[00:00:00] Host Create clip the following is a conversation with your dad Pearl Professor Use L. A and a winner of the touring award that's generally recognized as the noble prize of computing, is one of the seminal figures in the field of artificial intelligence, computer science and statistics. He has developed and championed probabilistic approaches today I, including Beijing Networks and profound ideas and Cause ality. In general, these ideas are important not just to say I but to our understanding and practice of science. But in the field of a I. The idea of cause ality, cause and effect too many lie at the core of what is currently missing well must be developed in order to build truly intelligent systems. For this reason and many others, his work is worth returning to often. I recommend his most recent book, called Book of Why that presents key ideas from a lifetime of work in a way that is accessible to the general public. This is the artificial intelligence podcast. If you enjoy subscribe on YouTube, give it five stars. An apple podcast supporting patri on are simply connect with me on Twitter Alexe Friedman spelled f R i D M. A N if you leave a review on Apple podcast, especially but also cast box or coming on YouTube, consider matching topics people, ideas, questions, quotes and sci tech and philosophy you find interesting, and I'll read them on this podcast. I won't call out names, but I love comments with kindness and thoughtfulness in them, so I thought I'd share them with you. Someone on YouTube highlight a quote from the conversation with no Chomsky, where he said that the significance of your life, it's something you create.

[00:01:44] Host Create clip I like this line as well. On most days, the existentialist approached. The life is one I find liberating and fulfilling. I recently started doing as a the end of the introduction. I'll do one or two minutes after introducing the episode and never any ads in the middle that break the flow of the conversation. I hope that works for you and doesn't hurt the listening experience. The show is presented by cash up the number one finance app in the APP store. I personally use cash up to some money friends, but you can also use it to buy, sell and deposit big coin in just seconds. Cash up also has a new investing feature. You can buy fractions of a stock, say $1 worth, no matter what the stock prices broker service is provided by Cash Up Investing, a subsidiary of Square, a member at Site B. C. I'm excited to be working with cash up to support one of my favorite organizations called First. Best known for their first robotics and Lego competitions. They educate and inspire hundreds of thousands of students in over 110 countries and ever Perfect rating and Charity Navigator, which means the donated money is used. The maximum effectiveness.

[00:02:55] Host Create clip When you get cash out from the APP store or Google play and use code Lex podcast, you'll get $10 cash. Apple also donate $10. The first, which again is an organization that I've personally seen inspire girls and boys the dream of engineering about a world. And now here's my conversation. Would you dare Pearl You mentioned in an interview that science is not a collection effects by constant human struggle with the mysteries of nature. What was the first mystery that you can recall that hooked you that

[00:03:30] Host Create clip capital the Christmas tree? That's a good one. Yeah, I remember that we'll have the fever for three days when I learn about the cart on a little geometry. And I found out that you can do all the construction in geometry using algebra, and I couldn't get over it. I simply couldn't get out of bed.

[00:03:58] Host Create clip What kind of world is and look? Geometry unlock

[00:04:02] Host Create clip It connects algebra with the younger geometry. Okay, so the cat had the idea that geometrical construction in geometrical theorems in the assumptions can be articulated in the language of algebra. Which means all the proof that we did in high school trying to prove that the three by sectors meet it one point and that okay, all these can be proven by general shuffling around notation. That was the connection experience. Traumatic. For me, it's

[00:04:46] Host Create clip the connection between the different mathematical disciplines that they all

[00:04:49] Host Create clip noted between two did the

[00:04:51] Host Create clip job languages. So which mathematic discipline is most beautiful is geometry it for you.

[00:04:58] Host Create clip Both are beautiful. They have, ah, almost the same power.

[00:05:02] Host Create clip But there's a visual element of geometry. Being

[00:05:04] Host Create clip individual, it's more transparent. But once you get over to algebra in the linear equations, a straight line, this translation is easily a blow job and the pass attention to a circle. You know, you have the basic theory. You can do it with algebra. So but the transition from one to another was really I thought that the card was the greatest mathematician off all times.

[00:05:36] Host Create clip So you have been at the if you think of engineering and mathematics is a spectrum. Yes, you have been. You have walked casually along this spectrum throughout your throughout your life, you know, a little bit of engineering and then, you know Ah, a bit. That done a little bit of mathematics here and there

[00:05:58] Host Create clip a little bit. I mean, we got a very starry background in mathematics because our teachers will. Geniuses are teachers came from Germany in the 19 thirties, running away from Hitler. They left their careers in Heidelberg and building and came to teach high school in his life. And we were the beneficiary of that experiment. So you and they taught us math the good way.

[00:06:25] Host Create clip What's a good way to teach math chronologically? The people,

[00:06:29] Host Create clip the people behind the exhumes, Yeah, the cousins and their nieces in their faces. And how they jumped from the bathtub when this cream eureka and ran naked in town.

[00:06:46] Host Create clip So you're almost educated. Is a historian off math?

[00:06:49] Host Create clip No. We just got a glimpse off that history together with the field. So every year, exercise and math was connected with the person and the time of the pills. The period

[00:07:03] Host Create clip period. Also, mathematically,

[00:07:05] Host Create clip Smith a medically speaking, yes. Not the politics.

[00:07:09] Host Create clip So and then in ah, in university, you have you have gone on to do engineering.

[00:07:16] Host Create clip Yeah, get the beers and engineering and technology. And then I moved here for graduate work, and I got did engineering. In addition to physics in Rather Gilles Andre. It would combine very nicely with my thesis, which I did in L. C A laboratories in superconductivity

[00:07:40] Host Create clip and then somehow thought to switch to almost computer science software even in the not switch but long to become to get into software engineering a little bit. Yeah, programming, if you can call it that in the seventies. So there's all these disciplines. Yeah. If you were to pick a favor, What? Ah, in terms of engineering and mathematics, which path do you think has more beauty? Which path has more power?

[00:08:08] Host Create clip It's how to choose No, I enjoy doing physics. I even have Ah, voters. Name Learned my name. So I have the investment in immortality.

