to tie this to what you said, to collapse the two things, you and the environment, because I think there’s only one there. We are part of the environment. The reason we can conjecture about the laws of nature, about mathematics, about pure mathematics that has no direct application to anything physical around you, is that you’re already embedding part of the environment in yourself.
The reason I say this is that in the review I argued that the kid shouted that the emperor was naked, and created the common knowledge that other people also knew that they knew, because people knew what the shape of the distribution looked like. They didn’t know who specifically was thinking it, or who specifically was going to say it. It was the fool, the most foolish person, who was going to think he would bring new information to the group by shouting that the emperor is naked. So maybe you don’t know who thinks what, but you know that the distribution has a shape, because it almost always has that shape. Even if you haven’t communicated with people, you have some embedded knowledge in yourself of what populations look like.
I think something analogous has to be happening when you’re able to think about theoretical physics, about pure mathematics, because you have a common ancestor with your environment. The laws of mathematics operating in the logic in your brain are the same as outside. The same dynamics that would govern an object colliding with another one are the ones you can simulate in your brain. I think that’s the reason why you can think about it. You and the environment are really one.
David Deutsch
I agree that we and the environment are one, and that the reason we can make progress in creating knowledge is that we have a shared problem situation that we inherit, and because we have a shared universal mechanism, including being a universal computer in our brain. We share all that, and that is what enables us to create knowledge.
When we think about pure mathematics, though, or any of these more abstract things, the shared knowledge we have is very specialized. If a pure mathematician is thinking about something, there may be only another 20 pure mathematicians who share that knowledge. And making progress may also consist of contradicting that knowledge. When somebody creates something new, it may be (rarely, but it can certainly be) that that person contradicts everything in the particular domain. He is then the only person who knows this knowledge. And he may have been the only person who, under the circumstances, even could have created it, because his personal variant of the shared problem situation differs by enough to make it possible only for him and not for anyone else.
Alex Buxeda
Yes, exactly. As I like to say, there are only apes and their credences in this universe. Well, an ape is too parochial a notion; let’s say the universal computing machine. So there are only universal computing machines and their credences. There’s nothing else. And we never really prove something right. We fail to disprove it. It’s through those independent failures to disprove something that we increase our confidence in it.
David Deutsch
Yes. Yes. Confidence isn’t really a thing to pursue. It’s error correction that is the thing to pursue: what’s left after you’ve corrected the errors. You may be confident in it in the sense that you would risk your life boarding an airplane because you have refuted all the ideas that maybe it hasn’t been maintained properly. But not because you have any absolute theory that the plane isn’t going to crash. We may at any moment be wiped out by a black hole moving at nearly the speed of light that happens to pass through the solar system. We can’t prove that won’t happen. We don’t have a sophisticated theory that says it can’t happen. On the contrary, it could happen. But that’s not a way to approach the truth, because I could have mentioned any number of other things which have the same kind of story behind them.
So confidence is the wrong word. Belief, too, is a concept that I don’t think has a place in rational thought. Popper says a scientist should not believe his theories, and I think that’s true not only of science but of mathematics and everyday life and everything. Belief is a concept that was introduced into epistemology maybe by the ancient Greeks and so on. We think that saying “I believe so-and-so” is the same thing as saying “so-and-so”. I believe that the Earth goes around the Sun; we speak as though that means the same thing as “the Earth goes around the Sun”. But in fact, “the Earth goes around the Sun” is our theory. Our theory of ourselves believing it is a theory about the human mind, and a very rickety one.
For example, the human mind must be full of contradictions, otherwise we wouldn’t have problems, and problems are the raw material of all kinds of progress. So the picture that the mind is, or ideally should be, a consistent or nearly consistent set of beliefs, and that when it isn’t consistent that’s the kind of emergency where we immediately have to update our prior or whatever: the concept of belief, I think, is a misleading one. And so, therefore, are the Bayesian concepts like prior and updating. I don’t think it works like that. Conjecture and criticism are the only two processes in an ideally rational mind. Everything else is not rational.
Alex Buxeda
Variation and selection.
David Deutsch
Yes. That’s a special case.
Alex Buxeda
I think the reason why all these words that aren’t very useful in your view emerged is that we needed a spectrum of confidence. We have resources; economics exists because we have a limited amount of resources and have to allocate them wherever the marginal return exists. As I said before, every feature of our biology is an implicit hypothesis: that light will come from the front, sound from different directions, that the radiation from the sun will be one level and not another. All those bets are made with a certain level of confidence, based on how much buffer, how much redundancy, is built into the system. How thick is my skin? Much thicker than would actually be needed to survive a normal day, because there’s a spectrum of days my ancestors evolved in, and on some of them they got very deep injuries. To survive those edge cases, a buffer was needed.
I think the same dynamics of mapping onto a distribution had to emerge in our language. When you say “I believe this”, what you’re actually saying is that you have less confidence than if you said “this is like this”. But you’re still talking about your confidence. So you’re adding information, in the sense of “I have failed to disprove this less”. But no one talks like that. They say “I believe this”, meaning “I don’t have a lot of confidence in it”, and if they had more confidence they would say “this is like this”. But the difference between the objective and the subjective there is a matter of how many resources we would bet on that specific conjecture.
