2026-08-23 Lucas di Grassi A Conversation with David Deutsch
YouTube Lucas di Grassi’s channel
Duration: 01:26:38
Participants: Lucas di Grassi and David Deutsch.
Topic guide
Timestamps link to the corresponding point in the recording.
- 00:00 - Introduction and Deutsch’s books
- 01:53 - Physics, philosophy, and Karl Popper
- 04:50 - Knowledge grows from problems, not data
- 08:21 - Children, astrology, and language learning
- 16:30 - Error correction in motorsport and creative work
- 24:24 - Democracy and nonviolent error correction
- 35:34 - Family institutions and disagreement
- 37:00 - Universal computation and explanatory universality
- 43:23 - Quantum computing, many worlds, and Deutsch’s early work
- 58:16 - Free will, determinism, and creative knowledge
- 65:16 - AGI, consciousness, personhood, and classical computation
- 70:49 - Atheism, truth, happiness, and the common good
- 77:30 - Error-correcting institutions from minds to governments
- 79:50 - Humanity’s future, AI, and optimism
Transcript
Lightly edited for punctuation and obvious transcription errors; false starts are retained where they affect meaning.
Lucas di Grassi
So today I have the honor to receive David Deutsch, an incredible physicist that in my opinion is one of the brightest minds in the whole world. And in Brazil especially, that is, he’s not very well known, but he wrote two incredible books that for me personally are my favorites. I’ve read them a couple of times, kind of trying to understand, because they go very deeply into the concepts of universal computing, error correcting mechanisms, the philosophy of knowledge, quantum mechanics, and many worlds and other incredible things that make the reality. And David aligns very much with my line of thought. So thank you very much, David, for being with us today.
David Deutsch
Cool, well thank you for having me.
Lucas di Grassi
David, I will start. This is a very easy conversation. The objective is to get, especially the Brazilians, like Brazil is 220 million and I discuss many topics with very knowledgeable people here in society in Brazil and your concepts, they are so powerful, yet that people don’t know that much about them or what is behind them. For example, one topic that is very discussed, which is quantum mechanics or quantum computing, actually recently, something that you developed the theory behind that so many years ago. So start explaining to us that somebody in Brazil that never heard of you, never studied quantum physics or doesn’t never study anything related to your books. How do you describe yourself and your worldview in a couple of minutes?
David Deutsch
Well, the reason I’m a physicist is that I wanted to, I always wanted to understand the physical world as much as possible. And the deepest, some of the deepest knowledge of the physical world is in the most fundamental theories of physics. But as I studied physics and found the various problems that exist in current physical theories, I realized more and more that one can’t really address these issues without also thinking about philosophy. And uh so the philosophy that that uh I found most helpful and uh I say most helpful but it was only the second philosophy that I tried that was Karl Popper’s philosophy the first was was uh Bertrand Russell which was also um nice it it uh it connected with science and uh it uh rejected mysticism and the explanations of things in terms of the supernatural and so on. But Popper seemed to me a whole other level of philosophy, a whole other level of getting to grips with philosophical problems. And they were relevant to the problems in physics, like in quantum theory. What brings about the very counterintuitive predictions of quantum theory. And there are no conservative options there. Every theory of how the predictions come about is amazingly counterintuitive.
Lucas di Grassi
On the Popper side the the just to explain for people in general right um the the idea um the idea of Popper is that is about how do you get how how how humans can know something how how can um how you infer how do you create knowledge and you have a very very interesting point is actually, I think, a very, when I discuss in this topic with some people, is very, very few people have this understanding that knowledge does not come from data alone, and knowledge does not come from observations alone, because you can, like you said in your book, right, you can observe the sun coming up and down every single day, but you don’t have an explanation about this, right? So, that’s, I guess, is where, it’s one of the key topics of your philosophical view of epistemology, right?
David Deutsch
Yes, exactly. Um so it’s uh usually uh um or or every time new knowledge grows from problems not from data at all data comes in after we have a problem like we have a problem like um we see the sun coming up every day for hundreds of thousands of years people saw the sun sun coming up every day much more than we do now by the way they really did see it coming up every day and maybe millions of people wondered why and we know that most of them made up explanations for why and we know that they were false and yet they had much more observation and data available to them than we do today so what was wrong with what they did and what is right with what we do now and what we started doing in the scientific revolution with Galileo and then with Newton and with the Enlightenment. What is it that we were doing differently? I think the thing we were doing differently was that we were trying, first of all, we were trying to solve problems, but more important, which they in a sense were also doing, but what they regarded as a solution was just a story, where sort of any story would do. But mainly we are looking for good explanations. That is an explanation which, if you change part of it, it wouldn’t be an explanation anymore. It wouldn’t account for the…
Lucas di Grassi
So specific explanations are more powerful than generic ones, right? Something that if you… Not generic on the sense that actually explains more, but the sense that any little change on that explanation actually changes the whole perspective of that specific…
David Deutsch
Yes, I mean sometimes it can explain more and that’s one, that’s another reason why maybe it can’t be altered because it explains more and you’d have to make it account for the other things as well. But my favorite example in layman’s terms if you like is if you go to a conjuring show, a magic show yeah and you and you see the the magician do the cups and balls trick with three cups and three balls or one ball or something and every time he lifts up the cup if you think it’s going to be there it’s not there if you think it’s not going to be there it is there and so you can predict what will happen you can predict he will pick up a cup and if we think it’s going to be there it won’t be there and you could predict every time what’s going to happen but that’s what you’re really looking for in thinking about the game is not the predictive it’s not a correct prediction it is how he did it it’s the it’s the explanation it’s seeing the things that it’s um understanding the things which happened that you do not see it’s the same with the sun and it’s same with every scientific theory that scientific explanations explain the seen in terms of the unseen.