[00:08:22] Host Create clip Yeah, s So what is a vortex? Vortex is in superconductivity in the super Planetary

[00:08:27] Host Create clip Have Hamelin comment is swirling around one way or the other. You can have a store 10 for computer. That was we worked on in the 1916 l. C. A. And the discovery a few nights phenomena with of Otis is pushing Helen Barone move Vortex pill vote. That's what you can Google it, right? I think nobody but the physicist picked up on my thesis on my Ph. D's and the the it became so popular with the thin film superconductors became important for high temperature superconductors. So they call it a pair of voters without my knowledge it discovered only about 15 years ago.

[00:09:14] Host Create clip You have footprints in all of the science is so let's talk about the universe a little bit. Is the universe at the lowest level deterministic or stochastic? In your amateur philosophy view? Put another way, Does God play dice?

[00:09:29] Host Create clip Well, we know it is stochastic

[00:09:32] Host Create clip okay. Today we think it is the castle.

[00:09:35] Host Create clip Yes, we think because we have the Heisenberg uncertainty principle, we have some. The experiments toe confirm that

[00:09:45] Host Create clip all we have is experiments to confirm it. We don't understand why.

[00:09:50] Host Create clip Why you you're a book about? Yeah, it's a puzzle. It's a puzzle that you have the dice of flipping machine. Oh, God. In the in the result of the flipping propagated with speed faster than the speed of light it we can't explain that. Okay, so But it's only governments microscopic phenomena.

[00:10:21] Host Create clip So you don't think of quantum mechanics is useful for understanding the nature of reality. The Village Inui. So in your thinking, the world might as well be deterministic.

[00:10:35] Host Create clip The world is a humanistic and the star is the new rules. No one firing his concern. It is this the money, Steve, tow first approximation. What

[00:10:47] Host Create clip about free? Will

[00:10:48] Host Create clip Free will is also Ah, nice exercise. Free will is an illusion. Then you re a I people are going to solve.

[00:10:59] Host Create clip So what do you think? Once we saw that, that solution will look like once we put it in

[00:11:04] Host Create clip a look, Nike felt what it looked like. A machine, A machine that act as though it has free will. His communicates with other machines as though they have free will. And you wouldn't be able to tell the difference between a machine that does admission that doesn't have favourites.

[00:11:24] Host Create clip So the illusion it propagates the illusion of free will amongst the other machines

[00:11:28] Host Create clip and faking it is having it. Okay, that's what tooling tests are about. Faking intelligence is intelligence because it's not. Is he so fake? It's very hard to fake, and you can only fake if you have it.

[00:11:47] Host Create clip Yeah, So that's such a beautiful statement. Yeah, Yeah, you could. Yeah. You fake it if you don't have it. Yeah. So ah, let's begin at the beginning with probability. Both philosophically, mathematically. What does it mean to say the probability of something happening is 50%? What is probability?

[00:12:16] Host Create clip It's a degree of uncertainty that an agent has, and about the world

[00:12:22] Host Create clip you're still expressing some knowledge in that statement,

[00:12:24] Host Create clip of course, is a problem with his 90%. It's absolutely different kind of knowledge. And if you just 10%

[00:12:32] Host Create clip am I still not solid knowledge? It's

[00:12:36] Host Create clip the third Knowledge Bay. If you tell me if it in 90% shoo ins, smoking really him give you lung cancers in five years versus 10%. It's a piece off useful knowledge.

[00:12:52] Host Create clip So the statistical view of the universe, Why's it useful? So we're swimming in complete uncertainty, mostly of everything on allows

[00:13:02] Host Create clip you to predict things with certain probability in computing. Those probabilities are very useful that the whole idea off prediction and you need prediction to be able to survive if you can't produce the future than you're just there, crossing the street will be next to me, uh, fearful.

[00:13:26] Host Create clip And so you've done a lot of working causation. And so let's think about Correlation. I started with probability, he started probability. You have invented the Beijing networks, and so you know we'll dance back and forth between these levels of okay, uncertainty. But what is correlation? What is it? So probability is something happening. It's something. But then there's a bunch of things happening, and sometimes they happen together, sometimes not their independent or not. So how do you think about correlation of things

[00:14:03] Host Create clip copulation or killed when two things very together over a very long time is one way of measuring gave? Oh, when you have a bunch off variable that they were very grace of Lee. Um, then because we have a correlation here and usually when we think about copulation, we really think cause I things that cannot be calling unless there is a reason for them to very together. Why should they? Very together. If they don't see each other, why should they? Very together. So

[00:14:35] Host Create clip underlying it somewhere is causation.

[00:14:38] Host Create clip Yes, Hidden in our intuition is a notional causation because we cannot grasp any other logic except is Asian.

[00:14:48] Host Create clip And how does conditional probability differ from causation? So what is conditional probability

[00:14:57] Host Create clip condition? Apple big. How things very When one of them hey, stays the same now, staying the same means that I have chosen to look only of those incidents where the guy has the same value as previous is my choice is an experimental. So things that are not calling her before could become Call it like, for instance, if I have two coins which I uncle related okay, and I choose only those sweepings experiments in which bell rings and bell rings when it is one of them is a tale. Okay, then suddenly I see correlation between the two points because I only looked at the case is with the bell ring. Crazy is my design with my ignorance. Essentially, with my ah audacity to ignore certain incidents, I suddenly create a correlation where it doesn't exist physically,

[00:16:07] Host Create clip right? So that's you just outlined one of the flaws of observing the world and trying to infer something from the meth aboutthe world from looking at the correlation.

[00:16:17] Host Create clip I don't look. It is a floor. The world works like that, which means, But if the flows comes, if we try to impose, um, causal logic on Correlation, it doesn't work too well,

[00:16:34] Host Create clip I mean, but that's exactly what we do. That's what that's has been. The majority of science

[00:16:39] Host Create clip is your Giorgio off naive signs the decision's No, it decisions. Now it. If you condition on 1/3 variable, you can destroy or create correlations among two other valuables. They know it. It's in the data. Nothing surprising. That's why they all dismissed the symptom. But, ah, we know it. You don't know anything about.

