2025-09-11 Into the Impossible David Deutsch - We Will Build Humans Before AGI
Duration: 00:31:09
Transcript
Brian Keating
And I want to ask you, is it even possible in principle to conceive of a machine that A has a happy thought, what does that even mean? And B could visualize the visceral sensation of free fall in an elevator or going over a roller coaster hump. Is that not something unique to the human embodiment of natural intelligence, therefore making it impossible for artificial intelligence to construct interesting new theories of physics, for example, like the Einstein equivalence principle?
David Deutsch
We are such machines, so obviously it’s possible for them to exist.
Brian Keating
But in computing, in silicon, in qubit form, in any form you like, in artificial intelligence to replicate that.
David Deutsch
Well, so an intermediate step before considering AGI would be, is it possible to build a human being from scratch out of atoms? I think that that is clearly not forbidden by any fundamental theory that we have. Obviously we’re nowhere near being able to do it. And then what is really operating to make the subjective sensation of falling and that kind of thing is not different between neurons and whatever your toy brain is made out of. If you could make the toy brain out of materials that did something analogous to the information processing in brains, then it would say that it experienced that as well. But I think that it’s obvious that it would experience it because when we say that we do experience it, we’re actually consulting our memory of experiencing it. We never experience what’s actually happening in a particular instance. We only ever experience what has happened to us one fifth of a second or more ago and what’s more we are interpreting it. And all that is computations. And we know no machine can perform computations that are different from the ones that can be performed by a universal Turing machine or if you like, the universal quantum computer. So I very much doubt that quantum computations are necessary for human cognition.
Brian Keating
So the notion of embodiment, I talked to Noam Chomsky about this many years ago, but he seemed to think it was critical to have an embodied, you know, and again, that’s not impossible. So therefore it must at some point be explored and a proper question. But in the near term, do you think it’s more likely we could get a Turing test that’s based on a new law of physics, not some new discovery or even an old law of physics that’s finally understood and more predictive and more explanatory, but only on the basis that it’s embodied in some sense? In other words, what part of what you do as a theoretical physicist is, you know, only made possible and enabled by your physical embodiment?
David Deutsch
Well my physical embodiment in my brain, so the way I look at it is that I am a computer program. In other words, I’m an abstraction. I’m not the brain. The brain is just the hardware on which I’m running, but I am software. So I am therefore embodied in the brain. If I were embodied in something else with the same computing power, then I would be embodied in that. If it didn’t have the rest of the body, like the arms and legs and so on, then that would be equivalent to being in a sensory deprivation tank. But a person in a sensory deprivation tank is still exactly as much of a person as when they’re outside. And, you know, if you lose a limb, you don’t say, I’m less of a person now. Or I mean, you might say that, but you’d be saying that metaphorically. You’re not less entitled to the vote or less entitled to human rights if you lose an arm. I think the same is true of the brain. That is, what counts in the brain is the running program. That’s what is conscious. That’s what has feelings. And that’s what is embodied. When you talk about something being embodied, I am embodied in the brain, mostly. I mean, it’s also the rest of the body also plays some role with chemicals and so on. But that’s, again, all just information processing. And the brain is the most important because it’s the only one in which error correction occurs.
Brian Keating
I want to ask a hardware related question. You mentioned software, but I can’t resist as an experimentalist mentioning hardware. And that’s this phenomenon known as lock-in. One of my favorite examples is the Hubble Deep Field could have been a lot deeper. And we could have learned a lot of the things that took until the James Webb telescope at much greater distance from the Earth at the L2 Lagrange point, a million miles from Earth, as opposed to 630 miles above the Earth. And that was set by the width of a horse’s rear end. And I don’t know if you know this story, David. Have you ever heard this? OK. So this is at least partially true. So the space shuttle was launched using solid rocket boosters. Those solid rocket boosters were made in Utah in the western United States. And the space shuttle was launched from Cape Canaveral in Florida. So they had to traverse from Utah to Florida, which is 2,000 kilometers plus, something like that. And on their way, they had to go through at least seven different tunnels in trains, train tunnels. And those train tunnels are a certain width. They’re set by the width of the gauge of the railroad. And two of those tracks side by side is what determines the maximum width of something that can go through such a tunnel. Well, now we go back another step. And well, what sets the width of the railroad gauge? Well, that was set by the width of horses that used to pull a standard gauge chariot going back to the Roman Empire. So two horses were used to pull chariots. And that’s the actual width. It’s something like 2.3 meters or something like that. And that’s the width of a railroad track, which is now half the width of the tunnel that the booster had to go through. And the altitude the booster gets to is proportional to its area, specific impulse, its jerk, if you will, is determining what altitude it got to. And that caused us the Hubble Space Telescope launched by the space shuttle to get to an altitude of about 400 to 600 kilometers or something like that. And that meant it went through the atmosphere and experienced something called the South Atlantic Anomaly. And it also had day night cycles every 90 minutes, which is bad for the thermal regulation of the camera. At any rate, David, the point I’m getting to is that no one would have thought that the Hubble Deep Field image of a galaxy would not have been as good because of the width of a horse’s butt was too small. If it had only been bigger, we would have had a higher, you know. But I want to turn that type of thinking into a question about A.I. because we talk a lot about A.I. again, you predicted immensely presciently in this book from 14 years ago. Many of the things we’re just now seeing and anticipated some things we haven’t yet seen. But I’m wondering if we’ll ever see them because of lock in. In other words, the LLMs that you and I use every day are based on GPUs married to LLMs. And those GPUs were designed such that my kids could play Minecraft a little bit faster than their neighbor and kill him, you know, before he kills my kid in the game.