Lucas di Grassi
Why do you think, I fully get it and I have two children right, I have one boy which is eight, I have a girl that is five, I have a third one coming and I try to explain, I use a lot your references and your lot especially the way of thinking to try to to try to bring them up in this already in this dimension right like for example I don’t like not that I don’t uh tell them that Santa Claus does not exist but I try I try to bring them to a world that you said look uh his name is Leo I said Leo look everything you try to you try to explain you try to understand why that thing is happening because for a child’s mind it’s it’s exactly on this what is exactly what you’re saying they see an observation and they they they try to predict or they try to they don’t try to rationalize behind but my question goes a lot of human behavior and a a lot of human, well, a lot of humans, they, when you explain to them something like that, or the way they see the world, is completely reversed from that. I give you the example, for example, of astrology, right? Astrology, it is a pseudoscience, or if you can call it even pseudoscience, I don’t even know how to describe that, but astrology is so popular. Astrology is basically trying to give like a very generic explanation or prediction of how your day is going to go, and if it goes that way, we are biased to believe that, okay, that was the explanation, or that was why the astrology actually gave me that. And if it doesn’t, ah, doesn’t mind, doesn’t matter, it can get it wrong. This is the totally opposite. But why do you believe so many humans, which we know that they are intelligent and they are cultured, and why they believe in this type of still, I would say, primitive type of reasoning?
David Deutsch
The correct theory of knowledge, Popper’s theory is the best we have, no doubt it will be improved on as well, but the correct theory of knowledge is very counterintuitive. It’s very hard to accept that when somebody tells you that the Earth is round, something like that or that the sun is going to come up tomorrow it’s it’s very hard to to realize that you are not getting that idea from them there is no transmission from their mind to their mouth to your ear to your brain that’s not how it goes their behavior is a problem that you are trying to solve. And the solution is a conjecture on your part. And every time you form a theory that seems to come from another person, you have really invented it yourself. And the first enormous achievement in doing this is by young children when they learn their native language when they learn their first language what they are doing it it almost never happens that you tell a child a table is a thing with four legs and this is one of them no most of the words that they learn they learn by seeing by hearing them used by using them themselves by seeing that they can make an utterance which somebody understands that that’s an amazing ability to be able to do that and they want to do that more and so they want to understand what is underneath the words everybody hears words but nobody hears grammar and yet the children learn grammar as well as words and so sometimes they learn a regularity and they use um a a rule which is a general rule that they have guessed. And we know that they’ve guessed it because it’s wrong. It doesn’t apply in the particular case when they’re using it. So they say, yesterday I goed to the shops or something like that instead of I went. And that’s because they’ve learned a rule. A rule is an intangible thing. They didn’t see it anywhere. And in fact it’s wrong, although it contains truth like many false theories, and in fact our best theories probably also contain falsehoods as well. So you say you would like to bring up children to have this Popperian scientific attitude. I think that kind of thing comes later. Actually doing it comes first, and they’re already actually doing it, already do it. And why do so many people believe ideologies that contradict it, like astrology and so on? Well, it’s because these theories tell you that what seems to be true is true and that is the opposite of the scientific attitude so if if children believe that what seems to be true is true then they would never learn grammar they would just just learn the words that they’ve heard and that that will that will uh never make them into fully functional language users And almost all humans become fully functional language users. And that is, in a way, the greatest intellectual achievement that we ever make. It’s the achievement of learning our native language. So why do people believe astrology? Because it’s kind of obvious. They have learned however not to do what they used to do as children because for example when they make a mistake they get punished when they uh say something that other people don’t don’t like they get mocked and so on so uh young children don’t mind they they will just you know if they say I go to the shops and somebody says no it’s I went to the shops then they will they will think about that. They will think, how does that fit in with a rule? Can I change the rule? Can I make a new rule? All that is happening unconsciously inside their mind all the time. You know, young children, I forget what age you said your children are, but around about eight, I think about eight is the time when they are maximally learning vocabulary. They’ve already learnt a lot of grammar and that at eight they learn something like 20 words a day of vocabulary so you don’t see all other 20 words going in you don’t tell them what 20 words mean but they learn it all real learning is is done by the learner not by the teacher
Lucas di Grassi
Yeah I I understand and what I try also always to to uh something that you pointed out there, which I like it very much, which is the fact of the error correcting mechanism. Yes. Right. We, I, I’m a racing driver. I’ve been racing cars. I’ve been, I was in Formula One, then we created Formula E. I mean, I’ve been racing my last, since I’m seven, so my 35 years of my life.
David Deutsch
Well done. I really admire that.
Lucas di Grassi
Thank you very much. But racing, there is a lot of physics, there is a lot of engineering. So you learn very early that if you understand how everything works, actually it’s going to give you an extra advantage because if you have a better car, you can actually, it’s a combination of the technology and the sport. And the error correcting mechanism is something that I always push my engineers and my team to do. I never had the base, the theoretical base and the understanding that actually you put on your books, but it’s once I read it of course then I clicked in my mind that this is actually much broader as a way to create knowledge and to falsify your theories than just the racing but we are using that in the sport already so one thing that I really like about the error correcting I try to that to my kids as well is the fact that um the the false the to make the a theory or to make a thought or to make an experiment falsifiable and to make it’s the only way to make progress you can’t if you can’t do this if you can’t either by authority or by whatever lack of qualitative data or whatever it is that you cannot falsify or you cannot you cannot prove your hypothesis or cannot falsify your hypothesis better than proving it because proving doesn’t really matter you actually making any progress but falsifying it actually makes so it’s very hard and exactly like you said the children they they are not and and the engineers in sequence they are not prone for that you have actually to push them to go to this direction of reasoning because they it’s not how the system actually teaches them how to behave it’s very it’s very funny but for me makes a lot of difference so one of the few things culturally we try to put inside the racing team is that the error correcting mechanisms is the best way forward even in simple so explain to us a bit more of the broader view of error correcting that fits into your theory in your books
David Deutsch
So I I very much agree that like when you’re when you’re making an an adult team that is that is focused on creating new knowledge and and succeeding in a new way you have to push them to become to adopt the Popperian uh way but I don’t think that’s true of children I I think they they already will and your job is to protect them from the many cultural forces including each other um that that makes them reluctant to do this afraid to do this um but but the people who have escaped like you evidently have uh yourself you run into the thing that as I often put it, everybody who is very creative in a certain field is a Popperian in that field. They might be superstitious or unscientific and so on in other ways, but in the way…
Lucas di Grassi
But in that space, exactly. We see that a lot. I see that exactly as you said, on that specific frame of reference and system that they operate they’re very Popperian and they’re very rational and then when they go outside they don’t apply the same reasoning for the other things in their lives
David Deutsch
That’s very, very common.