[00:17:07] Host Create clip Well, there's There's disciplines like psychology, where all the variables are hard to get to account for. And so oftentimes there's a leap between correlation to causation. You're

[00:17:18] Host Create clip imposing leap who is trying to get authorization from coalition That

[00:17:26] Host Create clip not your not proving causation, but you're sort of Ah, um, discussing it, implying sort of hypothesizing with Allah. Billy,

[00:17:35] Host Create clip which discipline you have in mind? I'll tell you, if they are absolute, who is their outdated or they're about to get outdated? Oh, yes. Tell me which one. You

[00:17:46] Host Create clip all psychology, You know, It's

[00:17:48] Host Create clip okay. What is this CM? You quit?

[00:17:50] Host Create clip No, no. I was thinking of a place Ecologist studying, um, for example, we work with human behavior and semi autonomous vehicles how people behave. And you have to conduct these studies off people driving cars.

[00:18:03] Host Create clip Everything starts with the question. What is the business question?

[00:18:07] Host Create clip What is the research question? The research question. Do people fall asleep when the car is driving itself?

[00:18:19] Host Create clip Do the fall asleep, although they tend to fall asleep more frequently,

[00:18:23] Host Create clip more fickle.

[00:18:23] Host Create clip And the car not

[00:18:25] Host Create clip driving is not driving it. So it's a good question, okay? And so you measure you put people in the car because it's really world. You can't conduct an experiment where you control everything.

[00:18:36] Host Create clip Why can't you could You could turn the Thomas automatic module on and off

[00:18:44] Host Create clip because it's on road public. I mean, there's yes, it's, uh there's aspects to it that's unethical because it's testing on public roads, you can only use vehicle. They have to, uh, the people, the drivers themselves have to make that choice themselves. And so they regulate. That s so you just observe when they drive it, Donald thing when they don't And then

[00:19:10] Host Create clip it may be the turning off when they were very tired.

[00:19:13] Host Create clip Yeah, that's kind of thing. But you don't know those there