David Deutsch
I think betting is only occasionally a useful metaphor, or a useful actual process. Again, if I say I’m going to fly on this airplane, and someone says, “Well, what do you think is the probability? What’s your credence for it not crashing?” I don’t know that number. Everyone who says they know that number is just making it up, whether it’s one in a million or one in a billion. You can look up the statistics for such airplanes crashing. But if my airplane crashes, it may not be for the same reason as all the previous ones did. It may be a new reason arising today, or arising only for that plane.
What I want to say (and I agree there isn’t really a natural way of saying it in prevailing language) is: I don’t know of an argument why this plane should be unsafe. If you have an argument, put it forward. But you have to take your place with the million other people who want to tell me that I shouldn’t go for a whole load of other reasons, like that the destination city has turned into a volcano. Which could happen. But I can’t make a decision based on bad explanations. I need a good explanation, one which can’t easily be varied or changed into some other explanation, for me to act, either by boarding the plane or by not boarding it. Either way, I need a good explanation. I never have a perfect explanation, but I have an explanation with no viable rivals. And once it has no viable rivals, I’d rather think about something else.
Alex Buxeda
This is basically Gödel’s incompleteness theorem. Because we are a subsection of the universe, we will never encompass the whole universe in ourselves. So whatever we use to build a model of how the universe works will by definition be incomplete, and whatever you’ve sampled, there will be things outside of it: a black hole crossing the solar system that would absorb everything. That’s something we don’t have the tools to test. And if we did have the tools to test it, there would still be something outside the sample we’ve analyzed so far that could surprise us. A black swan.
David Deutsch
Yeah, I think that’s quite a good analogy, although I have to say it’s not exact. Sorry to be so finicky here, but Gödel’s theorem in particular is about infinite sets. It doesn’t literally apply to finite things like experiments and the universe around us, which is finite. So the reason why our being part of the universe is relevant to epistemology is analogous, kind of similar, to the reason why a system can’t be proved true within itself. But they’re not the same, because reasoning about the physical world, for example, is always reasoning about finite things, and typically in mathematics we’re reasoning about infinite things which cannot be physically instantiated. As we were saying earlier, it’s important that we can reason about those just as well, because the same mechanism of conjecture and criticism can be applied to things that are purely abstract, things we can’t touch, and so on. Sorry to dot the i’s and cross the t’s.
Alex Buxeda
I love this. This is exactly why I wanted to have a chat with you: to finally put some rigor into the thinking, and extract the misconceptions I’ve had, which lazy thinking leads you to believe. This is exactly the kind of thing I really like.
So you said that the physical experiments we do are local, and mathematical thoughts are general, universal. But the physical experiments we do here are like a fractal. Well, not a fractal in scale; at a bigger scale the dynamics might change. But I mean the same dynamics would happen here and a light year in that direction. So it is local in your perception, because you are local, but it’s general in that the same experiment could happen there, and there, and there. So both are infinite.
David Deutsch
Well, those things are, assuming that the universe is spatially infinite, which it may or may not be, unbounded. We can’t actually do an experiment on infinity, whereas we can write a theorem about infinity. Unbounded is different from infinite.
By the way, speaking about parochial: I always say that problems, that is, the things we think about, are always parochial. Sometimes the solutions can be universal. They can be infinite in scope. We can suppose the universe is spatially infinite, and we can still have theories about the infinite space in the universe, and there are theories where time is also infinite. There’s nothing preventing us from creating and improving knowledge about those things, even though they are infinite, or even if they are. But the problems we address in order to create that knowledge are always finite. So they’re always parochial.
Alex Buxeda
That reminds me of how, if you throw a rock at a certain speed, it stays on Earth and all the energy gets dissipated into heat. But if you throw it with enough energy, it leaves the Earth, out of the gravitational well we’re in, and the range you get increases exponentially as you approach escape velocity, and after a certain point you touch truth. I’m trying to make an analogy with what you said: if your theory is true enough, you get an infinite amount of range. Well, not true enough, sorry: as you approach the truth, the range of the explanation increases exponentially.
David Deutsch
Yeah. I thought you were going to say, and you would have been right, that there comes a moment when you’re talking about an infinite range. So yes, it’s true that if you throw something fast enough, it may never stop, and may continue moving forever. And the theory of that is made by improving theories about what it will do when you throw it a finite distance.
Alex Buxeda
Exactly. And I’m wondering: if we touched infinity, what would that look like? Let me tell you my hypothesis about what it would look like, if we did touch infinity, in the sense of having found the truth, whatever it is. As we become more transcendent, as we become less parochial (and that’s the definition you use of what a good explanation is: if something works more, and if it really works, it works everywhere), if you figure out a way to solve all the problems and you go to the end, I think you would merge with the laws of nature. I’m not particularly religious, and I’m not trying to introduce any woo-woo here; maybe I should phrase this properly. But as you approach that infinity, if you do touch it, you would just merge with the laws of nature. It’s like a telescope: if it’s not properly calibrated, you don’t see the stars, you see the telescope. If you become aligned enough with the laws of nature, you wouldn’t see me, you would see the laws of nature.