Lucas di Grassi
So you’re 100% right. Sorry, continue.
David Deutsch
Yeah. Yes. No, I agree. So so the the reason they are of course is that that is the only way of creating knowledge is like backwards is the people who are like that who are the ones who manage to succeed creatively but as I keep saying um all children do this they must be we don’t see it because it’s not written in in neon lights on their foreheads but it must be happening because that is the only way of creating knowledge so they uh uh brilliant engineers brilliant racing drivers brilliant physicists all have in common that they’re kind of very different personalities very different in in the way they they lay their lives into the into the field but they have in common that they are fearless about conjecturing a novelty. They say, what if we do this? What if we didn’t have, well, I don’t know what things are like in your field, but you know, well, what if we didn’t have four wheels? What if we had five wheels? So, you know, something.
Lucas di Grassi
Yeah no there is that is exactly a great example people are constrained in their in their box uh of reasoning or or a lot of analogies right a lot of engineering goes from analogy not from pure conjecture it’s exactly what you it’s exactly how you how you you you put you put on one of your one of your theories which is instead of actually doing analogy and doing incremental steps actually thinking from yes let’s say conjecture of what is the best how many wheels—I mean doesn’t matter the cars have four wheels how many wheels are the best wheels possible like something absurd but then it creates some form of test that guides you to the right direction.
David Deutsch
Right. That’s another important thing which we haven’t discussed yet that is when you make a mistake it might be that it was a mistake and you have to go back and start again and have a different solution of the problem. But it might also be that making the mistake enables you to go on from there so that the mistake was a step towards the right answer. And again, that’s what we do when we learn language. If we somehow didn’t ever make mistakes with language, we would never learn the real grammar. We would hardly learn the real vocabulary and everybody makes makes bold conjectures as Popper puts it they make uh bold conjectures to solve the problem and uh when you’re learning language the problem is two problems it’s what are they saying and how can I say something that that will that they will understand. And you make mistakes. And the thing is, you know, when you learn 20 words a day, like I said, you don’t learn all of them correctly. At first, what you learn is 20 conjectures about those words. Some of them will be correct. Some of them will be half right. Some of them will be completely wrong and the next time you try to use those words you will see, wait, that person doesn’t understand me then you say it another way and then you learn your mistake usually they don’t it’s not a matter of being corrected it’s a matter of failing to solve the problem of communication and you make a new conjecture and that conjecture improves on the old conjecture which might have been half right
Lucas di Grassi
Yeah I agree and that leads me to a problem that we have in common me and you which is the which I’m gonna translate to this error correcting to a country level right we Brazil is one of the biggest democracies there is right and I know that you’re I’m also very in favor of democracy, not because of democracy itself, but because of democracy, it creates a mechanism, theoretically at least, it has a mechanism to actually error correct itself and move forward as a system of governing people, right? It’s not that it’s the best by default, it’s just that it keeps balancing, theoretically it should be balancing each other to correct and remove what doesn’t work. And England is going through an interesting phase now, right? With the changing leaders, a couple of leaders recently. Brazil, we have election this year in October. And of course, we are in these turbulent times, very irrational type of teams, politics, social media biases, people that are not using the rationale in the sense that okay what were the mistakes what was the error correcting they don’t even know that not even consciously not even unconsciously on how to deal with that but Brazil we don’t have a parliamentarism we have a presidentialism different from the UK but we are in the same situation right we are kind of and the more I see the more I see mechanism to suppress this error correction or contrary in the direction that for example some censorship some uh language that cannot be used some some points that cannot be challenged you know this type of stuff which goes back into the philosophy of error correcting which could be from for a physicist studying quantum mechanics or for a racing driver testing different things but also works for democracy and I know that you you’re very active on X actually I love your posts and thank you you’ve been talking about um politics recently quite a lot so explain to me what is your view there what is your uh it would be very interesting to understand your your mind on this topic
David Deutsch
Well you have just described it uh like Popper I I accept his his uh way of looking at politics at large scale politics that that um the most important thing in first of all the important thing is to judge political systems by their institutions rather than by the decisions that they make from day to day. The decisions that they make from day to day are going to be false, inconsistent, disastrous, sometimes good and so on. But the institutions are much more important and in particular whether they promote error correction, that is non-violent error correction yes and and uh or whether they suppress even non-violently error correction because if they promote error correct first of all because there is no such thing as a government or policy or a leader that doesn’t make mistakes even the best leader will make mistakes and if they can’t be corrected, then there will be a disaster. Some people think that Napoleon was a great leader, other people think he was a monster. But even if he really was a great leader, he set up a system where his errors could not be corrected, and then millions of people had to die. So even supposing he was the best possible leader, that’s not the important thing. The important thing is the institutions that would remove either him or his policies and so on, long before millions have died, find a different solution to the problem. So I think that and I also think that this feature of political systems has been brought to its best realization in Britain, basically. Although the actual politicians are not necessarily very good.
Lucas di Grassi
But I also think it’s in politics, it’s very, I think one of the main problems with error correcting is actually defining error in politics. So how do you define error in terms of short term, medium term, long term? It’s very hard to make sure that the people, whatever the decision is, how do you frame, how can you have the right data or the right feedback?
