Nassim Taleb: How Things Gain from Disorder [Entire Talk]

Thanks a lot. So let me give you my sort of biographical story and my obsession with what I call fat tails to explain fat tails. What you have here is the performance of an option portfolio. The variation in price of an option portfolio, daily variation over 57 years. Okay? So out of the money options, options that are remote, this is about 12,000 days, something like that, 14,000 days. And as you can see, there's a big spike for that portfolio. That day represents, depends on what you've been doing, between 80 and 99% of your total lifetime variation. If you're an option trader, one day in your life should represent 98% of the variation. If you're involved in remote payoffs, this, in other words, is what I call fat tails. And this is, of course, if you don't get that day right, you're not going to get anything right. And. It's not predictable, okay? They tell you it's predictable. If you go to the finance building, they think it's predictable. If you go to economics building, they're convinced it's predictable. It is not predictable. So that's the problem. They are events that have a huge impact and we can't tell ahead of time when they're going to come. And we have no idea about the structure. And people fool themselves thinking they have an idea about the structure. So you have to understand what domains are affected by these large deviations. And also, I don't know if someone here is from Brooklyn, but in Brooklyn they say that you should make lemonade every time someone gives you a lemon. So you got to make lemonade out of it. So this is the idea that we have. There's a huge amount of, of these events, the rare event representing a large share of the pie. Now you live here at the epicenter of rare events. Google represent this spike in its own business. Microsoft on Microsoft was Microsoft. It's still Microsoft, but, you know, even more Microsoft than it is today. There was 50,000 computer companies and, and it had more than half the sales in that segment. If you take the stock market over the past 100 years, a lot of company joined, a lot of company left. There are about 12,000 surviving companies that are listed. Of these 12,000 companies, how many companies represent half the capitalization? What do you think? Sorry, when I was trading, some years it was 50 to 100, some years it was 300 to 400. There's no structure. Another example, for example, is a book business. I don't know if you know about the book business. There are a million manuscripts in the English language on the market today by people who call themselves writers. But typically in the US they work for Starbucks. In Europe they do other things. They work for the government. So they call themselves writers. They're novels. Of the million novels, literary novels, about 20,000 will be published. Of these, half the sales will come from how many books? Sorry, 20%, maybe 20%? No, between 5 and 15. Some years less than 5. It's not 20%, really. We are in what I call extremistat. So let me give you an idea of what is extremistan. If I can figure out how to work this computer. If you imagine that you have a convention here and you bring in a random selection of people from the planet, a thousand people, bring them to Stanford, nice weather, you know, this is a great way to convince people to come and put them on a scale. You weigh them, you add to that sample the heaviest person you can find on the planet, who can still be called a person. How much of the total will that person represent? Sorry? OK, so three times the average, as we say in finance, 30 basis points, 0.3%. And if you move from 1,000 to 10,000, how much will that person represent? Nothing. Three basis points. Okay, so the maximum you can get will affect your average very little. This is commonly known as the law of what? Large numbers. About everything you do that's statistically based is based on that law that as your sample becomes larger, no deviation is going to mess up things for you. Agree. You eat? You know, I eat. I don't know. I consume 800,000 calories a year. Not a single day is going to make me, you know, double my weight. All right, thank God. But unfortunately, not a single day will make me lose half my weight unless, you know, I have surgery. All right? So the problem is that. But in finance, can you lose half your fortune in one day? Of course, when I became a trader, there's a. I walked in, there's something called new member. A badge means rookie. So you get abused by old traders who are bored, want to give someone life advice for life, so they grab you. So one guy grabbed me. Come over here, kiddo. He said, okay. I showed up. He said, that was before this event, incidentally. He said, listen, you see the guy over there? I said, yes. He said, his name is Ed. Yes. He said, at the time, it was a lot of money. He made 7 million in seven years. I said, yes. He lost them all in seven seconds. Okay. I said, kiddo, now you can go. All right? That was his advice, that you can you know that you're an extremist stand. The reasoning is different. So everything's moved by extremes in economic life. And visibly you're here for that, to understand that point. And we're going to take it to its conclusion. An extremistan. Okay. Things are unpredictable in mediocre Stan. You can predict things. So that's point number one. Unpredictability is a problem, of course, in the real world, unlike the textbooks. I guess you've seen this before. You recognize this item. This is commonly known as what? A wheel. Very