David Deutsch
I don’t think we can ever actually see or touch the laws of nature. Understanding infinity is an everyday thing, really. It’s not as mystical as people sometimes make out. Mathematicians can reason about infinite sets and infinite things, but that doesn’t mean they’ve reached the final answer about infinity. Mathematicians still argue about whether the axiom of choice is true for infinite sets, or the axiom of infinity. They have controversies, and they invent new theorems and new mathematical objects which are infinite, long after they have understood infinity itself. That’s just where they’re taking off.
Similarly with the laws of physics: we get a better and better understanding of them, but I don’t see that ever stopping. If it stopped, that would mean there were no more problems, and for there to be no more problems means the growth of knowledge would stop, and that’s not a thing to aspire to. For various reasons, I think it can’t stop. In fact, the scope for further improvement is itself increasing, not decreasing.
Alex Buxeda
Okay. I think I just saw an analogy with biology. Do you know the black hole fitness hypothesis: that our universe is finely tailored to maximize the number of black holes? If you look at the cosmological constant, or all the other variables, they’re very well aligned with what would maximize the number of black holes. And this is hypothesized to be the same dynamic as Darwinian evolution, variation and selection, conjecture and refutation, applied at a universal scale.
So what I was saying about merging with the laws of nature would be this. If I nail something, if I find the truth about a local problem I have (I’m a biological being, and if I properly predict the amount of radiation from the sun, I will survive and maybe reproduce, find a mate, and have descendants), but if I really nail it, I go one step up in the hierarchy. I stop creating another biological being, and I start creating another universe. My life presupposes things: having been born, parents, a whole lineage of humans, then other animals before me, the beginning of life, and the emergence of the universe. If I do something very right, I create another human, and it’s a functional life, and they can keep doing the human thing afterwards. But if I got it right infinitely, I would create not only what preceded me by one step, but all the way back: I would create a new universe.
David Deutsch
I’m not sure where you get to the word “universe” here. You would create something infinite, namely your descendants. If you create AGI, you will create an entity that is like you in many ways, but immortal, or at least potentially immortal. Again, this is not such a big deal as people often make out. That we at the moment can’t prevent ourselves dying is just a failure of engineering. It’s nothing deep and God-given or anything like that. And similarly, when we solve it, that also won’t be something deep. It will soon recede into the past as something that, with the benefit of hindsight, looks natural: of course, what else could we have done?
In the distant future we might create something that is as complex as we think the universe is now. But that’s not the same thing as creating a new universe. The universe is everything, and that includes all the things we create. And we won’t ever arrive. That’s an important feature of how I see the world: we will always be at the beginning of infinity, as my book title says. Our knowledge might increase. It might increase exponentially. It might increase in ways we can’t even conceive of at the moment. But it will always be just scratching the surface of infinity. And we should be pleased about that, rather than aspire to having the whole thing. And then what would we do? Sit back and watch TV?
Alex Buxeda
So yes, I agree with you. But we are only dealing with conjectures, and the level of credence we give to different hypotheses. Said in a clearer way: all knowledge is provisional. And that piece of knowledge you shared, that knowledge will always keep going forward, that we are at the beginning of an infinity: is that itself provisional? Is it something that could be falsified in the future?
David Deutsch
It could be, and you should expect it to be. But you should expect it to be superseded not by its negation, which is highly problematic and which we have many reasons to reject, but by something we can’t currently imagine. We should expect Popper’s epistemology to be superseded, but it’s going to be superseded by a better epistemology, not by something like empiricism or religious belief or Wittgenstein, or anything that has already been rejected, but by something better. We expect it to be replaced by something better, and all our existing theories of physics as well. We expect, and hope, that they will be replaced by something better.
Alex Buxeda
This reminds me of many examples I think you gave in The Beginning of Infinity, of some problem we were thinking about in one way, and then someone came from a different axiom. We were thinking in two dimensions, and then someone came with a new theory that of course took the two dimensions into account but brought in a third. It’s as if I were walking ignorantly down a street and a bus came from the other side and ran me over, because there are dimensions we’re not considering, and those could surprise us. One example would be Newtonian mechanics. Newtonian mechanics is a good explanation of what happens here, in this range of speeds. But if you go too far in speed, relativity starts taking over, and you have time dilation and things start changing. And if you take things further, not in speed, sorry, because you can’t, but in scale, since we were talking about relatively small things, but not so small that quantum dynamics happens, then you have another dimension on which something can surprise you completely. So this is the sense in which you were saying it could be superseded in some other way.
David Deutsch
Yes, that’s a good example. Special relativity had to be superseded by general relativity, not by Newtonian mechanics somehow turning out to be true after all. And general relativity, similarly, we expect to be replaced by something not yet conceived.