David Deutsch
That’s not a feasible property of a political system. You can’t make a political system that is intentionally short-term, intentionally long-term or medium-term. That itself gets decided by error correction. The important thing is, the way, for example, to avoid violence is that if somebody has an idea, that is unpopular and they feel very strongly that they’re never it’s although it’s unpopular it’s nevertheless true and uh they they have uh there’s an election and they lose the election and they get the idea that uh the people don’t understand it this and and we the this this minority that does understand it have to do something and that’s where violence comes from and so instead a good democracy in the sense of Popper will say to these people look the first thing is if you’re right why can’t you persuade anybody well they say well because they’re all the wrong class that never mind they are humans they are universal they can accept arguments There was a time in the late 18th century and the early 19th century when the upper class in Britain was persuaded by middle class people to abolish slavery. And so they abolished slavery and all the big slave owners, they were compensated, but they couldn’t make any more money. They had to invest their money in something else. And so this thing couldn’t quite happen in America and so they had a civil war in which in which one or two percent of the population was killed because they couldn’t solve the problem consensually and that was because there’s a flaw in their error correction process the the constitution is supposed to prevent the president from taking over and becoming a tyrant But it has the unfortunate side effect that if the Supreme Court rules that something is the case, then persuasion no longer works. If you persuade 50% of the people to vote against slavery, for example, you don’t abolish slavery because the Supreme Court has said that slavery is legal. And that’s that. To get rid of slavery, you’d have to get either a two-thirds majority or something which was specially designed to prevent the correction of errors. Now, you’ll say, well, yes, but if you don’t have something like that, you might try to correct an error and actually make a worse error. Yes, there is no way of avoiding that. That is going to happen in any system and as usual the way to deal with that problem is to have more error correction better error correction more sophisticated error correction more error correction built into the fabric of what ordinary people expect from their government from each other um from their um from the economy and so on. They expect things and these things that they expect have to have as much as possible of error correction in them. So that when you feel very strongly that Mr. Burnham will ruin the country and he’s a bad leader and so on and and and uh you’re let’s say you’re the uh previous prime minister and you think he’s going to ruin the country you still leave not only do you leave peacefully because he’s the constitution you know he’s he’s the one that the error correction error correcting processes have uh selected and you might think they’ve selected wrongly this time that’s less important that the important thing is that those errors can also be corrected and so in a in a good system like the British one not only does the previous leader who thinks there’s going to be a disaster leave peacefully he would actually fight and kill and die for the right of the new leader who he thinks is bad to take his place yes that’s that is the property of yeah
Lucas di Grassi
No, it’s exactly that. This property, which I think is instead of being the focus into the policies or the leader itself, like you said, and I think I completely agree, I never thought about that in this way, but the metric that you have to look in terms of democracies are eroding or not, are there a number of error correcting mechanisms that are in place. Regardless of which policies or which decision of the state of the country or anything. So I think that’s where we have to look at and we have to find where are the flaws or the mechanisms that are not working and change them to basically increase error correction with time.
David Deutsch
Totally agree. And we may not agree that I think this is also true of the parent and an eight-year-old. I think if an eight-year-old wants ice cream in the middle of the night, then what’s important is not whether he gets ice cream or doesn’t get ice cream or the ice cream harms him or doesn’t harm him. This is not important compared with whether the institutions that determine what happens are peaceful and whether you as a parent will defend the wrong decision, provided it was made in terms of family institutions that can solve problems. The particular problem going wrong, that’s always going to happen. You can’t focus on that. You must focus on the institutions which govern how policies are changed. So the institutions can be lucky in the best case. There should be persuasion. And if persuasion doesn’t work, then you try more persuasion. But in the meantime, he’s got to have the ice cream, in my view.
Lucas di Grassi
You can apply that from, so the beauty of this is that you can apply that from actually teaching a child to a giant democracy of a country. But I can’t agree more. But we are running out of time and there is so much to cover. I want to go to The Beginning of Infinity, right?
David Deutsch
Right.
Lucas di Grassi
Which is your latest book.
David Deutsch
Yes.
Lucas di Grassi
And which is an amazing, which is for me, I see that as such a, I see this with so much optimism and I use it so much in the current world of the AI discussions that they say, no, no, no, AI is going to solve all problems and there will be no more problems. So that’s not true.
David Deutsch
No.
Lucas di Grassi
Actually, there will always be problems.
David Deutsch
Yes.
Lucas di Grassi
There will always be a mechanism to solve these problems that will emerge other problems.
David Deutsch
Yes.
Lucas di Grassi
Regardless of how level is the intention. So there are many topics on The Beginning of Infinity. But first, I’d like you to describe for a layman again, the concept of, which for me are very key, which are the theory of universal computing, right? That we are universal explainers and that’s how we evolve. That’s how human brains actually work. This is not only humans, but every single analogous universal computer are solving the problem in the same way. There is nothing better or worse. It’s just a matter of time and memory that is necessary. But the explanation… People always come to me and say, oh, the aliens will come and the aliens will have something that we don’t even understand. I said, maybe we will understand if it gives us time and memory, regardless of what it is. And I always use this in my arguments, but people, a lot of people don’t have that as a base truth or they don’t understand exactly.
David Deutsch
Yes.
Lucas di Grassi
So explain to me about humans being universal explainers and why problems will always emerge and why AI is not going to solve all the problems or how do you see this?
David Deutsch
Uh um the first the first thing to distinguish is that there’s there’s hardware universality like the brain is a universal Turing machine among other things its hardware is capable of performing arithmetic and and you’re doing every kind of uh task that the Turing universal Turing machine can perform and then there’s software universality which is the one that people tend not to understand because it’s counter-intuitive. It’s one of these things like why irrational ideas are often more plausible than rational ideas. And that’s down to software universality, that is explanatory universality that you just referred to. So, first of all, imagine that they’re aliens that can compute a more complicated function than we can. Let’s say they can decode a code that has 256 bits and we can only decode one that has 128 bits and so on. Well, obviously, everybody understands now, if they know anything about computers, that the difference between those two cannot be a different kind of computer. There can only be two differences between any two computers, namely speed and memory capacity. So if we add some more memory to our computer and program the same program that they have, it will be able to do anything that their computer can do. Now a similar but different thing applies to software universality. So can they explain something that we could never explain? So is that possible? You know, they have huge heads like in the science fiction. Yeah. Is it possible that they have explanations that we can never even contemplate. We can’t even understand.
Lucas di Grassi
And we can take aliens out of this phrase that you said and put AI on it.
David Deutsch
AI, same thing.
Lucas di Grassi
It doesn’t matter, right? So we can say aliens, AI, whatever.