good. When was it discovered? Sorry, Long time ago. How long ago? At least 6,000 years ago. All right, Babylonian technology. All right, so this is technology. So let's see. Moving from this technology to this technology. You guys are too young to remember the days when it had to slept to carry our luggage because they didn't have wheel. How long did it take? 6,000 years to this discovery? Okay, so you realize that there is unpredictability. Not only that, but now we're going to look at something else. There are domains that are entirely dominated by the unpredictable. Medical discovery, for example, entirely dominated by the rare event. They don't know it. And we're going to talk now about luck, with a caveat. It's not really luck. It's a function of luck. How to exploit luck by which you get more upside than downside of luck. So the idea is, you know what this is? Windmill. The two approaches in life. The first one is that you hear that there are going to be winds, and the conservative attitude is to build what? A wall to protect oneself from the wind. A better attitude is to build windmills, if you hear the word. So the idea is to milk uncertainty and try to exploit disorder. But before getting there, we're going to have a little excursion into what I call fragility. There's something interesting that I discovered that took me about 21 years to figure out. 21 years. I was an option trader for 21 years and then realized something. I knew unconsciously, but only became conscious of it one day as I was looking at this patently ugly teacup. All right, we have a definition of fragility, very simply connected to what we saw about options. And let me propose the following. Is fragile, what doesn't like volatility? If this teacup is on a table and there's an earthquake, will it benefit from it? No, you see, Very simple. It is. It wants fragile because it wants calm. And predictability. It's harmed by disorder. You guys have earthquakes in California? It doesn't want an earthquake. It doesn't want to be in California. It'd rather be in New York. You see, very simply. And now from this concept of fragility, we're going to get into innovation and optionality. So this, the way I see it, is short optionality. And let's look at what short optionality is. This is a payoff of a financial structure that is very vulnerable to phone calls at night, when in the middle of the night, the chairman of the company gets a phone call saying, oh, there's an event. So this payoff, this package, you know, you make or lose some money, but typically the losses, all right, are very large and the profits are very small. Exactly like the coffee cup, except the coffee cup has no variation. Except for the loss, the coffee cup will never improve. Let's see if you can connect this. This package is robust. And this package has an opposite payoff where all the gains are large. The kind of thing that only people in California seem to understand. People in New York, they don't get it because people in New York have this payoff. And let me explain. The banks are in a business of hiding risks. So they make, make, make, make, make, make small. They have no volatility, you know, most of the time. And once every 10 or so years, guess what? They lose everything they made, and then they come to the taxpayer for support. They have the setup for people to give them monetary policy to help them and all that stuff. And meanwhile, they keep their bonuses because it's a very efficient strategy where you make everybody think you're smart. And then when you lose, guess what? It was a bad environment. It was only a quarter when you lost, but visibly, you've lost more than you ever made. Banks lost in 2008 more than history of banking. And of course, they paid themselves a lot of money. They didn't return it to us. So this is a payoff that's very similar to a short, an option. You sell an option, you have very little return, you earn some money. But then something happens. You have unlimited losses or very large losses. That payoff resemble the threshold. The opposite of that payoff would be something that's long volatility. So when I went to try to explain this, people weren't understanding that the opposite of fragile isn't robust. It's not solid. It is not the opposite of fragile. The opposite of this. It's not the straight line. It's not a payoff where you end up unharmed. You agree? If I'm sending a package to, say, Mongolia Interior. And the package is fragile, what do I write on it? Fragile. If the package is robust, what do you write on it? Nothing. So if the package wants volatility, wants to be harmed, wants something, wants stressors, wants disorder, wants earthquakes, it's not the same as robust. It would be something on which you'd write, please mishandle. I am antifragile. It's the opposite of fragile. So the opposite of fragile is not something that is robust or something solid. It's something that wants disorder. And guess what? There's something called a cluster. And if you like one, you like them all. Uncertainty, variability, incomplete knowledge, trance, chaos. If you benefit from all of these, you are classified as antifragile. And it's very similar to someone who owns an option. So we're going to talk about how you guys can milk randomness, because that's what you do in California, right? And we're going to be a little more technical and a little more rigorous in the way we analyze an option payoff. So there are things. Now, let's view the world in three categories. There's a fragile, like this computer. You pour water on it, it's