By the way, I think the problem of quantum gravity, of unifying our two deepest theories of physics, is a conceptual problem. They will be unified by something that requires a concept, a way of thinking about the world that hasn’t been thought before. At the moment, theoretical physicists tend to think it’s a mathematical problem: that if we just apply the right mathematics, the right axioms, the right equations, we will solve it. But I think (and, you know, I could be wrong [laughter]) that even if we had the right equations, we wouldn’t know they were the answer unless we also had this new conceptual framework for understanding them. So I think the solution to quantum gravity is not going to be an equation. By the way, the solution in Newton’s theory, and special relativity, and general relativity, and quantum mechanics was never an equation. It was always a new physical concept, which was then expressed in the form of equations, which we then tried to understand further, which we then tried to supersede, and so forth.
Alex Buxeda
Could you try to define mathematics on one side and a conceptual framework on the other? Maybe give an analogy, to cement the meaning of those two things. You’re saying that with more mathematics we won’t figure out quantum gravity; we need the other thing. So what is mathematics, and what is a conceptual framework?
David Deutsch
Yes. Well, the example I give in The Beginning of Infinity is Zeno and his paradoxes of motion. To put it in modern terms, there are an infinite number of points between me and the other side of the room. So it follows that to get to the other side of the room, I must first get halfway, and to get halfway I must first get a quarter of the way, and so on. So in order to move at all, I have to do an infinite number of things, and therefore I can’t move.
This is a valid inference with the wrong conceptual framework. The mathematics may or may not be wrong. I don’t think Zeno knew about real numbers, but even if he had, the argument would be the same. The problem is that whether something involves an infinite number of steps or not, whether it involves something infinite or not, depends on the laws of physics. The laws of physics say that walking across the room is a finite operation, and that’s why we can do it. And they say that other things are infinite operations, and we can’t do them. But you can’t tell whether walking across the room is a finite or an infinite process just from the mathematics. You need the physics, and to have the theory of physics, you need the physical concepts that relate the mathematics to reality.
Alex Buxeda
Yes. I think there was a mathematical disproof of Zeno’s paradox: the rate at which you move through those segments increases exponentially, and the rate at which you would have to go through them also increases exponentially. So what you have to do and what you actually do both increase exponentially. The numerator and the denominator are infinite, but they cancel each other out, so you end up with the actual thing, which is the physics, and you can move.
David Deutsch
Yeah. But it’s physics that tells us which number to divide by which number. There’s an infinite number of ways of parameterizing this, and not all of them lead to the answer that it’s actually finite.
Alex Buxeda
Do you think there’s more than one game in town? Let me give you my understanding of it, and then you can tell me what you think. I think there’s only science. There’s only the pursuit of explanations. Of course, maybe it’s a bit biased that I’m telling this to you, but I think there’s really only the pursuit of explanations, and whatever is not furthering our creation of more explanations is not contributing to us making cosmological progress, defined in the broadest of senses. And I don’t know exactly what I’m saying, because what about art, poetry? Are they not contributing cosmologically?
I want to be consistent with myself, because before, we said that you have to say what something is not, and I’m saying there’s only one game in town, the pursuit of explanations. To say that, I have to say at the same time what is not the pursuit of explanations. Coincidentally, I have a conversation later with [my next guest], and I think it’s very related. What is not an explanation is like a blurry telescope where you’re seeing an afterimage. I understand why we like sugar: 10,000 years ago it was adaptive to eat sugary fruits. But now, in our environment, with many more calories and many more sugary things, it’s not adaptive anymore. So I can explain why you like the sugary thing while it’s negative for you. It’s a maladaptive taste. The bet you’re making with your taste, that this is good, is actually wrong. That’s one of the things I wanted to talk with her about. So: one, do you think there’s only one game in town? And two, what would be the other things that are not that game?
David Deutsch
Yes. You’ve said two different things and said they’re the same, but they’re not the same. The only game in town is not science. Science is a very narrow part of seeking good explanations. Seeking good explanations is the only game in town, because that’s the only thing that can ever improve things. You might call that reason. So not everything is science, but everything is reason, or should be. The scope of reason includes philosophy and mathematics and art, and whether sugar is pleasant.
By the way, I don’t think much of explaining human behavior in terms of our evolution, because we routinely do things that are biologically maladaptive, and it’s quite normal. Teenage girls get anorexia and die, for entirely parochial, social, idiosyncratic reasons of their own. And if we want to override our genes, we just do. That includes sugar as well. We can choose the aesthetics of gastronomy, and the aesthetics of gastronomy are not directed towards maximizing sugar. It’s not because we think it’s unhealthy. It’s because gastronomy is a field of art which points in certain sophisticated directions that have nothing to do with sugar.
Alex Buxeda
I think you nailed it when you said that if we want to override our genes, we can. And the reason we can do that is that we have a prefrontal cortex that is context-dependent. Memes, culture, emerged out of genes. Genes couldn’t adapt to the environment fast enough, and that’s why culture emerged out of them, to be able to engage with things faster. It’s like a hand that doesn’t move fast enough, so you need something with a finer granularity, and that’s culture, our memes that emerge out of it. So the whole reason we have a prefrontal cortex is that we needed context-dependent behavior in order to be adaptive. And if you’re not context-dependent, if you’re doing what your genes blindly liked 100,000 years ago, like the social cues of the beauty standards, that can lead to a problem, because you’re not being context-dependent enough. You’re trying to apply a solution from 200,000 years ago that’s not adaptive anymore. You have to override that with your memes and realize that you should be eating more; you should listen to your grandmother, who probably says you’re a bit too thin. All of those dynamics emerge out of exactly what you were saying: we have to override the genes.