Brian Keating
They weren’t designed for this. They happen to be exceptionally well suited to matrix multiplication, inversion and other things. Right. So I’m wondering, did that lock us in to a finite ceiling on what computers can ultimately do because of the trillions of dollars that are going into GPU plus LLM and not into alternative modes of artificial intelligence that could be actually useful for physics and determining new laws of physics and so on. Hey, everybody. I’m usually the one that asks my guests to judge their books by their covers. But today I’m asking myself to judge my own book by its cover. My newest book, Focus Like a Nobel Prize Winner, is chock full of advice, life tips and focus and productivity tips from nine of the world’s greatest minds, Nobel laureates ranging from economics to peace to physics. Of course, it launches September 9th, which is also my birthday. I’ll go check it out. And my publisher’s got Amazon to run a special just for listeners of the Into the Impossible podcast. You can get the Kindle edition for only 99 cents. That’s less than a new pocket protector. So go to Amazon and get the Kindle copy today because this special only lasts for the first week after launch. What do you think about that?
David Deutsch
I think that sort of thing is bound to be ubiquitous and it’s bound to slow things down compared with not having it, not having that thing. But it cannot impose a bound. So it can make things happen later than they otherwise would. But it can’t stop things happening. For example, with your horse’s rocket story. Yeah, if that horse’s butt thing were really the impediment to space travel, then by now, you know, it would have it would have prevented the Hubble Space Telescope from being better. And you know, you could imagine a scenario where it prevented all sorts of other things. But by now, if there was a factory in one part of the US and you needed to get the rocket to the other part, then Elon Musk would make a reusable rocket into which you could put that rocket and would transfer it that way, would simply take off in one place and land in another. And then the problem would be solved, except probably it would be expensive. But then that would the problem of expenses also solved in the long run by creativity.
Brian Keating
One thing I’ve wanted to ask you for 12 years or so, what I call the unreasonable effectiveness of the square root in physics. And that’s if you look at the Poisson bracket in classical mechanics, you get a commutation relation that happens to vanish. But then if you add the square root of negative one, now all of a sudden it doesn’t vanish and you get this very interesting and rich behavior. Likewise, if you add the square root in a certain sense to a group called SU(3), you get rotations in SO(2). If you add quaternions, which are sort of the next level up in the Grassmann algebra for imaginary numbers above, you know, second and third, fourth dimension. And then if you do matrix multiplications in higher dimensional spaces, you encounter the LU decomposition and you get square roots as well. What do you make of the square root?
David Deutsch
Imagine that there are many universes and the reality consists of many universes in each of which physical quantities have different values like the position, you know, is one thing in one universe and it’s a different number in another universe. So you might think that the way you’d represent that mathematically is as an array or a vector where each element of the vector is the value of that variable in that universe and then the second is the value in the second universe. Turns out that doesn’t work because the real descriptors of the multiverse are algebraic. The real relationships are algebraic relationships between what are often called operators, linear operators, whatever. But the important thing is that they are an algebra and you can’t represent an algebra in a vector, but you can do the next best thing and represent them in a matrix. So instead of having an array of real numbers, you want to have a matrix which somehow represents a set of real numbers. Well, what can they be? What set of real numbers can be represented by a matrix or by an element of an algebra? Well, the answer is its eigenvalues. Those are the only set of real numbers that are invariant that you can construct out of an algebra. So you have an algebra and you want eigenvalues of things that describe reality to be real numbers. Therefore, they are Hermitian. Symmetric or general matrices won’t do. Only Hermitian matrices have real eigenvalues. So that’s my answer. That’s where complex numbers enter into physics.