gone, right? You understand? And now we have to give a lecture, maneuvering with a mouse and maneuvering so it doesn't blow up and hoping doesn't blow up before the end of this lecture, right? That's a fragile, okay? That's the definition. There's a robust. You hammer it, nothing happens but doesn't benefit from it. And there's the anti fragile that likes disorder and benefits from disorder and benefits from error. So happen that now we can identify the fragile. There's a method to figure out what is fragile, and there's a method to figure out what is anti fragile. How? Very simple. Something related to convexity. Let me explain what I mean by convexity or concavity. Okay? If I jump, I am fragile. My body is fragile. Okay? If I jump 10 meters, what would happen to me? You guys are engineers and smart people. You would know what would happen to me. Sorry, Squish. I die. I mean, say it, you die. Your hospital here won't be any good, right? You die when you fall 10 meters. But if I fall 10 times 1 meter, what would happen to me? Nothing. It means that every additional meter harms me more than the previous one. No? Aha. Everything fragile, we saw it in time series space, but if you look at it as a response, everything fragile has to have accelerating harm at some point. And we can actually show that everything that has accelerating Harm doesn't like volatility. And the opposite of it is something that has accelerating benefits. The nonlinear is harmed more. You see, this accelerating harm is harm a lot more than the linear. And visibly the linear for small variation show more response than the non linear. This is easier to understand if you look at it this way. You saw that the harm I get from hitting the floor, okay, it has to be nonlinear for me to be harmed at 10 meters, but not to be harmed at 10 times 1 meter. You agree? Okay, it has to be that way because walking, you know, from on campus would kill you otherwise, because being harmed by that large. But you get the. If the thing were linear, you'd be killed. So harm for anything that has not been broken has to be. In a nonlinear. The coffee cup sees thousands of hits, very small intensity, like very small pounds per inch hits. It doesn't really care, but one big hit and it breaks. So things that are fragile have to be nonlinear. And this is what's central about this idea. You have to be accelerating on harm. Another example is look at the world around us, okay? If we were linear to harm, we'd blown up long time ago. The world has 3 to 4 million earthquakes every year of low intensity on a Geiger scale. You see, it would be like a Fukushima every few minutes if it were, if the harm were linear, but it's not. So this is quite central because now we can figure out whether it's fragile from nonlinear and look at its opposite. Simply we have the following. If you have more gain, say if the market goes up 10%, you make more money than the market went down 10%, then you lose. If the market went down 10%, you're convex, you see, and otherwise you're concave. The best way to figure it out is with this. This is convex, this is concave. If you want to remember, this is convex and this is concave, okay? So if you look at the payoff of the convex, you make more, then you lose. You're here, you make more than you lose here for an equivalent one, or you make more the second time than you make the first time. If the market goes up 20%, you make more than twice. If the market went up 10%, if you are in that situation, then you like volatility. If you're in that situation, then you are like benefiting from randomness, what I call the disorder brothers, uncertainty, all these things. Let me take one or two questions here before continuing. As you're asking the question, I can look at, I can see if I Can fix the computer? Yes. Are you considering only the sloping upside of the convex curve? Because it seems to me that on the other side of your smiley face, you would lose. No, no. Yeah, increasing. It has to be increasing. But you're not going to be here also. I mean, it's not going to be harmed here. This is F, F of X. And this is X. And if you're here, you see or decreasing, but not very at the decreasing rate. You see, for the increasing, all right, you make more than you lose. And for the decreasing, it slows down. Any other question? The idea is for you to see the connection between convex, the idea of connection and likes volatility. Think about why the convex likes volatility. You can do it here. Instead of getting X all the time, you're getting X plus 50% sometimes or X minus 50% sometimes. That linear combination is much better than just getting X. So you like volatile, you like variability. I did a simple, very simple experiment on my computer. I generated two kind of series. One where a person knows where he's going, very intelligent person who knows where he's going but doesn't have any convexity in his payoff. The other one, the person who doesn't know where he's going has no idea, but has convexity in his payoff. You see? Look at the difference between the two. Trial and error is an option. Why? Look at trial and error as something like this. You lose, lose, lose, lose, lose, lose, lose. And once in a while you make a lot of money. You see exactly like the opposite of the coffee cup, like what we saw before, something that gains from disorder and gains from black swans. That's trial and error. Okay? You're in the business of trial and error. Someone who does trial and error versus someone with directed ideas is going to perform extremely well