[00:19:16] Host Create clip so that you have now uncontrolled, uncontrolled experiment. We recall it observational study, and we form the correlation and detected we have to unfair causal relationship. And whether it was the automatic piece has caused him to fall asleep. Oh, so that is a issue that is the about 100 and 20 years old. Yeah, but I should only go. Energy was old. And the count. Oh, my, maybe. No. Actually, I should say it's 2000 years old because we have this experiment by Daniel. But the Babylonian King that wanted them the exiled, the people from Israel. It was taken in exile to Babylon to sell. The king wanted to serve them King's Fool, which was meat and Daniel as a good. You couldn't eat a non kosher food, so he asked them to a vegetarian food. But the king over Seals says, I'm sorry, but if the king see that your performance falls below from that of other kids, no, he's going to kill me. And then you said, Let's make experience. Let's take four of us from Jerusalem, Okay? US vegetarian food. Let's take the other guys with the King's food in about a week's time will test our performance. And you know the answer. Of course he did the experiment and they will so much better than the others. And the king's nominated them to super position in his. So it was the first experiment so that there was a very simple It's also the same research questions we want to know. A vegetarian food, um, assist, obstruct their your mental ability and, uh, okay, so that the question is very old even Ah, democratise the If I could discover one cause off things Ah, I would rather discuss a one cause and be a king of Persia. Okay, that they task off discovering causes what's in the mind off engine people from many, many years ago. But the mathematics of doing that was only developed in the 19 twenties, So science has left us often. Okay. Science has not provided that with the mathematics to capture the idea off ex coaches. Why? And why does not cause X? Because all the question off physics are symmetrical. Algebraic. The quality sign goes both ways. Okay, let's look a machine learning machine learning today. If you look a deep neural networks, you can think of it as, ah kind of conditional probability conducted fascinators beautiful. So where did you say that? What condition? Poor Brady estimators. None off the machine learning people troubled you. Ah, most people in this is why this today's conversation, I think is interesting, is most people would agree with you. Are there certain aspects that are just effective today? But we're going to hit a wall and there's a lot of ideas. I think you're very right that we're gonna have to return to about causalities. Uh, it would be Let's try to explore it, Z even take a step back. You invented Ah, Beijing networks. That looks awfully a lot like they expressed something like causation, but they don't, Not necessarily. So how do we turn Beijing networks into expressing closet? How do we build causal networks? This a Cosby because of See, how do we start to infer that kind of thing? Who start asking yourself questions? What other factors that would determine the value off ex ex could be blood pressure, death hungry angle. But these air hypotheses that we propose processes everything which had to do this cause authority comes from a theory. Hey, differences on Lee. What kind of how you interrogate the theory you have in your mind. So it still needs the human expert to proposal, right? You need the human expert toe specify the initial model. Initial model could be very qualitative. Just who listens to whom? By whom, Listened. I mean, one variable listened to the other, right? OK, the tide is listening to the moon and not to the rooster crow. And so far, this is our understanding of the world in which we live scientific understanding off reality. We have to stop them because if we don't know howto handle cause and effect relationship. When we do have a model and we certainly do not know how to handle it when we don't have a mother. So that start filth in a slogan is representation first Discovery Second, What if I give you all the information that you need? Can you do anything useful with it? That is the first presentation. How'd you represented? I give you all the knowledge in the world. How do you represent it? When you presented, I ask you, Can you in fair X, y or Z Can you answer certain inquiries? Is the complex is it paranoia? It's called a computer Science exercises. We do. Once you give me a presentation for my knowledge, then you can ask me Now it understand how to representing How do I discover that the second anything There's a lot. We I should echo the statement that mathematics and the current much of the machine learning world has not considered causation that a Cosby just in anything that seems like us. Ah, that seems like, ah, non obvious thing that you think we would have really acknowledge it, but we haven't. So we have to put that on the table. So, uh, knowledge, How hard is it to create a knowledge from wish to work in certain area. It's easy because we have only 45 major variables. In the end, epidemiologists or an economist can put them down. Ah, what minimum wage? Ah, unemployment policy. X Y Z and start collecting data and quantify the parameter that were left unquantifiable with the initial knowledge. That's the oh Tim Well that you find in experimental psychology. Yes, in economics, everywhere. In the health science, it's a routine things, But I should emphasize you should start with research. Question. What do you want? Toe estimate. Once you have that, you ever have a languish off expressing what you want to estimate. You think it's easy? No. So we can talk about two things. I think one is, um, how the science of causation is very useful for at, ah, answering certain questions. And then the other is how do we create intelligence systems that need to reason with causation? So if my research question is, how do I pick up this water bottle from the table? Ah, all the knowledge is required to be able to do that. How do we construct that knowledge base? Does it do we return back to the problem that we didn't solve in the eighties with expert systems do have to solve that problem of automated construction of knowledge. Who you're talking about the task off, eliciting knowledge from an expert task of eliciting knowledge, an expert or this self discovery of more knowledge, uh, more and more knowledge so automating the building of knowledge as much as possible. It's a different game in the causal domain because it's essentially the same thing. You have to start with some knowledge and you're trying to enrich it. But you don't enrich it by asking for more rules in which is by asking for the data. For him to look at the data and quantifying and ask queries that you couldn't answer when you started, it couldn't because the the question is quite complex. And it's not within the, um, capability off ordinary cognition of ordinary person on the expert, even to answer. So what kind of questions do you think we can start to answer even a simple supposed Yeah, and start with the easy one? Okay. What? The effect off drug on recovery. What did the aspirin? That because my headache too be killed. So what did the television program or the good news I received? This is already the difficult question because it's find the cause from effect. These you want to find a fix from cause that's right. So forced you construct a model saying that this is an important research question. This is the one question. Then you do that. I didn't construct the market. I just said it's important. Question one question in the first exercise, his expressive mother matter. What do you want to go like if I tell you what the what will be the effect of taking this drug? He has to say that in mathematics. How do you say that? Yes. Can you write down the question, Not the answer. I want to find the effect off the drug on my headache. Right? Right down, Right. What I did. That's where the do calculus comes. Yes, cooperated when you do, a boy do operator is nice. It's the difference in association and intervention very beautifully sort of constructed. So we could we have a do a pointer. Lou Carter's connected on the door operator itself connect the operation off doing to something that we can see right it was supposed to. The purely observing you're making the choice to change variable. That's what it is it expresses. And then the way that we interpreted the mechanism by which we take your query and we translate into something that we can work with is by giving semantics, saying that you have a model of the world and you cut off all the incoming arrow into X and you're looking now in the modified mutilated model, you ask for the probability why that is interpretation