David Deutsch
Yes. In this context, I think even more important is that during the ice ages, humans spread further and further into new habitats they hadn’t been in before. Almost every other species retreated. Some became adapted to the cold weather, but humans did not. Humans created things like fur coats and campfires. So when you say they overrode their genes, it sounds as though they just decided to power through the cold and walk into it anyway. But that’s not what they did. They solved the problem of living well in the cold, and presumably of living better in some sense than they had before. Maybe they were leaving behind more static cultures, and were able to attain more freedom with their fur coats and their better spears and their better campfires and so on.
The mystery is only why there was not explosive growth of knowledge at that time, or as soon as our species existed. Why was there not explosive growth of knowledge? I think it’s because creativity was not used for innovation as we think of it today. It was used for transferring memes. To transfer a complex meme, it’s not enough to copy the other person. You must form a theory of why the person is doing what they’re doing. For example, if they’re collecting dry wood for kindling, you can’t see that. Suppose this is before language; or even if there was language, you can’t express all the things in language. You can’t tell somebody how to ride a bicycle in words. So when somebody is collecting dry wood, you have to guess that it’s dryness they’re looking for. Only then can you copy the behavior of making a campfire as they do. If you don’t understand that the kindling has to be dry, your campfire won’t work.
So this is why we needed explanatory creativity. But it’s also usable for innovation. And I think that for most of human history it was not used for innovation, or hardly ever. They hardly ever produced an innovation. What they did was accumulate more and more cultural knowledge. Some of it was bad for them.
Alex Buxeda
Yes. Just like when throwing a rock, you have the precursor of what would give you infinite range. The precursor of cultural knowledge already had the base materials needed to touch infinity culturally and achieve the most flourishing civilization: from the moment life emerged, the moment variation and selection started, with protein folding, when the pattern that started replicating could be subject to natural selection. Well, I don’t know about this, because matter was itself the ancestor of this replicating piece of matter, so I think the ultimate ancestor is the universe itself. I don’t know if I’m sure about what I’m saying, but whatever those people had was already the precursor of what allows this. There’s no distinction in category. It’s all a matter of a gradient, right?
David Deutsch
I think there’s a qualitative difference. At some point our ancestors acquired the capacity for explanatory creativity. At that time they maybe didn’t have enough memory in their brains, or enough sophistication in their speech apparatus and so on, to be able to cooperate and do these things as well as we can. But they had the ability to do it, if only slowly. And if they had done it, even slowly, they would have had explosive growth of knowledge. But they didn’t.
A qualitative change happened with the introduction of explanatory creativity into the evolutionary scheme, and once it did, it changed everything. But if you had been there and looked at it, if you were an anthropologist, a time traveler, or somebody from another planet looking at this, you wouldn’t have been able to see anything at first, because apes have memes. Early humans had memes. You would have seen that the human memes were a bit more complex, but no big deal. You might have thought there was no big change you could see in the first people. I use the word “people” for entities that have explanatory creativity.
If you’d looked carefully enough, in detail, you would have seen it, because you would have seen that this person is not copying the behavior of the other person building a campfire. He’s copying the idea. And that idea is not being copied by any kind of ribosomes or any biological process. It’s being copied by explanatory creativity. He’s explaining what the person is doing. Sometimes he’s explaining it wrong, but he’s explaining it correctly millions of times more often than random. And so the memes of Stone Age culture were very much more complex than those of pre-people apes. By the way, I think some of the apes before humans were already people, but they never made it. They never had the takeoff that humans had.
Alex Buxeda
I’m having trouble understanding the distinction between people and not-people, because from what I know about ecology, there’s a gradient. From the genotype the phenotype emerges, and from the phenotype we can have an extended phenotype: the dam of a beaver, for example. That’s part of the genome, embedded in the genome: the ability to build a dam given the right context. The beaver plus the environment creates a dam; that’s the extended phenotype of a beaver. But then you can go one step above that and have adaptability to certain types of environment, like chimpanzees have. Chimpanzees have not only the extended phenotype, the things they construct, but also the ability to replicate whatever they’re into, to adapt to a certain ecology that you can arbitrarily mutate. You can change it, and they would find the tweaks that solve the problems locally.
David Deutsch
No, that’s not what they do. They always acquire such things by copying, or by randomly messing around. But randomly messing around is extremely slow. So the number of memes in a chimpanzee culture or a gorilla culture is tiny. They only know a handful of behaviors that they have learned from other gorillas. And very, very rarely does a gorilla invent a better variant of some meme, and they do that by sheer chance, not by explaining it.