when you have randomness. And what we're going to see is why, when you have trial and error, you outperform someone who knows. Because convexity matters a lot more than knowledge. You don't know what's going on. You keep trial and error and you're rational enough, rational enough not to make mistakes by doing the same trial twice or having a result and giving it up. So we can generate rules from this. And I was thinking of seven rules, because people like the number seven. Look here, seven money, rules for life. There's here seven rules for success, all right? Seven rules for weight loss, all right? You can do that as well. And seven rules for change management, okay? So I also said, okay, I'm going to come up with my seven rules. But then I realized you guys will not remember the seven rules as you leave this place. So what you have to understand is how an option works. And once you understand how an option works, then you understand one rule. How to milk. How to be in a position where you are antifragile and benefit from randomness. It's a big, big, big misnomer that trial and error or luck benefiting from luck is a good. The word trial and error is a misnomer. We should not say trial and error. We should say convex function of randomness or trial with small error. Trial and error. The error have to be small. So the antifragile. You're an environment where the errors are small and of small cost and the gains are large and unlimited. And then you are positioned properly towards the black swan. Unpredictability, all these things, this is what's central. And you can model it exactly like an option. And an option increases monstrously in value when there is volatility. Now there's some good news and bad news that people don't quite understand that optionality concept in this book. 4 I'm attacking the notion that we got to where we are because of the superb intellect and brilliance of the ancients. That's not the case. We got to where we are thanks to people who were taking risks, involved in trial and error. The problem is, I call it lecturing birds how to fly people, birds, you know, if you lecture birds, tell them this is how you should be flying and birds will fly according to your principles, then you can claim credit, you agree? And then everybody would believe that because you never see. It's called epiphenomenal because you never see the birds flying without someone lecturing it. Okay? Now it's the same thing with the problem we have with education, right? People have this impressive, this idea that science creates technology. So if you take a book, any book, you have this idea. Science, technology. All right? Practice. It looks like it's exactly backwards in history, but the problem is that birds don't write books. Last time I checked, who writes books? Those who lecture birds. So we have the next generation of people who will forget how it was before. Now let's take the history of technology. To me, it comes from optionality discoveries and then someone formalizes it and dresses it up as you know, this is directed research, this is from the top down. But in fact all comes from trial and error. You don't know where you're doing, you don't know where you're going. But the harm is small. So it's much closer to cooking, you see? How do you cook? It's trial and error. You agree? You don't take a chemistry class, except if it's with Tina, because she wrote a book on the chemistry of cooking. But you don't really take a chemistry class. Then work out the equations, the chemical, and then get your perfect dish. How does it work? You try, you taste. The harm is small. If you're wrong. And by little, by little. And after three or four generation of collective trial and error, you get a good dish like choukot, or you get whatever or sour dough, bread or kind of thing you have here. You see, this is cooking. Well, it looks like technology resembles cooking, but we read the wrong books. And matter of investigating how things were derived. You see, whether it's optionality or whether it was knowledge. In book four, I start book four with this story by Aristotle. He was talking about Thales of Miletus, okay? And Thales made a killing. He was a philosopher and he was tired of the people who were telling him, listen, philosopher, you know, you're a philosopher, you're a schmuck. You get the idea you can't make money. Plus, the guy was Phoenician, so you can't make any money. So he's tired of these people. He wanted to prove that he was philosophizing because he loved that business, not because he was incompetent. All right, now how he's moving by itself, it's moving left, if you've noticed. It's going backwards to the seven rules. Okay, so what did he do? He went in and put some money, some bets on olive presses. Weight loss business, okay? So he put some money on olive presses. He went in and made a bet that it was going to be a demand for olive presses. He put a little money on it. And of course there was a lot of demand. The season was great and he made tons of. He made a fortune, okay, now what does Aristotle write? That Thales made a fortune because he read the stars and predicted, okay, a great season for olives, high demand for olive prices. But in fact, what did Thales have? He had an option. He had a down payment, a call, like the right to use the olive press by putting very little money on it. You see, had he been wrong, it would have cost him nothing. So he'd have tried 10,000 times or not 10,000 times. Maybe 10, 20, 30, whatever. A high number of times at no big harm. You see, he would