Brian Keating
And then higher generalizations in terms of spinors and so forth are naturally now again, hearkening back to my first question about mathematical inductions, limitations and how minimal surfaces have singularities less than dimension seven. And so is that are we limited then? Or would you expect that every mathematical structure, whatever the next level up from a Grassmann algebra is from dimension three or four, rather complex dimension two. But is this a possibility? In other words, should we spend a lot of our time exploring all mathematical structures and how we could force them to instantiate themselves as physical observables?
David Deutsch
No, we should only explore the ones that we think might solve a problem in physics or mathematicians should explore only those ones that they think solve a problem in mathematics or interesting to mathematics. I think that’s the same thing as solving a problem in it. If there is some mathematical limitation, it’s because solving the problem in physics requires that limitation, like the finite speed of light. It’s necessary that the laws of physics impose that limitation because of the structure of space time, because time has to be different from space and that kind of thing.
Brian Keating
OK, OK. So you may not want to talk about this. This is going to go into Israel, Palestine. But your thought process and your frameworks for assessing memetics and memes, as I’ve had the honor of hosting Richard Dawkins in person here and on the West Coast. I want to ask you this question. Memes, as you frame them, are replicators of human epistemology, effectively, in some sense, that they shape the modality of human thought, the way genes and the extended phenotype, for example, can shape our culture, our biological development. And I started to think about the Israeli-Palestinian conflict through this lens only recently. So my thoughts are necessarily well, well-unformed, shall we say. But I often wondered, you know, how can this conflict persist, you know, for decades and decades? And I wonder if it can be explained through memetics in that these are enduring, that they are entrenched, that they are simple to explain. It’s easy to say free, free Palestine. These are very catchy things. And as you had in your country yesterday at Glastonbury, we’re speaking in late June, this horrific event where these antisemitic rappers were singing about the death and threatening death to the IDF. And, you know, basically Israel would be soon to follow. Is there a way to interpret the Israel, the endurance, the persistence of either side of the conflict? But obviously, I’m Jewish. I’m very Zionist. Is this possible to be understood through the lens of memetics as to its endurance long past the, you know, kind of date for which other cultural and even wars are being fought with much greater human genocide and so forth, if to the extent that Israel even is complicit in such a thing. But can you understand it? I’m sorry for the long-winded question, but do you interpret it in any way as a memetic battle as an addition to an actual?
David Deutsch
So the memes of what used to be called anti-Semitism, but I want to get away from that terminology. So I call it the pattern. It’s the pattern of legitimizing hurting Jews. It’s not hatred. It’s different from many other memes. Obviously, it is a meme because, you know, there were people who were enacting this behavior long before anyone alive today was alive. So it didn’t get from one generation to another by genes. It got there by people telling each other stories, telling each other ideas about how the world works. Why this? But it’s certainly not explained by that, because as you said, most memes, it is very rare for a meme to last 100 years, much rarer, a thousand years.
Brian Keating
Without evolution?
David Deutsch
Well, evolutionary things can last for a long time, but that’s because they have a different mechanism. Memes have to run the gauntlet of being represented in minds and then enacted in every generation. Genes can survive for generations without ever being enacted. But like, you know, I have genes to heal broken bones and I may never have a broken bone in my life, but I could still pass it on to the next generation. Memes aren’t like that. They have to be enacted and they have to be enacted faithfully in every generation so long as they last. Why this particular meme? And I don’t think it’s only a meme. I think there’s also a logic of the situation involved, which we also don’t understand why this has lasted for at least 2400 years. So that means it was not caused by Christianity, like many people say it existed long before Christianity, probably more than 2400 years ago. I don’t know why. It’s less important to know why than to know that it is there, because only then can we see what it isn’t, like that it isn’t racism. It isn’t xenophobia. It isn’t picking scapegoats. All those things can be caused by it. That is, the pattern can cause hatred because people will want to explain in their own minds why they have what Sartre called the predisposition to the pattern. And they will try to explain it in terms of things in their culture that are deemed in their culture to legitimize hurting people. So in cultures where being heretic is deemed to legitimize hurting somebody, people explain their behavior towards Jews in terms of religious beliefs. But atheists don’t explain it that way, but they explain it in terms of things that in their culture justify hurting people like imperialism or genocide and whatever. And it’s important, again, as Sartre said, I was very surprised, by the way, at how much sense he talked about this. Very surprised. The impulse comes first and the explanation for how they’re purporting to justify it or why they believe it comes second. And a lot of the justifications just don’t make sense in their own right. But people’s rational faculties are disabled by this pattern. It’s like a state of mind that has that effect on morality and has that effect on reasoning. Why it happens, why it is so long lasting, I don’t know.