David Deutsch
Exactly the same thing. And by the way, also children, adults, you know, it’s all the same theory, all the same theory of knowledge. So imagine that then that the AI with its huge banks of of memory and of processing processor power. Suppose, and it’s going to be a reductio ad absurdum this, but suppose for the sake of argument that such a computer can understand something that we can never understand. Well, why? Why can’t we understand it? Well, this understanding consists of a program in its memory. There can’t be anything else going on there other than a running program. What can prevent that program being understood by a human brain? Well, only memory capacity and speed. But we can increase our memory capacity and speed in exactly the same way that the computer does, by adding more memory. And I already have additional memory and additional speed. And with this, I can do things, some things I can do a million times as fast as I could do it without this. And the task of understanding this task, which is supposedly done by this superhuman AI, cannot be anything other than a computer program because of the first, because of hardware universality, because of Turing universality. It can only be a Turing program running. And therefore, the only reason we can’t run it is lack of memory or speed. But we can augment our memory and speed using…
Lucas di Grassi
So, I was not the original question, but I would like to make a parallel here. So, when you say… So, what is the next question that I have actually also? Then the quantum computing capability. Is it a different form of computing or it obeys also these laws?
David Deutsch
It obeys these laws. It is a technically a more powerful form of computing than Turing computing. Yeah, so however, there are ways for a Turing computer to emulate a quantum computer with a lot more memory or or with a quantum computer as a subroutine.
Lucas di Grassi
Okay, understood. So the quantum computing, if you give a classical… So basically a quantum computer, it is a classical Turing-universal computing machine, but it’s just structurally the same, but it just calculates in different…
David Deutsch
Yes, it’s structurally superior. It can perform operations that the classical one cannot perform without an exponential increase in speed or memory capacity required. So an exponential increase means without trying every possibility, which is not really a feasible computation. However, a classical computer can be given a quantum computer as an ancilla, as they call it. It can program the quantum computer and the quantum computer can give it an answer. So there’s no fundamental obstacle there.
Lucas di Grassi
Okay. And you argue that we’re going to go into that, I hope, a little bit later on, that basically which was one of your amazing work, that given the fact that quantum computing, the way that quantum computing works is that it calculates or it processes information in parallel universes, right? In different universes. But before actually we go there, I would like you to explain to us, for people that does not, Because we’re going to go a little bit into quantum computing. There is a lot of woo-woo. And every kind of discussion I go, anything that is not known or there is many possibilities, they go into quantum because nobody really understands anything. I also don’t understand anything, but at least I try to figure out what actually is weird and counterintuitive but actually happens in quantum and what is just complete nonsense that has nothing to do with anything else. So, I would like you to just to explain to us, also for people that does not know quantum physics so much, there is a discrepancy that I could say between the, I think it’s not common sense, but a lot of scientists believe on the Copenhagen explanation, the Copenhagen interpretation of quantum mechanics, right? That there is this infinite of possibilities. And when you observe something, it collapses into the reality that we are on. And there is your, and which I believe also, and other scientists are more, I see at least more people going towards that direction, that there is, that actually everything happens in infinite universes in parallel, which is where the math leads us, which is the Everettian way of interpreting the way the quantum mechanics work. So explain to us a little bit these discrepancies between collapsing of the wave function and the multiverse, where the multiverse comes from.
David Deutsch
Yes, well, I think the easiest way to enter this area is to stay with quantum computers for a moment. And if I tell you how they work, so you say they do different computations in different universes, that’s a kind of shorthand for something that’s not quite so bizarre, which is that the description of subatomic events in quantum theory is extremely complex. It is exponentially more complex than the classical description of the same thing. So when the things that we can see, we have, like we can see that one particle is there and one particle is there, and for each particle we can give some numbers that say where it is, and then we can apply classical laws to determine where it will be and so on, and a classical computer works in that way. It makes the particles go around. Now, the quantum description is exponentially more complex. So that’s, you know, really mind-boggling how much more complex it is. But most of the time, the extra complexity is just washed out, and we can’t see the effect of the additional complexity, and it just resolves itself into. Now what does it resolve itself into um uh I can put it this way inside the quantum computer that there’s this exponentially complex complex stuff going on but it resolves itself into autonomous strands of events each of which um this is inside the computer each of which um proceeds as if the others were not there. So there are like two to the power of a hundred different computations going on, each of which is going on as if the others were not there. It’s independent of the others. Now normally, if we open the…
Lucas di Grassi
That means the parallelism of the computation.
David Deutsch
Yes, yes, but it’s not like a classical parallel computer where you need a hundred different computers to do a hundred different things. This is just one computer which can do two to the power of a hundred different things simultaneously.
Lucas di Grassi
Where does this computation is actually happening?
David Deutsch
Inside the computer.
Lucas di Grassi
I know, but there are the ones which are not actually happening there.
David Deutsch
No, no. Let’s talk first about the ones which are happening. Okay. Two to the power of a hundred of them are happening inside the computer until you look at it. If you measure it, that is, you send in a laser pulse or something like that, which detects which thing is happening, then this complexity washes out again. The very fact of the laser hitting the interior of the computer makes the um I didn’t finish the story that there are these two to the power of a hundred different computations going on that would be useless if we couldn’t bring them together again so if we want to ask a question that that is only answerable by all of the two to the power of a hundred computations we we need to know some aspect of all of them like are there are there more of a of one kind than another or something like that where where you would have to look at all of them to get the answer then we can perform a quantum operation that brings that answer into all of them and then that’s that’s the answer to the quantum computation but if we don’t do that extra computation at the end to bring them together then all we will see is just one of them selected randomly so that doesn’t help that would be just as much as doing a one yeah yeah so uh now if you look a bit more closely then what really happens when you make a measurement when you send the laser in and the laser then bounces off and hits the detector and you look at the detector and the photons from the detector hit your eye and so on that whole process remains highly complex it’s just that the different parts of it are now no longer affecting each other so if they don’t affect each other you can only at most see one of them so it’s for you it’s as if the others stopped existing you can’t make that move for for the process inside the quantum computer because all of them contribute to the answer
Lucas di Grassi
But that means the observer gets entangled with that measurement.
David Deutsch
Exactly. But it doesn’t need a conscious observer or anything like that.
Lucas di Grassi
No, no. A conscious observer is where the woo starts, right? People use that. It could be any set of atoms, basically.
David Deutsch
Yes. Or a laser.
Lucas di Grassi
It could be a laser beam, or a laser, or whatever—or anything getting entangled with that specific—
David Deutsch
Quantum state. Yes. So all the all the uh autonomous channels um are realized in reality and because we are part of that multiplicity if you ask how many are you seeing we will say one just like if you ask the um quantum um quantum computation in the middle carefully ask it how many computations are you performing they will all answer in unison one but there are two to the power of a hundred of them and yet they’re answering one because each of them is only performing one the difference is that with a quantum computer you can then make them um interact with each other in such a way that there a certain function of them gets transferred to all of them and then then whichever one of them you look at you will you will see the answer.