have been breaking even, you know, breaking even. Breaking even. Breaking even and making a killing once in a while. And that's where really what he had is a big payoff. That was what Thales had was a big payoff. The problem with the story of Thales is that they were convinced. People were convinced that all this great thing came from knowledge, when in fact it was from trial and error. And then if we continue what I did here, and I can find it here, probably in the book, which also is wet, but you see much less fragile because you can still wet book doesn't play tricks on you. You can figure out that in a history. All right. There are a lot of things that were discovered by trial and error that later masqueraded as coming from top down research. And then nobody gets credit for trial and error. We've known, for example, a lot of things think of Euclidean geometry. You think that we need geometry to put buildings? No, that's what people think. No. All right. We had Euclidean geometry since the second century. Okay, all right. Did people use it to build, you know, these big. Did the Romans use it? No, nobody used it. They had their own heuristics, their own tricks coming from trial and error. And they only used it later. Okay, after the 15th century. By the 15th century, there were only four people capable of doing making a longhand division in Europe. So I take that example. Another example is another example. Patent example. Other than, sorry, pyramids. Which one? Pyramids. Pyramids. Well, the pyramids. Yeah. Before you take a lot of things in medicine, sorry, penicillin. Penicillin. Penicillin. What? Penicillin was discussed. We know everything in medicine is discovered like Viagra by error, a side effect of something else. And then people, you know, the Viagra. As the nurses were, were complaining from behavior on the eighth floor when they were doing the trial. So you get the idea for 8. And it was a blood pressure medicine. So there are a lot of examples of stories backward, even cybernetics, you think is wiener. Nor wiener. In fact, just someone sent me a book as I was writing this about that it was practiced by technicians way before and of course, no credit for them. So let me take some questions here. As a computer, as I find my bearings, Let me take a few questions on this. Yes. Isn't there in general a mix of knowledge and luck? That is true. So there's something called. Yes, the bacterium in the stomach that causes ulcers. It was always believed that there was no bacteria in the stomach. Was discovered, quote unquote, by chance by two guys went on vacation, left things in the lab came back, but it was not by chance that they had left the stuff in the lab. So there's a mix of the two. There is a mix coming for optionality is when you mix the two. But effectively there is some kind of improvement in scientific knowledge built from optionality. You see improvement that helps you conduct other experiments. But typically the role of knowledge is highly overstated. So we have this impression that things were driven a lot more by design than they were by luck, you see? But the point is, it's not luck. It is a convex function of luck. It is by tinkering, it's a convex function of luck. This is what is quite important in this, is that you have more upside than downside into whatever you're doing. The biggest improvements we had in drug discovery, when people had no idea what they're doing, but knew that they kept trying dyes, okay, Till they found something. And today we're directing and we haven't been able to find much. Now, we know the genome, we know so many things. We understand bacteria, viruses. We haven't been able to find as much as we did in a period when it's purely trial and error. So let me continue with this idea of trial and error has optionality leads us, remember the seven rules, okay. To maybe you can make 77 rules, you can make a lot of rules, but the principle should always be to be on the right side of convexity, because convexity will pay you a lot more in the long run than direction. So what do you do it. The first thing is make sure that you have optionality. You're not locked into a business plan. It's like you're not locked on a highway with no exit. Yes. So I agree that I think people know that if you're convex, you benefit from extra volume. But you've now popularized this idea, right? This is. No, I mean, maybe people knew about it before. So who are you going to buy these options from? This is very interesting. Now when we talk about antifragile is gains from randomness and volatility in financial option, you buy the option from someone you agree, but in real life, you don't buy it from anyone. Trial and error, you're not buying it from anyone. This is not the financial option. This is a real option. That's number one. Two, you still have an option that you're buying very cheap whenever you buy. If you buy a house and you have a mortgage, you have an option. You see, people don't understand you have the option and for a long time For a long time people were getting even greater option from banks. So the built in, the explicit option is not what I'm talking about here. What I'm talking about is embedded option and in things or option like characteristics in things. And if you take the formation of wealth and history, it all came from this optionality. Other questions? Yes, how do you argue for trial and error in an industry like for the airline industry where the errors are costing would cost an error. There is explicit trial and error where you are willing to