Brian Keating
I wonder if it’s maybe the first kind of instance of such a behavior that not only oriented towards hatred of others, but also hatred and willing to sacrifice oneself. And I’m really touched by the ending of your book of The Beginning of Infinity is actually the beginning or the middle of my Bar Mitzvah Parsha, which I never had my Bar Mitzvah at age 13. So I had at age 52, four times 13 with my family at the Kotel. And my Bar Mitzvah Parsha is Nitzavim where it says the following. It says, I’ve put before you life and death, blessing and curse so that you may choose life and blessing. And the end of beginning of infinity, what lies ahead of us is in any case infinity. All we can choose is whether it is an infinity of ignorance or knowledge of wrong or right. Blessing and curse effectively. Death and life. Was that intentional? Am I reading too much into it?
David Deutsch
I did not know that passage. I learned that passage only recently from Douglas Murray.
Brian Keating
Yes, I talked to him about it on my podcast.
David Deutsch
So that’s when I first heard it. So as I also say in the book, when people seek the truth, they also converge with each other.
Brian Keating
That’s right. Because there is only one physical truth and one physical world to describe. But I wonder the reason I bring that up is because when I read the Bar Mitzvah Parsha throughout my life or when I should have done it, I thought, well, this is kind of foolish and superfluous. Surely there are no people who choose death. And choose curse. And yet I read that Bar Mitzvah Parsha portion at the Kotel on September 7th, 2023, a month before October 7th. And I was there with my family and I just thought, I’m hopeful, etc. This is a new day. And nothing could be further from the truth. I guess I wonder, are you optimistic? And my last three questions involve you. And I hope you’ll give me a forbearance of five more minutes of your time. But and then we’ll wrap up because I know it’s late there. But is there a limit to infinity in that we if we have elements and memes that choose death and choose curse willingly and they say so themselves, is there a limit to what humans can achieve? Are we universal constructors at all?
David Deutsch
So there’s no limit to the size of error we can make. That’s a feature of being general purpose. The G in AGI. We are GIs, not AGIs. But the G is there. And that means not only is the future potentially unlimited, there’s also no limit to the size of mistake we can make, including destroying ourselves. There’s no limit to the size of mistake an individual can make. And it very often happens that people make a mistake of valuing something above their own lives. Sometimes they’re right. Sometimes they’re wrong. More often they are wrong. So I don’t think there’s any upper bound, but there’s also no lower bound.
Brian Keating
Very good. OK, the last two questions come from Sir Arthur C. Clarke, your late countryman, who actually I don’t know, did you know that he’s responsible for the word podcast?
David Deutsch
No, no, yeah.
Brian Keating
So you’ll know that from 2001 A Space Odyssey, obviously. Open the pod bay doors. So when the first iPod was made, it had this white shape and it had the circle like the eye of Dave, you know, and the pod bay was the repository of protection and knowledge for the astronauts and the engineer who created the iPod in 2001 told Steve Jobs that, you know, his favorite name for it was the iPod after the pod in 2001 A Space Odyssey. So it comes from Sir Arthur C. Clarke and Steve Jobs accepted that name. So behind me is the famous saying open the pod bay doors in neon sign if you’re watching. So I want to ask you two questions. The name of the podcast is Into the Impossible, and it comes from Arthur C. Clarke’s famous phrase that the only way to determine the limits of the possible is to go beyond them, go beyond into the impossible. And I want to ask you in the form of a question, advice to your former self or what you might say to your former self. If you had 30 seconds with a 20 year old David Deutsch and you could talk to him about anything for 30 seconds only, what would you tell him to give him the courage to do as you’ve done to go into the impossible?