off doing X because by doing things, you've liberate them from all influences that acted upon them area and you subject himto the tyranny off your muscles. So you're you remove all the questions about cause ality by doing them. So you're another one level of questions answered questions about what will happen if you do think if you do, if you drink the coffee if you take the aspirin. Right. So how do we get the once? How do we get the doing data? No, The question is, um, if you cannot one experiments right, then we have to rely only observational study. First, we could decide to interrupt. We could run an experiment. Well, we do something where we drink the coffee and Oh, and this the do operator allows you to sort of be systematic about expressing who Imagine how the experiment will look like even though we cannot physically and technologically conducted. Given example, What is the effect of blood pressure on mortality? I cannot go down into your vein and change your blood pressure. But that can ask the question, which means I can't even have a model of your body. I can imagine the effect off your how the am blood pressure change will affect your mortality. How I go into the model and I conduct is surgery about the blood pressure. Even though physically I can do I cannot do it. Let me ask the quantum mechanics question. Does the doing change the observation? Meaning the surgery of changing the blood pressure is, um I mean, no, that surgery is a hey called, um, very delicate. It is very infinitely dark. It decisive interrogate, which really do means do X means I'm gonna touch all the X on the X directly into X. So that means that I change only things, which depends on X by virtue of exchanging, but I don't depend things which are not dependent on next. Like I wouldn't change your sex or your age. I just change your blood pressure. Okay, so in the case of blood pressure, it may be difficult or impossible to construct such an experiment. No, both physically years, but hypothetically know if we have a model, that is what the model is for. So you conduct a surgeries on a model, you take it apart, put it back. That's idea of a model it a view of thinking, counter factory imagining and that idea off creativity. So by constructing them all you constructed for if the higher the blood pressure leads to mortality, which increases or decreases by construct the model, I can still cannot answer it. I have to see if I have enough information in the model. It would allow me to find out the effects off intervention from a non interventional study, the moves of a hands off study. So what's needed to be near tohave assumptions, but who affect whom? If there is the Graf had a certain property, the answer is yes. You can get it from off provisional study. If the graph is too messy, bushy bushy. The answer is no. You cannot. Then you need to find either different, kind off observation that you haven't considered. 01 experiment so basically does that puts a lot of pressure on you, too. Encode wisdom into that graph, but you don't have to include more than what you know, God forbid if you put the light. Economists are doing this because identifying assumption, they put assumptions even they don't prevail in the world. They put the something so they can identify things. But the problem is, yes, beautifully put. But the problem is, you don't know where you don't know. So you know what? You don't know, because if you don't know, you say it's possible it's possible that ex affect the traffic tomorrow. I know it's possible to put down a era which says this about every a row in the graph says it's possible. So there's not a significant cost toe adding arrows that the more air Oh, you add about, the less likely you are to identify things from purely observational other. So if the whole world is bushy and it was, everybody affect everybody else. The answer is you can answer it ahead of time. I cannot answer my query from observational that I have to go to experiment. So you talk about machine learning is essentially learning by association, reasoning by association. And this do calculus is allowing for intervention. Like that word, uh, with an action. So you also talk about counterfactual sze and trying to sort of understand the difference in counterfactual is an intervention. Ah, what's the first of what is counterfactual is and why are they useful? Why are they especially useful as opposed to just reasoning? What what affect actions have conflict if actually contains what we know, Willie Kool Explanations can give an example of that acting one way. Affect something. I didn't explain anything it, But if I If I asked you, Was it the aspirin? Cuma headache Amassing for explanation would kill my headache. And putting a finger on aspirin provided financial. It was aspirin. It was responsible for your headache going away. If if you didn't think the husband, you would still have a headache the bys by saying if I didn't take aspirin, I would have a headache year there by saying that aspirin is the thing that remove the headache? He is. But you have to have another import of information. I took the aspirin, and my headache is gun. It's very important information. No reasoning, Backboard. And I said, Well, did ask mine? Yeah, by considering what would have happened if everything else is the same. But I didn't take aspirin. That's right. So you know that things took place? No. Joe Kidd Schmo and sure would would be alive. Head John, not use his gun. S o that is the counter factor. Is it considered confliction here the conflict here or clash between observed fact. But he did shoot. Okay. And the hypothetical predicated which says heady No, Chad, you have a clash. Logical clash. They cannot exist together. Let's count affection. And that is the source of our explanation off the youth responsibility, regret and free will. Yes, it certainly seems. Ah, that's the highest level of reasoning, right? Yeah. And penises do it all the time. Who does it all time? Physicists, physicists in every equation of physics, let's say you have a hook slaw and you put one kilogram on the spring in the spring is one meter and you say had this way being took telegram. The spring would have been twice as long. It's no problem in four physicists to save it. Accepted. Mathematics is only is in the form of equation. Okay. Equating the wait proportionality constant in the Lincoln of the string. So you don't have the a symmetry in the equation of physics. Although every physicist things can't off actually ask the high school kids had the weight being three kilograms. What will be the length off the spring? I can answer it immediately because they do the counterfactual processing in their mind. And then they put it into equation algebra equation, and they solve it. Okay, but the robot cannot do that. How do you make a robot learn these relationships, Evelyn? The values would learn. Tell him, Can you do it before you go learning? Yeah. You have to ask yourself. Suppose I did hold information? Okay. Can can the robot perform the task and I asked him to perform. Can he reason? It's a no. It wasn't the aspirin. It was the good news you received on the phone, right? Because well, unless the robot had a model Ah, a causal model of the world. That's I have to linger on this. That's now we have to link, you know, we have to see How do we How do we do it? How do we build? Yes. How do we build a cause? A model without a team of human expert on Facebook running around. Why don't you go to a learning right away? You're too much involved with looming because I like babies. Babies learn faster than you ever got. Hardly. D'oh! Good. That's another question. How did the babies come out with counterfactual model of the world? And babies do that. They know how to play with in the crib. They know which balls hits another one, and they literally by, um, play for manipulation off the world. Yes, simple World involved only toys and bulls and times meth is ever if you see a way the complex world we take for grand. Yeah, how come and get do it by playful manipulation, plus Parent's guidance Pew wisdom. And he'll say they greet each other. Can they? You shouldn't have taken my toy, all right, And they these multiple sources of information they're able to integrate. So the challenge is about how to integrate, how to form. He's causal relationship from different sources of data could. So how how how much is that information is it to play how much causal information is required to be able to play in the crib with different objects? I don't know. I haven't experimented with the crib. Okay, Not a crib. Picking up very manipulating physical objects on this very opening the pages of a book. All the task, the physical manipulation tasks give a sense because my