take the error, there's explicit trial and error. This is a system that has explicit trial and error where you tinker voluntarily and you have to maintain your error small and make sure it's not like a lottery ticket where the upside is known, someone's overcharging you for it as a financial option and it's open ended, you see. But also there are businesses where you make mistakes, in which case you want to make the mistakes as small for the system as possible and never let a mistake go to waste by exploiting them. So let me give you an example. Just mentioned the transportation industry. Every plane crash which is an error leads to the improvement of the safety of traveling. We haven't had a plane crash in America, the commercial plane crash in more than three years. And before that it was more than three years or a lethal one. So you realize that, that there are businesses that even if they can have a large error to make sure it's one at a time. Not like monetary policy where you have big mistakes or politics where we have one person make a big mistake that cost so much in lives and money like a war, you see, you want your mistakes to be small. When you can't avoid having mistakes, you want them to remain small. And typically top down things make large mistakes. This is what my idea of decentralization of you have to decentralize the errors by making sure they remain small and distributed, okay? But when you're willingly making errors because they're small and because they're not of huge cost, when you're willingly, then you want, you see, then you want to make error. You want to make a lot of errors because they're not as harmful, you see? And the payoff comes from it. But remember one thing, you saw my first graph where I was showing you the returns in this business, okay? The strategy that immediately comes to mind from that is that you can't be directed. If a lot of your returns are going to come from one, what are you going to do? What do you need to do? You have to be as broad as you can in your strategy. That's number one. But you can't do it if you're an entrepreneur. An entrepreneur by design is not broad. And if you're going to have trials, you could probably. You can't have too many trials in a lifetime. But collectively, society has been benefiting from this army of entrepreneurs. Very few of them managed to buy the big houses around here. I hear real estate, she's telling me, is hugely expensive. You see, very few win. And then the remaining people are like soldiers who fall in battle for the sake of the system. Which is why I was calling for National Entrepreneur Day. And then suddenly it happened, by the way. All right, but didn't get credit for happened on publication, on the pop day of the book, by the way, so called National Entrepreneur Day. Because hey, you know what? We salute fallen soldiers as some because they're needed for the system, for the collective. They're needed for the system. To improve the system, we need soldiers to fall. We should do the same for entrepreneurs because realize how few win and you don't want the other ones to be demoralized. You want to thank them for getting involved in their side of trial and error without having anything but errors all their lives. But they're doing it for the sake of the system. And this is what sees there's something moral about what you're doing if you try to improve the system. Something moral even if you fail. Especially if you fail. You see, and this is what people don't understand. We bail out. We tend to bail out corporations that are rotten and fragile. We don't bail out entrepreneurs. We should be bailing out entrepreneurs, which should also make it more honorable. The only place I'm honored to be here, the only place on the planet where it's honorable to fail is here. Before that, it used to be, guess where. A couple of centuries ago. Where? The battlefield? No, no, in entrepreneurship. In entrepreneurship, sorry. England. England that pulled out of the Industrial revolution thanks to these aggressive tinkerers, you see. So you have. That was when. And now the problem is, look at England. What happened to England? The minute you get rich from tinkering, you had the illusion of being able to replicate it through Whitehall and economic policy and Whitehall, a research policy and this and minister of policy and stuff like that. They didn't have that when they pulled up, you see. And the minute they had that, of course, they went in relative decline compared to other places. So you see, you have to let the things develop organically on their own. And tinker the max because that's how systems operate. Let me take more questions. Yes, I'll conclude with the question. Yeah, go ahead. Monetary policy a few minutes ago and today they had a leak of the Fed minutes where they're talking about ending quantitative easing. Ending quantitative easing, Sorry. He's asking me, he's telling me that something not related to the stock, but he knew that I'm very interested in topic which is federal policy today leaked that they're ending quantitative easing. Now, I'm not involved in journalism or commenting on events, but I'm going to talk about monetary policy. Okay. The Fed making it too easy for people to. Shouldn't they be raising rates already? The thing is, what the Fed has been doing, the idea of the government, they don't understand anti fragility. They don't understand that when you have a crisis, you should come out of it better than you were before. You guys had your crisis In California, remember 2000, you weren't