On the other hand, the early precursors of humans, who did have explanatory creativity, hardly used it for that, as I have just explained. Just like the gorillas and the chimpanzees, they hardly ever did anything new. They usually only copied, even though they had the ability to do something new. Therefore, if such a person, a pre-human but post-gorilla ape, were somehow to exist today, there would be a way of allowing them to participate in our society and to be creative in our society. They might be mentally slow, and not able to participate in the most complex parts of culture. But they would be able to participate positively, to have ideas, and to improve their lives, without it being biologically entailed for them. In other words, they would be capable of fully human thinking, even though impaired.
Alex Buxeda
Oh, yes. The base material from which all this is built is memory and compute. If you have both, if you have a long enough piece of paper and the ability to do calculations with enough time, you would be able to do everything. But you said they would be able to contribute positively to society. “Positively” is an economic term. It relates to marginal benefit. If this person next to me is going to cause more trouble, cause a deficit economically or financially to the company, they’re not contributing positively.
David Deutsch
Well, yeah. I mean, if it were possible to bring a Homo erectus back to life by gene editing or whatever, I think it would be highly immoral to do, because that person would not be leading the kind of full life we know how to provide. But from a scientific point of view, it would be possible. And it’s not true that they wouldn’t be paying their way. They could earn a fortune by being studied by scientists. So there’s no problem with them contributing a lot to society. It’s the cost to them that I think would make this immoral, not the economic cost to the rest of society.
And it’s not true that you only need memory and processing power. You need the program. You need the program for explanatory creativity. That is in us, and it was in some of our ancestors, and it’s not in gorillas. That program can run, and it’s creative. That’s all it does: it creates things. But whether it actually does what it’s capable of doing depends on all sorts of things, including all the existing knowledge, the memes, the society, and the environment: whether they allow or facilitate somebody inventing a new meme. They might not, even though they can.
Alex Buxeda
I want to attempt now to give you a proof of the opposite, because I think the opposite is actually true. You said that chimpanzees only figure new things out randomly, one in a million times, because they don’t have anything from which to build towards the next conclusion. But I think that, by definition, this is always true at the edge of our knowledge. At the edge of knowledge, if I’ve exhausted all my knowledge and don’t have any more, the next step will always be random. So this is like a fractal; it happens with everything. It comes back to my point from before: everything is a spectrum, and there was no black-or-white distinction from the beginning of life. The universal precursor of universal explanations was already there, because everything is variation and selection.
David Deutsch
Okay. Again, there are two different things there, one of which I agree with. It’s true that at root, all variation has to be random. I think in the brain there’s a hierarchy of criticism and conjecture. At the lowest level the conjecture is random, and then there’s criticism, which is automatic, unconscious criticism, and then, less random but still undirected, and so on. In that sense, the basis of human thought is randomness at root.
Similarly, imagine taking billions and billions of gorillas and arranging them in steps, so that they could only copy the level underneath them, and you gave all of them some stimulus. The lowest ones would do the gorilla things. The next level would do the gorilla things, but with a not completely random starting point, and the next level would have an even better starting point. Then you would have several of these levels cooperating to produce a better criticism than merely rejecting or accepting the theory, and so on. And once you got to a height of, I don’t know how much it is, but it can’t be that much, once you got to a height of a hundred, you might in effect have a human mind run by gorillas, very slowly. Rather like Searle’s Chinese room, if you know what that is: it’s a room that knows Chinese. Does it really know Chinese?
Alex Buxeda
Exactly.
David Deutsch
So yes. Yes, it does know Chinese. The system does know Chinese. And this whole thing would be a person, even though its constituents are not. Just as within one brain, neurons are not people, and clumps of neurons are not people, but at some point some level of neurons constitutes a person. And there’s a lot more going on in the brain apart from that, like memory and everything. So it’s only a small part of the brain that is a person, and it’s using the rest of the brain as a computer, a dumb ancilla.
Alex Buxeda
I think there’s only compute, and I’ll tell you why. I think your “at some point” is doing a tremendous amount of work there, and it’s easy to vary. At some point the gorillas would become a person, by aggregating all their randomness and all their directed or undirected figuring out of what is true. That “some point” is easy to vary, because you could say, oh, a bit later, a bit sooner. It’s like the definition of AGI: people thought AGI would be a program you could talk with at PhD level about many different topics. We achieved that, and it’s still not AGI. So I think that “some point” will keep moving the goalposts, because it’s not fundamental. I think you’re putting your finger on an atom that isn’t the atom of figuring things out.
Maybe I’m completely wrong, but let me tell you why I’m saying this, because I think I’m onto something. I think figuring things out, random variation and selection based on fitness, on what matches the environment, is a fractal. It happens at all scales. It happens at the smallest of scales, with matter, seeing which pattern starts replicating, the emergence of life. It happens at the level of the amoeba, the bacterium, the chimpanzee, the human, and the AGI we might create. So that is the atom: the atom is random variation and selection. And at the edge of whatever your knowledge is, you are by definition always throwing things around, not knowing what you’re doing, because you don’t have the knowledge to explain it. Gödel’s incompleteness theorem again: because you don’t have the knowledge to explain the next step, you take a random one. One works and the other doesn’t, and you select the one that works. It’s random from a sophisticated set.