senses, the world is extremely complicated, Complicated. I agree. And I don't know how to organize it because I've been spoiled by ease the problems such as cancer and death. Okay, 50 50 First, we have to start the major easy. The easing of scented You have only 20 variables, and they're just very words are not mechanics. It's easy. You put him on the graph and they they speak to you. And you you're providing a methodology for for helping them speak. Yeah, I'm working only in the abstract. Have knowledge in knowledge, out data in between. Now, can we take a leap to trying to learn in this Very when the saw 20 variables but 20 million variables trying to learn causation in this world, not learn but somehow construct models. I mean, it seems like you would only have be ableto learn, because constructing it manually would be too difficult. You have ideas off E. I think it's the middle of combining simple models for many, many sources from many, many disciplines in Millimeter Falls, metaphors are the basics of human intelligence basis. So how do you think of about a metaphor in terms of its use in human intelligence? Meta fools is the next boat system, and next time it's mapping problem with which you are not familiar. Two. A problem with which you are familiar. Like I give a good example. The Greek believed that the sky is no picture. It's not really as in free space, it's in a picture, and the stars are holes poked in the shells. Through is you see the eternal light. That was a metaphor way because they understand how you poke holes in shares they're not. They're not familiar with infinite space, and, uh, and we are working on ah, share off a total. And if you get too close to the head, you're gonna fall down too hard this wherever. Yeah, and that's a metaphor. It's not true. But these kind of metaphor enable Czarist Otterness to measure the radios off the Earth because he said, Come on, If they were working on a turtle shell, then array off light coming through this anger be different. This place will be different. Angle that coming to this place. I know the distance. I'll measure two angles on and then I have them radio off the shell of the turtle. And he did. And he found his measurement very close to the measurements we have today through the ER but 10 6707 100 kilometer there. That's something that would not occur to Babylonian astronomer. Even though the bologna experiments were the machine learning people of the time, they fit curves and they could predict the eclipse off the moon much more accurately than the Greek because they feel cared. Okay. Ah, that's a different matter for something. You're familiar with a game? A total shit. What does it mean? If you're familiar, familiar means that answers to certain questions are explicit. You don't have to derive them, and they were made explicit because somewhere in the past, you've constructed a model off that Oh, yeah. You're familiar with the Chinese familiar with billiard balls? Yes. So the child could predict that if you let loose off one ball, that the one will bounce off. These are you. You obtain that by, um, familiarity. Familiarity is answering questions and you stole the answer explicitly. You don't have to divide them. So this is ideal for beautiful All our life, all our intelligence is building on Mr Falls, mapping from the unfamiliar to the familiar. But the marriage between the two is a tough thing, which I which we haven't be ableto algorithmic ties. So you think that process of of using metaphor to leap from one place to another? We call it reasoning. Is it a kind of reasoning? It is reasoning by metaphor, metaphor for do you? I think of that as learning. So learning is a popular terminology today in the narrow sense it is, it is. It is definitely if you may not dog erect one of the most important, very taking something which theoretically is driveable and store it inaccessible. Former and given example kiss Okay. Finding the winning, winning, starting, moving just who is hard. But the ah If there is an answer, either there is a winning move for white or the reason all these oh draw. So it is the The answer to that is available for the rule of the Games, but we don't know the answer. So what are the chest muster? Have it we don't have, he has told explicitly an evaluation off Southern complex pattern off the board. We don't have it. Ordinary people like me, I don't know. But you know, I'm not the chess master. So for me I have to derive yes, things that for him is explicit. He has seen it before where you've seen the pattern before. Well, similar bottom is immutable, and he generalize and said, Don't move with the dangerous move. It's just that not in the game of chess but in the game off billiard balls. We humans are able to initially derive very effectively and then reason by metaphor very effectively and make it look so easy. And it makes one wonder how hard is it to build it in a machine. So in your senses, how far away are we to be able to construct. I don't know. I'm not a futurist. Can, I can tell you is if we're making tremendous progress in the causal reasoning domain, something that I even they're to call it evolution the Cultural Revolution. Because, ah, what we have achieved in the past three decades is something that, ah, dwarf everything. It was derived in the entire history. So there's an excitement about current machine learning methodologies, and there's really important good work you're doing in causal inference. Where do the what is the future? Word of these worlds collide, And what does that look like? First, they're gonna work without collisions. It is gonna work in harmony. Harmony's that the the human is going toe to toe jump start the exercise by providing qualitative, known committing mothers of how the universe works, universe that we ever how the in the reality the domain of discourse works. The machine is gonna take over from that point of view and derive whatever the calculus says can be derived. Neighborly, quantitative answer to our questions. These are complex question. I'll give you some example of complex question that the wood Bugel your mind if you think about it. You think result off studies in the verse population under the reverse condition. And you mean for the cause. Effect off a new population which doesn't even resemble any of the one studied. And you do that buying do calculus in Did it by generalizing from one study to another. See what's what's comin with Beto. What is different. Let's ignore the differences and pull out the commonality and you do it over maybe 100 hospitals around the world. From that you can get really ah ah, mileage from big data. It's not only do you have many samples, you have many sources off that, so that's a really powerful thing. And I think for especially for medical applications, cure cancer, right? That's how from data you can cure cancer. So you were talking about causation, which is the temporal temporal relationships between things. Not only Temple is was structural and temporal, then put her enough temporal presidents by itself cannot replace causation is temporal precedence. The era of time in physics is important necessary. It's important official yet is it? Yes, I I never seen because we get back. But if we call if we use the word cause. But there's relationships that air timeless. I suppose that's still forward in a row of time. But the are their relationships. Logical relationships that fit into the structure whole do. Calculus is logical. Relationship that doesn't require temporal hasn't just a condition that it's you're not traveling back in time. Yeah, so it's really a generalization of a powerful generalization. Ah, off Why Boolean logic have Boolean logic. Yes, that is simply put, it allows us to ah, you know, reason. Ah, reason about the order of events, the source the about between not deriving the we are given cause effect relationship. Okay, They ought to be obeying the time president's relationship we're giving it. And now that we ask questions about other cause of relationship that could be derived from the initial ones. But you're not given to us explicitly there like you. The case off the firing squad they give you in the first chapter and I asked what if rifleman eh declined to shoot with the prisoners still be dead, uh, to decline to shoot. It means that this obey order and the the rule of the Games were that he's ah ah obedient and Ma marksman. OK, It's how you start. It's a issue. Not, But now you ask question about breaking the rules. What if he decided not to pull the trigger? He just became a pacifist. Oh, and you can you and I can answer that. The other rifleman would have killed him. Okay, I want a machine to do that. It's so hard to ask, wishing to do that This is a simple plaza. But I also