born. Okay, so almost. All right, so you came out of it better off than before. No, it cleaned up the system. So you should never let a crisis to waste. Now, what happened in Japan and here and in Europe, they have a crisis now. They shoot to establish a system, to bring it back to what it was before without letting what is fragile break and new things come in its place, you see. So the monetary policy aims at preserving the statu quo. And then also a side effect is of course to enrich those who were involved in the crisis, in causing the crises. And of course, it's not good to have such a, you know, to have very easy money because you don't want people to take money and go gamble with it in the stock market. You want it to go use the money for what you're doing. The real thing, the trial and error that you don't get respect for, for losing a little bit all the time. Other questions? Yes. On a related note, then, how would you think about Chapter 11 bankruptcy, which essentially grants the original proprietors the ability to keep on doing what they do? What do I think about chapter 11? I have written two things about the notion of limited liability. If it's used for this trial and error, then it's very good of limited liability. If limited liability is used to transfer your losses to society, so you get the bonus, the upside and society gets the downside, then it's not good, you see. And what has happened in the capitalistic system is instead of having incentives and disincentives, we have some people who have the incentives and then when they lose money, take banks, they make bonuses when they have good years, and then when they have a bad year, they transfer it to society, to the taxpayer, and to keep their bonuses. As we saw before. Again, remember the two kind of payoff. The payoff where you lose a lot to make small, and the other payoff, which is what you have here, where you lose small to make big. You see, now those who lose a little bit, who lose a little bit to make big, are not gaming the system because the taxpayer doesn't bail you out. Have you heard of a bailout of any company in California? No, because you guys, all your big volatility are on the upside, not the downside, you see, but the other ones have their big volatility to the downside and they lose more than their capital and someone has to pay for the losses, ends up being a taxpayer. Other questions? Yes. Yes. What do you think about the rise of. So you have all these things to say about the nature of probability and being able to predict things, but we also use probabilistic AI a lot, and we are using it more and more in every field. Okay. He's saying that I'm skeptical about the use of probability to predict, and he's telling me that in AI, they predict. In Mediocre Stan, you can use probability very well. It works very well outside Mediocre Stan, when you have things dominated by rare event, okay, these things fall apart. And they fall apart. I just put a text because nobody was understanding my black swan. They would read it and comment and write articles. I'm not even wrong about it. My point is that large deviations occur. They are unpredictable. No matter how many PhDs you put on it, they're unpredictable. And all the people who relied on predicting rare events have blown up. And this is how actually you can make money on Wall Street. You identify those who rely on probabilities to compute the risk of rare events and make sure they're going to go bust like Fannie Mae or like others. And you can make a buck and celebrate being right with these guys. So I'm saying, but for AI, you can apply it for small problems that don't have big tail effects. Yes, you can definitely apply it. We use probabilities for anything. For insurance, on your house, for insurance. Medical insurance probably works beautifully in these domains and in casinos. It's perfect. Okay. Yes. Can you comment on the convex function of luck in drug discovery? Given the fact that the average drug in the US costs about a billion dollars in 25 years of discovery? If You've got any insights on that end? Probably helpful, yes. When you have a. Okay, how did we get to where we are today? You lose small to make big. So because you have to have a lot of trials to get to destination, okay, if you are what I call teleological in the slide that I can show and think you know where you're going, then you can bypass the trial and error because you have great scientists and great scientific understanding. And it looks like pharma got to where it is today through accidents. The side effect of drugs, side effect of things. And now suddenly they woke up and said, okay, now we're going to target results, you see, when in fact it's a collective effect of a lot of firms trying. And those who succeed thought that they succeeded because their brains, not that they succeeded because of their taking advantage of luck. You see, that's the problem of pharma today. And pharma has more problems than that. In the past you came up with a drug, okay, you had to look at side effects and probably a few other drugs. Today we have 850,000 authorized drugs, something like that, and 100 and some thousand existing drugs. All right? So every time you introduce a drug, you got to look at side effects on a patient and you got to look at cool, you see co side effects and then you have to look at tri side effect and all these drug interactions that you had becoming more and more unpredictable. So pharma is in trouble in that sense. Plus there's another thing in my book here that I discuss about