Lucas di Grassi
When did you start theorizing about this process it was because you were the the first or one of the first ones to theorize yes in computing right What was that when did you start having
David Deutsch
It’s a it’s a slightly weird story because the question you ask isn’t quite the right question I first theorized about quantum computers and realized and proved that there is such a thing as a universal quantum computer in theory whether or not we can build one but there is a universal quantum computer the same way that there is a universal Turing machine that was in 1984 or 1985 something like that
Lucas di Grassi
When I was born
David Deutsch
Right yeah it takes a long time to catch on yeah um but uh already in 1977–78 I I I wrote a paper in which I described a quantum computer a universal quantum computer but I didn’t know that’s what it was I just described it as a machine to test the many-universes interpretation of quantum theory. So with this and to make this machine work, it had to have a computer that operated according to quantum theory. I described how it would work and I described how you would get an answer. You get one answer if the wave function collapses and a different answer if the wave function doesn’t collapse, like Everett said. And so that’s what that paper was about and I never used the words quantum computer or anything like that because I was just thinking of it as a physical object I wanted it for some other purpose not for computation and it was only several years later that I realized the connection with computation in general and with universality
Lucas di Grassi
Do you think the scientific consensus is going towards more this vision now? Because I think when you wrote it, it was quite revolu… I mean, it was very revolutionary today, but I guess that was not the consensus at the time, right?
David Deutsch
It was not. It depends consensus among whom… I think if you ask people who work on quantum computers, they will typically say that they adopt the Everett interpretation. A few of them don’t, but most of them do. But if you ask people who work on, let’s say, lasers, solid state physics, something like that, they will usually nowadays, they will say, I’m not interested in that kind of question. They would say, I’d rather work out what happens rather than try to understand why it happens. So that’s not a scientific attitude, in my opinion. That’s a very bad attitude and can’t lead to progress. But unfortunately, I think most actual working physicists would take that view nowadays. By the way, that’s not the Copenhagen view. But the difference between the Copenhagen view and what they say nowadays is is a technical thing that is probably not of interest here
Lucas di Grassi
Good so I’m going back towards more I would say more mundane and normal and normal topics that that I’d like that I’d like you to to explain um which for me then it it is interesting so we’re talking about universality and how knowledge is created and so on um my question is on on creativity and the AI and large language models and where we are now right there is a big discussion can the LLMs achieve AGI do you need a new technology does um how human creativity I mean you can go and even says I don’t know are you a determinist do you believe do you believe do you believe in free will as a
David Deutsch
Yes I believe in both
Lucas di Grassi
You believe in both
David Deutsch
Yeah
Lucas di Grassi
Okay explain so before we go into that to explain to me so creativity in the terms of free will determinism and creativity and then maybe do a link with artificial intelligence maybe in this sense
David Deutsch
Uh yes I’m not sure if the connection is so strong between those two I mean I have I have views on both so first of all free free will and determinism uh people like to say we don’t have free will because we can only do what the atoms make us do but that is a mistake because and they say you know if you say you’re you’re you’re you’re doing something new in other words your your your thinking is bringing something new into the world if you think that it isn’t because it was all determined by the by the atoms that make up your body and and you think it’s new because you you just uh you you just see it but unconsciously it was just caused by the atoms now that is that is a um quite fundamental mistake about how the physical world works we’re we’re not forced to do things by atoms atoms are just doing things and sometimes they do free will things they they always do deterministic things but sometimes they do free will things and which are those they are the ones that create new knowledge
Lucas di Grassi
How does an atom do a free-will thing? Sorry.
David Deutsch
No. Atoms do not have free will.
Lucas di Grassi
Yeah, no. But you just said that atoms do free-will things. What does it mean?
David Deutsch
Together. So, together—like in the brain. The brain is made of atoms. That collection of things can do something radically new, fundamentally new, which is not predictable from just knowing the laws of motion of atoms. You can’t predict, not just it’s too difficult, but you can’t even in principle predict how these atoms are going to behave. If you thought you could, then you would be reduced to ridiculous claims like mathematics is not the study of abstract objects, and some people do say this, mathematics is not the study of abstract objects, mathematics is just the study of mathematicians’ brains. Yes, yes, yes. But that is untenable. When mathematicians that there is a reason why a mathematician prefers a valid proof to an invalid proof and he can’t choose he can’t decide to wake up on a Wednesday and say well now I’m going to flip I’m going to prefer invalid proofs that he can’t do that because the the thing that he is thinking about this mathematical object he’s thinking about is autonomous now I’ve used the word autonomous twice once for quantum computers and once for for um uh collections of atoms that that are creating new knowledge and they and they are linked um so so it’s it’s not true that we’re forced to do things by atoms atoms aren’t forced to do things at all they just do things and some of the things that they together can do is create new knowledge and that is a thing that is fundamentally unpredictable not random random is also unpredictable
Lucas di Grassi
But does unpredictability necessarily mean free will in this sense? Because it does not mean—
David Deutsch
It is the other way around. Free will necessarily requires unpredictability. But unpredictability is not enough for free will. Unpredictability could simply give randomness.
Lucas di Grassi
But, for example, let us say a neuron is gonna it’s gonna shoot an electrical signal or not that depending on the chemistry and the basically the the electron gradients you have between these two that creates the different so how can you interfere on this how does it how this free will is created in this process that’s always intrigues me for example
David Deutsch
So, you know, if I knew the exact answer to that I could make an AGI so I don’t know the exact answer but I know like many bad answers and I know why they’re bad so so uh the reason why why um so in the brain there are many things going wrong all the time just like in a government so the the important thing is the thing that makes a brain a knowledge creating system rather than just a random collection of atoms is its error correction abilities so if you if something accidentally were to go into your brain uh what were to happen in your brain that accidentally changes your memory of what let’s say uh with the difference between left and right for a moment then then you would do a double take because you have many error correcting processes happening an enormous torrent of error correction is happening all the time in your brain which has other knowledge about left and right and if you suddenly think, wait a minute, that’s not left, this is left. And so you correct it, and so the momentary thought, left and right were the other way around, gets erased before you even consciously are aware of it. And this error correction is the reason why molecular indeterminism or molecular randomness can add up into creativity at a higher level. It’s because of higher level error correction
Lucas di Grassi
Interesting. I never thought about it this way. But it yeah it’s so so you’re saying that any mechanism—so then creating this analogy to AI so if you create the right structure or the right so the right structure of error correcting of you could say that you could say that AI could be as creative or have the same type of
David Deutsch
Well, it would be then it would be AGI then yeah and we don’t know how to do that but yes I hope that one day we will know how to do it and then we will be able to make AGIs and by the way they will be people with rights and owning property and integrating into a society.