David Deutsch
The difference between a chess grandmaster and a chess computer is that the chess computer mostly pursues all the lines it can, and to make it better you try to trim down the criteria it uses. Whereas a chess grandmaster looks at far fewer possibilities. So although at the root level both of them are random, or rather both are trying every possibility (in chess you don’t have to be random; you can try every possibility at some level), at a higher level the chess grandmaster is not trying every possibility, and is not trying them at random.
Alex Buxeda
Exactly.
David Deutsch
The computer is trying them at random, or trying every possibility out of some set. So they are achieving the same functionality in completely different ways. The human chess player is using explanatory creativity.
And I admit that I don’t know where the line is. I don’t know how to program an AGI, and therefore it’s possible, or likely, that I have misconceptions about what an AGI is. So you may be right that being an AGI is a matter of degree after all. But the argument I gave seems to me conclusive that it has to be all or nothing.
An analogy I could draw: we already know this in regard to universal computation, which is a different thing. It’s not to do with solving problems; it’s to do with calculating mathematical functions. People have been working on the smallest possible universal Turing machine, and they’ve got it down to 17 states, and 15 states, and seven states. I don’t know what the record is now, but the smallest universal Turing machine is a very small thing. However, to calculate any useful function, it would have to do a lot of simulation of bigger machines, and it would need a lot of memory. Asymptotically it would need the same amount of memory, but for slightly more complex things it might need a lot more. And of course it could always do with more speed. Therefore, being a universal Turing machine is an all-or-nothing property.
Alex Buxeda
Let me try to dovetail your example of the grandmaster who isn’t randomly processing options. Doing it non-randomly is knowledge, up to the point at which you have knowledge. Knowledge is like a tree, and you weed out all the options that aren’t good. Because the grandmaster knows a lot, he knows what’s not good, and goes straight to the point of his competence. The reason I said this is like a fractal is that wherever your edge of competence is, it will be random there. The grandmaster doesn’t do a million things that are bad, doesn’t need to spend his compute on that, but wherever his compute goes is at the level where his knowledge stops.
An analogy I could bring you is a compression algorithm. If you’re trying to compress something, you find the regularities, the things that aren’t information, and you weed them out. You focus on the signal, the bits of actual information. And the moment you’re forced to store information is when you can no longer find regularities. If there are no more regularities, because you don’t have the right knowledge, you just have to deal with the bits you’re left with. This is the sense in which I mean this is a universal feature that applies like a spectrum from zero to infinity, and the edge of your knowledge always has to be taken a step forward by randomness. I don’t see an alternative.
David Deutsch
It’s an interesting idea. I agree that at the edge of your knowledge you’re always faced with ignorance, absolute ignorance, because your knowledge has run out. But whether that is random is a separate thing, because random means, for example, equal probability. But what do you choose equal probability over? John Maynard Keynes had a very nice analogy. If you know that somebody is either British or French, but you don’t know which, then you might think he has a half probability of being British and a half of being French. But in those days Ireland was part of Britain. So if you don’t know whether he’s British, Irish, or French, then you would conclude that being French has a one-third probability. That shows you that even if you have zero knowledge of what he is, how you split up the conceptual space is still an additional thing you have to decide. Just saying it’s random doesn’t specify it. You could be random between Britain and France and then random between England and Ireland, or you could be random between Ireland, England, and France. I’m mischaracterizing the various countries now, and people are going to get angry with me, but mathematically you know what I mean.
Alex Buxeda
Yes. But I couldn’t have given you a better example, because randomness is the definition of lack of knowledge. Something is only random when you lack the knowledge to find the regularities.
David Deutsch
If you absolutely don’t know where somebody is from, you have zero knowledge, but that cannot be expressed as randomness.
Alex Buxeda
And that’s why it’s random.
David Deutsch
Random means that there’s a set which has certain probabilities. If you use the word “random” just to mean undetermined, then yes, it’s undetermined. But random is a structure. It has a structure, and different structures give different answers.
Alex Buxeda
Yes, precisely. Having zero knowledge does not mean random. Zero knowledge in the next step, because the structure is always there, because we exist. The fact that the universe exists is a given, from which we emerged as life, and from which we humans emerged out of that life. That’s the structure we have. We cannot conceive of all the options. We are forced by our knowledge, by our biology, into a set of options, and from that set of options we choose randomly. Our knowledge carries us to a filtered set.
David Deutsch
If you know the options, there’s a set of options. In other words, you know all the options. You still don’t know how to structure those options, what to give equal probabilities to. There’s a very large number of ways you can assign probabilities to a set whose probabilities you don’t know; I’ve given the example.
Alex Buxeda
I’m running out of time; in seven minutes [my next guest] is going to join this studio. So I wanted to ask you one last question. Maybe we could catch up on this specific point later, because I think there’s a lot we could discuss about this eternal finding of new hypotheses, variation and selection, which I think is a fractal, while you think there’s some exponential point where everything starts. But let’s leave that aside for a moment.