David Deutsch
I think that philosophical advice like that is almost impossible to follow. So I would want to give him some factual knowledge, such as think about the fact that interference processes could be used to perform computations. That sort of opens a massive door to understanding not just quantum computers, but all sorts of things about physics. Feynman, when asked this question, he was asked, what would you tell primitive people? He said that matter is made of atoms. But I already knew that when I was 20. So there’s this other thing I could say.
Brian Keating
That’s great. Yeah. He said that what contains the most knowledge and the least amount of characters or words. Yeah, I would say that the cosmic microwave background has a power spectrum that you could describe with 2500 numbers and that includes quantum mechanics within it because of the physics of hydrogen atom. But this is not my podcast. So the next question is kind of maybe an offshoot of one of Sir Arthur’s other phrases. He was quite quotable. He said for every expert, there’s an equal and opposite expert. I like to drop that on my department chair every now and then. But the final question is one where he said the following. He said for any older scientist that says something is possible, he or she presumably is very much likely to be correct when they say something’s possible. But when they say something is impossible, they are much more likely to be wrong. And so I want to ask you, what have you been wrong about? What have you changed your mind about? What would you like a mulligan on? If anything, what have you most found at odds with your former way of thinking?
David Deutsch
So I think my former way of thinking, sort of worldview or epistemology, I think I didn’t understand again when I was 20, I didn’t understand the importance of explanation. I needed to understand that more. I would have been flabbergasted if told that the pattern exists. I would have resisted that and I would have been wrong. I think the main things I’ve actually been wrong about are more prosaic. When I first encountered Everettian quantum theory, I thought it needed an extra structure, which I called the interpretation basis. Other people had the same idea. They didn’t call it that. And it took a long time before actually some philosophers, notably Harvey Brown, Simon Saunders, Michael Lockwood persuaded me that not only is this not needed, it is a mistaken theory of the multiverse. And it has all the problems, all the problems of that you were mentioning earlier about things like splitting and how fast did it happen and what is a universe and what else is there in the multiverse apart from universes and so on. Those are all caused by that kind of thinking. So I was wrong about that. I was also wrong about the multiverse being able to explain free will. So at the end of my first book, The Fabric of Reality, I give a story about how the multiverse can explain free will because it can explain counterfactuals. But what I got wrong is that it can only explain some counterfactuals. But to explain free will, we need to explain all counterfactuals, including counterfactuals of the form, if quantum theory weren’t true, then this would happen. Or if this theory that some crank has invented were true, then this would go wrong. And in other words, we have to understand a theory, we have to understand counterfactuals that aren’t anywhere in the multiverse. They are just imaginary. We have to understand those too. And since then, I’ve come to think of free will in terms of creating knowledge. So we have free will if we can bring something objectively new into the universe.
Brian Keating
Will computers have free will?
David Deutsch
It’s not computers, it’s computer programs. Computer programs will have free will when the problem of AGI is solved. Of course, if we destroy ourselves before we solve that, then they won’t. But I expect that we won’t destroy ourselves and I expect that the problem of AGI will be solved and the resulting programs will have free will.
Brian Keating
Oh, David, you’ve expanded my fabric of reality. You’ve put many, many more wrinkles into my brain. And I have another two dozen questions about technical details of quantum mechanics and the Deutsch states and Deutsch gates and all sorts of fun stuff. But maybe we’ll do that in person. I’d love to meet you someday. And this has been a great honor for me. Thank you so much for your courage, for your intellectual championing of rational thought and just the way that you’ve transformed not only what we know, but how to think, and I think that’s incredibly rare. You’re a real treasure.
David Deutsch
That’s very kind. And again, thank you for having me.
Brian Keating
Hey everybody. I’m usually the one that asks my guests to judge their books by their covers. But today I’m asking myself to judge my own book by its cover. My newest book, Focus Like a Nobel Prize Winner is chock full of advice, life tips and focus and productivity tips from nine of the world’s greatest minds. Nobel laureates ranging from economics to peace to physics, of course. It launches September 9th, which is also my birthday. I’ll go check it out. And my publisher has got Amazon to run a special just for listeners of the Into the Impossible podcast. You can get the Kindle edition for only 99 cents, that’s less than a new pocket protector. So go to Amazon and get the Kindle copy today because this special only lasts for the first week after launch. If today’s conversation with David Deutsch made you question everything about modern science, you need to watch my recent episode with Sabine Hossenfelder. We got into how mathematical beauty can guide scientists astray and what the true power of quantum computing and artificial intelligence may turn out to be. Click here and don’t forget to like, comment and subscribe.
Markdown