Lucas di Grassi
So that’s actually a very good point. So do you believe to go to AGI, we need to create a new technology that is basically on this, will be around the self-correcting mechanism between the, let’s say, the nodes or the structures of AGI.
David Deutsch
But the self-correction that happens in an AGI is almost the opposite of what happens in an AI. That’s the thing. An AGI has got free will. It can decide for itself. So to make one, you’ve got to make an AI, you have to make sure that as closely as possible, it meets the criteria. If you say, what is the best chess move? It must produce the best chess move. If you ask it, is this picture a cat or a dog? It must say the right answer with a high probability, or most of the time it must produce the right answer. If you increase the number of things that it can do of that type, you are actually going further and further away from an AGI, not closer towards it, because an AGI doesn’t have to play chess if it doesn’t want to. So an AGI that plays chess can one day come up with the answer, I don’t want to play chess anymore, I want to play a different game. Put me into a robot and I’ll play tennis. Or it can be silent. It can choose silence, just like a human can choose to be silent. Because the human can decide that they don’t like the society and they want to be alone.
Lucas di Grassi
But that AGI that you’re framing on, that’s almost like you’re talking about consciousness as well.
David Deutsch
Yes, well the G means that it has to be capable of consciousness among other things as well, otherwise it won’t be general.
Lucas di Grassi
Well, but it could be that you have a general machine that actually creates knowledge without being conscious or that’s a necessary step?
David Deutsch
I don’t know. I’d guess that the answer to that is definitely yes. But I don’t have this theory. I just think that all these things, free will, consciousness, qualia, we haven’t mentioned, creativity, capacity to do science, all of them evolved remarkably fast in an evolutionarily very fast time. All of them evolved together. So humans, at basically the same time that they were able to invent campfires, they were also able to think about the sky and invent gods and invent new cultures and invent all sorts of new ideas because they had free will as well. I can’t prove that these that these five or six things could not happen separately which would raise a number of important problems if they could but I I think they can’t I think they’re all and and the AGI the the general one is definitely capable of all of them even if it doesn’t actually do so. For example, it could decide not to have free will. It could decide that from now on, yeah.
Lucas di Grassi
Do you think necessary, not necessarily, but do you think that because I think it’s Sir Roger Penrose that actually most of his talks, he frames a lot of the consciousness with the quantum the way quantum operates in the brain back and forth I’m not going to go into these details because I don’t even understand exactly I I read a couple of articles but I don’t but my point is um do you believe that for to create AGI also in this process that you’re describing you need quantum uh you need quantum computing you need quantum operations I think I guess no because or or or yes what is your view
David Deutsch
No no I I think AGI can be achieved with a classical computer
Lucas di Grassi
Okay
David Deutsch
Um and I I think the brain is also a classical computer in the relevant sense I think the the um thermodynamics of the brain which is the kind of the lowest level of error correction um is is already classical at that that level and there are many many levels of error correction above that before it gets to the conscious level or the creative level
Lucas di Grassi
Uh uh now just to finish off so thank you so thank you very much I have a few questions actually that I that I normally I’m trying to put together these questions to try to understand better the person so um first do you are you a religious person or do you believe
David Deutsch
I’m an atheist.
Lucas di Grassi
Ah, okay. So I’m too. I’m an atheist. So that’s very straightforward. In sense, so my second question is, which is just to understand you better. Do you believe that the search of truth is more important than the search of happiness in the sense of that? I had a lot of this philosophical discussion with friends of mine that, do you believe that sometimes not knowing something that actually does not, that makes your life less stressful or less, because you are going to, you’ve spent your life going to so much. Trying to understand truth at the deepest level of reality, right? Yes. Does it make you happier as a person? Or do you believe in general, maybe not knowing some stuff actually would make life easier for the humans?
David Deutsch
So I think the two things are not separable. You can’t have a conception of happiness that is independent of problem solving. It’s not so much truth, because we never reach the truth.
Lucas di Grassi
Yeah, truth on the frame that you just said, so actually problem solving, the next level of truth.
David Deutsch
Exactly. So problem solving is the same as happiness. So if you just seek happiness in the abstract and don’t try to solve problems, you will fail. No matter how much that was your priority. And we can’t define happiness independent of problem solving anyway. Because some people enjoy doing things which other people absolutely hate so for example most people would hate to have to race in a Formula One race
Lucas di Grassi
Yes.
David Deutsch
They would find it terrifying they wouldn’t they would get they would give up a lot of things they would they They would give up food and so on in order to not have to do that. And yet other people, that is the prime source of joy and happiness in their life. So this shows you that you cannot define, this idea that there’s happiness on one side and truth on the other side is a mistake. Our attitude to truth and happiness are both determined by our attitude to problems and to problem solving. Whether we are solving them, whether we hope to solve them, whether we think we can solve them. And when you’re in the process of solving problems, that’s when you are happy. Something makes you unhappy like um you know an illness or a bereavement or something like that what you’re doing when you’re experiencing that is you’re trying to solve the problem you’re trying to enter a state of mind where you are happier
Lucas di Grassi
Yes. Which leads me to so it’s following this train of thought which is the next question so do you believe so for example if if a group of people or governments or the way society is organized. So if some people are ignorant in a certain area, is that acting upon their willingness to do something or is that the goal or the objective of governments or is the rights of governments actually to choose by having more knowledge to choose what what people can do or cannot do in in the sense that if it’s for the common good if it’s the good of society do you believe that government should act on it?