I have a hypothesis called the cockroach theory of fitness. The Enlightenment is beautiful. It’s what creates beautiful things, things worth fighting for, dying for, and putting your effort into. The civilization we’ve built, the solutions we’ve built, the accumulation of knowledge are as valuable to me as anything I can conceive of. I’m a huge advocate of the Enlightenment. I try to join you, Steven Pinker, Michael Shermer, and Matt Ridley in the fight for the values of the Enlightenment. But what if the Enlightenment is not fit? Not biologically, but in terms of fitness thought about existentially.
I could give you an example, North and South Korea, where this could be the case. Maybe it’s not general, just a specific example, which could be extrapolated if it were true. North Korea has a fertility rate of 1.8, or 1.7, children per woman. South Korea has a fertility rate of 0.7. If time goes on, one will collapse towards zero, and the other will take over the other country just with its population. One is more enlightened than the other, and the enlightenment happens to come together with demographic collapse. So if we run out of a substrate on which to run the software of the Enlightenment, there won’t be an Enlightenment anymore. Do you have any thoughts on how the Enlightenment can actually build the substrate?
David Deutsch
Some, yes. Again, I don’t think your example quite matches the general point you’re making about fitness. Yes, I agree that the Enlightenment may not be fit. We may be doomed. It may be that the Enlightenment just isn’t going to work, and that the human species won’t work. After all, if I’m right about previous species having been people as well, most species that have been people have not survived. They’ve gone extinct, despite having the ability to go to the moon and do all the things that we do, perhaps more slowly. But still, according to my view, they went extinct unnecessarily. They didn’t find the right social organization. They didn’t find the right aspirations, and so they went extinct.
We could go extinct. If our species doesn’t go extinct, our civilization could. We could have a dark age lasting 10,000 years, and only then resume; that sort of thing has also happened before. So there’s never a guarantee. If our civilization is destroyed, or put into recess for a time, it will be because we failed to solve the problems. Therefore what we have to do is not hope that we’re fit. If somebody is guessing that we lack fitness in a certain way, we must try to solve the problems of that: see whether that theory is right or not, and if it is right or partly right, solve the problem it has raised. That’s all we can do. We may fail, but all we can do to prevent failing is the usual thing, conjecture and criticism. That’s all we have.
As for North and South Korea, that’s an argument from demography. You’re young and I’m old, and I remember when all the arguments from demography went in the other direction. Demography is something that can change on a sixpence, as we used to say in England, or on a dime, as they say in America. It can turn round in less than a generation and start going in the opposite direction to the way it was going. The reasons why the birth rate is low are multiple and complex, and what we ought to be doing is trying to address those. For example, we ought to be making it easier to have children. We ought to be making it easier to assimilate immigrants.
Once North Korea has got 100 million people in it and South Korea only has 10,000, it’s not a foregone conclusion that North Korea will overtake South Korea. South Korea would still win the war you are imagining. So you’re saying, “Okay, well, then they just have to wait another couple of generations.” But no: by that time the South Koreans will have noticed that something is terribly wrong. Israel has a similar problem, and is one of the few Western countries with an increasing population. This is a matter of culture. It’s a matter of the problems people are solving in their own lives, which mesh with the problems of society as a whole: the general problem of how to make a viable culture and a viable civilization. We have to be doing that all the time. It’s not a special thing to be doing in regard to population. We need to be doing it on a wide scale, in all sorts of different ways.
Alex Buxeda
So it’s a matter of culture, I agree with you. But it’s also a matter of game theory. We have to coordinate to be able to cooperate, and what happens is that maybe there’s a market failure there. The Enlightenment-oriented society creates positive externalities for the retrograde, authoritarian command economy, which imposes from the top down the recipe the West developed. Let’s say it works. They become relatively rich, and because they are oppressive, they keep having more children than the enlightened West. So if the tragedy of the commons, the market failure, is big enough that it doesn’t allow people to coordinate on a good enough outcome long term, then North Korea could very well dominate long term, and South Korea might disappear.
David Deutsch
Just as we don’t have to obey our genes, we don’t have to obey market forces either. We can do what we like. We can see a problem and address it. If addressing the problem as best we can involves reducing the population, we can do that. If it involves increasing the population, we can do that. As I said, the outcome is not determined by population or by any brute parameter like that. It’s determined by the decisions people make, by the knowledge they create. So long before this overwhelming happens, the problem will be obvious and extreme, and people won’t go on having too few children to retain the existence of their culture until their population is only a thousand. They will see the problem long before that. In fact, they are seeing it already now. You and I are talking about it, because people have recognized this problem, and solutions are being proposed. We don’t want to go immediately for an authoritarian solution just because the authoritarians might be better than us at doing certain things. We have to find solutions that are compatible with our values. In fact, improvements on our values. That’s what we need to find.
Alex Buxeda
What a wonderful and optimistic note to finish on. David, it’s been an absolute pleasure talking with you, and I hope to catch up sometime on that specific disagreement we had.
David Deutsch
Fun chatting, and let’s do that.
Markdown