some systems that sometimes curing people is much easier to do by subtraction. Subtraction is much easier. You see these seven rules, all these rules are positive means. Do this, do this, do this. Charlatans typically have nothing but positive rules. People who are scientists or philosophical have negative rules. Don't do this, don't do that. It's much more rigorous actually, and they don't have side effects. So instead of curing people by giving them drugs, every time we looked at subtractive methods, remove things like, for example, remove cigarettes from society, you saved more. It's more effective than anything we're spending on drugs. And cumulatively more than anything done since penicillin. You say, I didn't believe the statistic when I saw it and in fact it turned out to be true. If you just remove cigarettes, if you remove corn sugar, okay, you remove by removal. If you remove diabetes, the first thing people do is give people drugs. If you send them to Siberia where they didn't have the money, put them on 500 calories a day for three months, come out of it fluent in Russian or conversant in Russian. And it's more effective than any drug initially. All right, but what the protocols you start instead of starving people because it's a stressor that someone needs, we are antifascial to variation. If we are deprived of these variations, we get weaker. So people, the first thing they do is try to give you a drug instead of removing, instead of removing food, try to give you something and effectively. Now epileptic, if you have a child has epileptic seizures, the first thing they do is give them drugs. Whereas the most effective treatment is removal of all sugars. And we know, we've known now for 70 years. But pharma doesn't make any money removing things from your system. Nobody's going to make any money if you stop smoking directly. If you stop smoking, they're not going to sell you something unless they sell you a method to stop smoking, you see? So this is the problem. There's a gentleman in the back. In fields where it's hard to disentangle luck and ability, such as trading, how do you devise an effective compensation of traders? I believe in something I call skin in the game, okay? Skin in the game is defined skin in the game. It's both moral and effective to remove risks is that nobody should ever put someone else at risk. I don't really care about compensation, okay? I don't really care about ranking because you can't have a trading competition because someone has a strategy that pays off very rarely, okay? He will lose in the competition, he lose every battle and win the war, you see? So can't really rank traders. But there's one rule I have, I call it skin in the game is that nobody should put others at risk without having harm to himself. In other words, if you lose money to your clients, you should be exposed to the same risk. That's sort of both moral and risk management rule. Risk management because in Hammurabi's code it was simple. If the architect builds a house and the house is fragile but hidden fragilities in a basement. You get the idea. When the foundation, like in a bank system, banking system, they look very stable, but they have the cut corners that nobody will see, okay? To make the bonus. And if the house collapses, the architect is penalized. That was in Hammurabi's code, actually he's put to death if the house collapses and kills the owner of the house. So this system, this is the best Risk management rule, because as Hammurabi discovered, but something they forget. Today in Washington, the architect or engineer knows a lot more about the risks. He doesn't know a lot about the risks, but he definitely knows a lot more than the inspector, you see? So if you make people eat their own cooking, you see, they're a lot better off. Okay? And someone sent me, read my book and sent me something, a story in Brazil where they discovered that they could lower the rate of helicopter crashes by forcing helicopter engineers randomly to take a ride, a half an hour ride, once a month in a helicopter. All right? Now, for example, something the Romans knew and Victorian knew, that you make engineers sleep under the bridge. So in trading, what do you do? So long as whoever is involved in a strategy has losses. Small, okay, doesn't matter. All right? Has losses if he can harm others, in other words, he has some incentive, but some disincentive, then we should be okay. It's when people don't have this incentive, when they lose, that the system blows up. Yes, you said that top down systems are more prone to errors and mistakes. Politics today is top down. Does your theory of anti fragility have some implications or some suggestions on how politics should move forward? Okay, the idea that I'm proposing is very simple, along the rules. If you look at these rules very simply, that you want mistakes to be small and gains to be large, okay? And on the mistakes side, you want to distribute mistakes, then you necessarily need decentralization. The most stable country in the world is Switzerland, where nobody knows who the president is. They can tell you that there's Gaddafi's president, of this Assad's president, but they don't know their own president. The system works well. It's so decentralized, it's completely bottom up, you see. So mistakes are small. So this simple concept pervading the book, convexity means mistakes are small, concavity means mistakes are large and rare. You see? And that simple system. So we don't have any more time. I have another lecture, but thank you very much.