David Deutsch
I don’t believe there is such a thing as the common good there are only problems and okay and so at a particular time we have various laws, we have various means of enforcing the laws, we have various means of changing the laws, and all these things are imperfect. And all these things, if they’re not corrected, the imperfections will grow and will destroy the system. So we, and it’s not true, it’s not right to think of the government as something with more knowledge. They might have more knowledge of you know how uh how uh scientific theories work or at least they can hire people who have more knowledge and so on but I I again as usual side with Popper on this matter he said um we one some of us may know a bit more about one thing or another but in our infinite ignorance we are all the same so we are all alike in our infinite ignorance and that’s the important thing it’s not the knowledge the different knowledge that we have that should determine how we interact with each other what should determine it is the institutions that that make error correction because we are all infinitely ignorant including the government including the people, including the parents and the eight-year-olds. All of them are infinitely ignorant and some of them have little bits of knowledge that the others don’t have. Everybody has little bits of knowledge that nobody else has. And so we all build, try and solve problems and build up our own knowledge in order to solve problems and become happy.
Lucas di Grassi
If you if you could say to to us what what do you believe so what do you believe people should um understand more study more uh how what would be your recommendation if you’re in a position of of deciding of how education or how people should uh You talked a lot about the decision-making process and error correcting and so on. How do you resume that into an actionable practice that people should proceed with their lives to have a mentality like yours? What do you think is necessary?
David Deutsch
They have to have institutions both inside a single mind and also in a family and in a in a subculture and in the culture and in the government and in the world all those things need to have um error correction non-violent error correction um where where um particular theories are not entrenched and unmovable. So there should be no particular… This is an ideal, of course, in real life everything is imperfect and doesn’t have this property well enough. But the more a mind or a family or a culture or a government or whatever has this property, the more it will be able to create knowledge and therefore solve problems and therefore proceed towards more happiness and the more it has aspects of the institution that thwart or sabotage error correcting processes the more it will be thwarted in the solving problems and therefore the more it will be thwarted in making making itself or making people happy happier always happier there’s no such thing as happy
Lucas di Grassi
Yes. Happier. And my last question, I promise. Uh but what is are you excited about the future of humanity a lot of people I see a lot of parents and even people saying like I’m not gonna have kids because the future of humanity is doomed because of AI there will be no jobs and this and that what is your view you’re probably you know spend the last I don’t know how many decades studying knowledge studying the progress studying uh how how information is processed and and work so what is your view uh for the for the future of humanity and are you optimist are you uh how do you see um
David Deutsch
I can’t predict of course the future of humanity because the future of humanity is the most unpredictable thing in the universe because it is the most affected by the growth of knowledge or the prevention of the growth of knowledge and it’s it’s it’s it’s the most thing in the universe that is most affected by knowledge so I can’t predict how it will go but I can predict what it will do first of all I can say that up to now, up to the present time, the present time is the best society that has ever existed on this planet. By far the best. Now this is not a close-run thing. It’s by far the best that has ever existed on the planet. What will happen next depends on what choices people make and what knowledge they create I don’t think that uh the technology of AI of LLMs is is any kind of a threat to humans of course there are problems with it there will be problems with it just like there are problems with computers in the first place and with the internet and so on and you know there there are people who have died because they have been the victim of a scam on the internet and so in a sense they were killed by the invention of the internet but many more people were saved by the internet in the sense that they would be dead if it were if it unless if it were not for the internet and that that’s true of every technology even atom bombs even atom bombs have up till now
Lucas di Grassi
Prevented more deaths than actually—
David Deutsch
Yes. Yes. And and those those are like the usual examples that people take of a technology that is bad, but it’s not. And of course, it all depends on how it was used. And it could have been used badly. It could have been used to create badness and to destroy the human species. It could have, but it wasn’t. And it still isn’t today. Whether it will be tomorrow will depend on the decisions that we make we are decision-making animals knowledge creating animals problem solving animals the problems that come out of AI um and I argue in the book that the problems are inherently soluble yes even though one one cannot by an act of will necessarily solve them on a particular occasion, but if you fail to solve them it’s not because they were insoluble, it’s because you didn’t think of the right idea in time. So it’s incumbent upon us to, again as Popper says, to be optimistic in the sense that we think that problems are soluble and that our job is to remove the obstacles to solving them. So we have problems will definitely arise. And when AGI happens, completely different problems will arise. Problems that people, you know, people who think that a bit more AI will make AGI, they kind of completely misconstrue the kind of problems we will have when there really are AGIs one day, which are the problems like, what are their rights? Do they have a vote? Are you allowed to unplug them? Is that murder? You know, the problems of that type. But we’re nowhere near that. I mean, of course, I can’t predict. So I can’t predict that tomorrow somebody won’t have the idea of how to make an AGI. They might say, wait a minute, we shouldn’t be doing this like LLMs. We should be doing this.
Lucas di Grassi
It is a problem. It is a problem that eventually will be solved.
David Deutsch
It could be solved. Yes. I hope it will be solved. And, you know, I hope I am still alive when it is solved. But when it is solved, that will not be the end of problems. There will just be more problems and better problems.
Lucas di Grassi
Very good. So I think it’s a great way to finish. David, it’s been a big honor. Thank you very much for this one and a half hour talk. A talk you are uh in my opinion uh uh one of the brightest minds uh we have in the world today I really recommend your two books for anybody that has interest about well about anything really because I think your theories like we discussed they go from children to governments they go from educating children they go from how do you see the world how do you operate in a corporate environment And the ideas are so broad. And when you go fundamentally to understand the first principles of these ideas and why they come along, the explanations you give in the book, they are so powerful that I think it changes how people approach life and will make their lives better. They will solve problems in a different way, faster probably, and in a better way. So I think you contributed so much for humanity in general, and you’re a true legend. So thank you very much for this talk. And, well, I’ve lived in Oxford for two years, by the way. So great memories from the UK, from Oxford. And I wish you’d get better. I wish you get well and one day you will be able to meet in person because you’re one of a kind human that deserves all this recognition.
David Deutsch
Well, thank you for all that and it was a very pleasant conversation. Very interesting.
Lucas di Grassi
Thank you, thank you very much, I appreciate it.
David Deutsch
Okay,
Lucas di Grassi
Bye-bye.
David Deutsch
Bye-bye then.
Markdown