have a calculus for that. Yes, yeah, but the curiosity and the natural curiosity for me is that yes, you absolutely correct and important. And ah, it's hard to believe that we haven't done this seriously extensively already. A long time ago. So this this is really important work. But I also want to know. You know, this maybe you can philosophizing about how hard is it to learn? Okay, lets us women learning. We want to learn it. Okay, What I learnt. So what do we do? It? We put a learning machine that watches execution trials in many countries and many locations. All the machine can there is to see shot or not shot dead not dead. A court issued an order or didn't. Okay, that fact for the fact you don't know who listens to whom. He don't know that they condemned person. Listen to the bullets and the bullets are listening to the happen. Okay, All we hear is one command, two shots dead. Okay, triple off variable. Yes. No. Yes. No room that you can learn who lists to home and you can answer the question. No, definitively. No, but don't you think you can start proposing ideas for humans to review? You want machine to die doing the robot. So Robert is watching trials like the hit 200 trials. And then he has to answer the question. What if riflemen a refrain from shooting? So I don't do that? That's exactly my point. Looking at the facts don't give you the strings behind the fact. Absolutely. But do you think of machine learning as currently defined as on Lee? Something that looks at the facts and try right now the only look at the fact Is there a way to modify? Yeah, in your sense, Playful manipulation play for motivation. What do want interventionist kind of thing? Yes, intervention. But it could be a trundle. For instance, the rifleman is sick of the day or he just vomit. So whatever Rethinking Observe this unexpected event which introduced noise. The noise still have to be, Ah, random Toby ableto and related to randomize experiment. And then you have a, um, observational studies from which to infer the strings behind the fact it's doob to a certain extent. But now that we expert in what you can do want to have a model, we can reason back and say, What do you kind of data? You need to build a model. I got it. So I know you're not a futurist, but are you excited? Have you moved back in your life long for the idea of creating a human level? Intelligence is Yeah, I'm driven by that all my life until you just buy one thing. But I go slowly. I go from what I know to the next step incrementally so without imagining what the end goal looks like. Do you imagine what? Hey, then go is gonna be a machine that can answer sophisticated questions, counterfactual of regret, compassion, um, responsibility and free will. So what is a good test? Is a touring tests. It's a reasonable free wheel doesn't exist yet. Ah, there's no would you test feel on? That's so far we know only one thing. Me. If Roberts can communicate with reward and punishment among themselves um, hitting each other on the wrist and say you shouldn't have done that. Okay? I am playing better soccer because it can do that. What do you mean? Because they can do that because it can communicate among this because of the communication that can do because of the communicate like us reward and punishment. Yes, you didn't possible the right the right time. And so far there are four. You're gonna sit on the bench for the next two If they start communicating like that, The question is, will they play better? Soccer is opposed to What is it? Put what they do now without this ability to reason about ah, reward and punishment, responsibility and factual. I can only think about communication, communications and in a non insane natural language. But just communication, just communication. And that's important to have a quick and effective means off communicating knowledge. If the coach tells you should've passed a ball pink, he conveys so much knowledge to you as opposed to work. Go down and change yourself. That's the alternative. But the court doesn't know your software. So how can it coach tell you you should have passed the ball but that our language is very effective. If it is possible. You know you're so awful. You tweak the right more jewels. Okay, Next time you don't do it. Now that's for playing soccer. The rules well defined. No, no, not the will defy when you should pass the ball is not what the fuck? No, it's a Any software that is noisy. Yes, you have to do the pressure. It's art. But in terms of aligning values between computers and humans, do you think this cause and effect? Ah, type of thinking is important to align the values, values, morals, ethics under which the machines make decisions is is the cause effect where the two come together gonna figure is necessary Component toe build. Ah, ethical machine because the machine has empathize toe understand what's good for you, Toby? A model of your view of the recipient. We should be very much What is compassion the Imagine if you suffer pain as much as me as much as I don't have already a model of myself, right? So it's very easy for me to map you to mine. I don't have to rebuild the matter if matches there. You like me? OK, Therefore I will not hit you and the machine is to imagine Has to try to fake to be human. Essentially. So you can imagine that you're the year like me, right? And moreover, who's me? That's the feather. That's consciousness. They have a model of yourself. Where'd you get this? Mother? You look at yourself as if you are a part off the environment. If you build a model of yourself veritas the environment, then you can say I need to have a month off myself. I have abilities, I have desires. And so for Okay, I have a blueprint of myself, not a full detail, because I cannot get the halting problem. But I have a blueprint. So that level of a blueprint I can modify things. I can look at myself in the mirror and say room. If I change this more trick, this smile, I'm gonna perform differently. That is what we mean by free will and consciousness. What do you think its consciousness? Is it simply self awareness? So including yourself into the model of the world. This way. That way some people tell me, No, this is only part of consciousness. And then they start Anybody really mean and I lose? Yeah. For me, consciousness is having ah, blueprint of yourself. Do you have concerns about the future of a I All the different trajectories of all of our research? Yes. Um, where's your hope? Where the movement has, where your concerns I'm consumed because I know we're building a new species. It has the capability off exceeding our exceeding us ah, after the AL capabilities and can breed itself and take over the world. Absolutely. It's a no species. It is uncontrolled. We don't know the degree to which we control it. We don't even understand what it means to be able to control this new species. So I'm concerned I don't have anything to add to them because it's such a great area that unknown. It never happened in history. Yeah, quick. The only the only time it happened in history was evolution with human being right. It was a very successful would it simply says It was a great success for us. It was, But a few people along the way are Few creatures along the way would not agree. So Ah, so it is just because of such a gray area. There's nothing else to say. We have a sample of one sample of one. It's us. But we don't know how people will look at you and say, Yeah, but we were looking to you to help us make sure that the sample to works out. OK, we have more than a sample of Mark. We have theory theories, and that's a good that's we don't need to be statisticians. So something off mine doesn't mean you Poverty of knowledge it's not sample off one plus ceiling conjectural theory of what could happen. Yeah, that we do have. But I really feel helpless in contributing to this argument because I know so little and the and my imagination is limited. And ah, you know how much I don't know in the I. But I'm concerned you're born and raised in Israel Moon reasoning and ah later served in, um, Israel Military Defense forces in the in the Israel defense. Full. Yeah, What did you learn from that experience? Well, this is business. There's a kibbutz in there as well. Yes, because I wasn't enough, which is a I am combination of agricultural work and military service we will support. I was with the idealist. I wanted to the a member of the Khumbu throughout my life and to leave a community life. And ah. So I prepared myself for that, uh, slowly, slowly. I want the greater challenge so that Zeus Ah, far world away both what I learned from that it was a miracle. It was a miracle that I served in the 19 fifties. I don't know how we survived. The country was under austerity. It tripled its population from 600,000 to a 1,000,000 0.8. When I finish college, no one went hungry. Aw, 30th when you wanted to buy to make an ambulance In the rest one, you had to bring your own egg and, uh, the imprisoned people from bringing the food from the far from a farming here from the village is to the city. But no one was hungry, and I always add to it high education did know suffer any budget cuts. They still invested in me in my wife, in our generation to get the best education that they could. Okay, so I'm really grateful was opportunity, and I'm trying to pay back now. Okay. It's a miracle that we survived the war off. 1948. They were so close to a second genocide. It was all implant. But we survived by Miracle And then the second miracle that not many people talk about the next phase how no one went angry in the country, managed to three police population. You know, it makes the principal imagine United States going from, what, 350 million, too? Yeah, I'm blue. There's a very tense part of the world. It's a complicated part of the world. Israel, if in all around the religion is, is at the core of that complexity. One of the components is a strong motivating course. Who many, many people in the Middle East. In your view, looking back is religion good for society? It's a good question for robotic. You know, it should have a cause of that quick, probably with religious belief. Suppose we find out who agree that religion is good to you to keep you in line should be given about the A metaphor for God, the metro for the world. But we'll get it without us also. Why was more a robot with reason by metaphor and what is the was primitive metaphor? A child grows with mother smile, follow teaching, father image and mother. That's God. So I wanted Oh, no, no, no. But with. But it's simply a summary of the robot is gonna have a mother and father. It may only have a programmer. We doesn't supply warmth and discipline Discipline. It does zero but will have this model off the trainer, okay. And everything that happens in the world cosmology and so is going to be mapped into the program. If God man the thing that represents the origin off everything for that writes the most primitive relationship. So there's gonna arrive there by metaphor. And so the question is, if overall that metaphor has served us well as humans, I really don't know. I think it did. But long as you keep in mind, it's only metaphor. So if you think we can can we talk about your son? Yes, Yes. Can you tell his story story? Well, Daniel is so is no news was objective in the Pakistan by Al Qaeda driven six and, um undergoes various pretences. I don't even pay attention to with the pretence Will. Originally they wanted toa toa Every the United States deliver some promised airplanes that it was all made up A list of demands. We're bogus. I don't know really. But eventually he was ah executed in front of a camera. At the core of that is hate and intolerance. They're calling it absolutely Yes. We don't really appreciate the depth off the hate. It's which which billions off people's are educated. We don't understand it. I just listened recently. So what they teach you in Mogadishu When? When the wood will step in the tap. We know exactly who did it. The Jews, the Jews. We didn't know how, but we knew who did it. We don't appreciate what it means to us. The depth is unbelievable. Think all of us capable of evil and the education. The indoctrination is really were created to keep it off, even if you're indoctrinated sufficiently long and in depths capable off Isis capable of Nazism. Yes, we are. But the question is with a we After we have gun through some Western education and relented. Everything is really relative. It is no absolute God is only a belief in God. Whether we're capable now of being transformed under certain circumstances to become, ah, brutal. Yeah, that's why I'm worried about it. Because some people say yes, given the right circumstances. Given ah, economic Lou Bad, economical crisis, you are capable of doing it too. That's what worries me. I want to believe it. I'm not capable. Seven years after Daniel's death, he wrote an article in The Wall Street Journal titled Daniel Pearl. The Normalization of Evil. What was your massive A message back then? And how did it change today? Over over the years I lost. What was the message? The message was that Ah, we are not treating terrorism. Is it taboo we're treating? It is a bargaining device that is accepted. People have grievance and the going in the bomb Merced ones. It's normal. No, you even not not surprised when I tell you that 20 years ago. So what? For grievance you go on blow restaurant. Today's becoming normalized, the finalization off evil and that we have created that tow ourselves by normalizing by a lie by making it part off political life. It's a political, uh, debate. Every terrorists yesterday becomes a freedom fighter today into more become terrorist idiot switchable. And so we should call our evil when there's evil. If we don't want to be proud of it, become if, yeah, if we want to separate good from evil. That's one of the first thing that what was the Garden of Eden? Remember the first thing I gotta tell you, it was a You want some knowledge? Here's the three off good and evil. So this evil touched your life personally. Does your heart have anger? Sadness Or is it hope? Oh, I see Cem. Beautiful people coming from Pakistan. I see beautiful people everywhere. But I see horrible propagation off evil in this country too. It should help. Populistic slogan can catch the mind of the best intellectuals. Today is Father's Day. I didn't know that. Yeah, What's Ah, what's what's ah fond memory you have of Daniel? Oh, many goodly with ah humans. Hey, was my mental? What do you He had, um sense of balance that I didn't have Yeah, He said the beauty in every person he was Not as emotional is I'm more look isn't things in perspective. You really like every person he really grew up with. The idea that four nil is, ah, reason for curiosity notes for fear. At one time we went in Berkeley and, um, homeless came out in some dark alley and said, Hey, man, can you spare a dime? I with written back, you know, two feet back. And then I just hugged him and say, Here's a dime and juggle safe maybe wants, um, um, money to take a bus or whatever, we're dedicated. Not for me. Do you have advice for young minds today dreaming about creating, as you have dreamt, creating intelligence systems? What is the best way to arrive at new breakthrough ideas and carry them through the fire of criticism and and past conventional ideas? Ask your question freely, your questions on level dumb and so the only. And don't take no for an answer. If they're really them, you'll find up quickly by trying an arrow to see that they're not leaving any place but follow them and try to understand things your way. That is my It's all right. I don't live is gonna help anywhere now There's brilliantly by the lot off. It's the initial in science in academia. It is slowing down science. Yeah, those two words your way. That's a powerful thing. It's against inertia, potentially against the phone against your professor. Okay, I want the book off. Why? In other to democratize common sense in little too in still rebellious spirit in students so they wouldn't wait for until the professor get things right. Asi wrote the manifesto of the rebellion against the professor and the ensuing befriends. Oh, yes. So, looking back at your life of research, what ideas do you hope ripple through the next many decades? What would do you hope your legacy will be? I already have. Ah, Tombstone. Hey, boy, element the fundamental law off contra factors. That's what it's Ah, simple equation. What can't affection in terms off model surgery? That's it. Because everything for us from that if you get that, all of us, I can die in peace and my student can derive all my knowledge. My mother medical means the rest follows. Yeah, Dad, Thank you so much for talking. Really appreciate it thank you for being so attentive and instigating way. Did it really? The coffee helped. Thanks for listening to this conversation with Judea Pearl and thank you to our presenting sponsor. Cash up. Download it. Use code, Lex podcasts. You'll get $10.10 dollars will go first stem education on profit and inspires hundreds of thousands of young minds toe learn and to dream of engineering our future. If you enjoy this podcast, subscribe on YouTube. Get five stars, an apple podcast, support on page Iran or simply connect with me on Twitter. And now let me leave you some words of wisdom from Judea. Pearl, You cannot answer a question that you cannot ask and you cannot ask a question. You have no wars for Thank you for listening and hope to see you next time.