Antifragille: Things That Gain from Disorder | Nassim Nicholas Taleb | Talks at Google

MALE SPEAKER: Thanks for coming, everybody. Nicholas Nassim Taleb has devoted his life to problems of uncertainty, probability, and knowledge. He spent two decades as a trader before becoming a philosophical essayist and academic researcher. Although he now spends most of the time either working in intense seclusion in his study or as a flaneur-- I love that word-- meditating in cafes across the planet, he is currently distinguished professor of risk engineering at New York University's Polytechnic Institute. His main subject matter is decision making under opacity-- that is, a map and a protocol on how we should live

in a world we don't understand. We'll be selling books after the talk and Mr. Taleb will sign them for you. He is going to pause for questions occasionally during the talk as well as after, but please go to the mic if you want to ask a question. Please welcome Nicholas Nassim Taleb. NASSIM NICHOLAS TALEB: Thank you for inviting me. So this is actually the first time I speak at a place knowing what it's about, because I've watched your Google talks on YouTube. They're quite interesting because they're long-- no, they're one hour, which means we have

an hour and 10 minutes, because there's always a bonus that the speaker he likes to give himself. So thanks for inviting me. An author should not give you a substitute for his book. A book is a different product-- it's not a talk. So in other words, it's not something that I can substitute. So what I'm going to do is mostly-- what I have here as slides is things to scare you a little bit, just graphs of things cannot be captured fully in a verbal conversation. And I'm going to show you graphs of what is this

idea of anti-fragile about. So I will speak for about, what, 20, 25, minutes, which means probably 30, and then we're going to have a Q&A. But if you're extremely angry with what I have to say, do not hesitate and interrupt. Raise your hand. If you have severe disagreements, disagreement is always very good. This is why these things should be different from the book, because with a book, the author rarely disagrees with himself, you see. Whereas here, you can have disagreements and they're welcome. OK, so we start. If you asked your mother or cousins, someone who

hasn't heard about this book, what's the opposite of fragile, what do you think the answer would be? AUDIENCE: Robust. NASSIM NICHOLAS TALEB: Robust. What else? AUDIENCE: Stout. NASSIM NICHOLAS TALEB: Stout, durable, solid, adaptable, resilient, what else? OK, it's not, simply because if you're sending-- let's look at the exact mathematical opposite. I don't know-- you guys work for Google. What's the opposite of negative here? Positive. It's not neutral. OK, very good. So what's the opposite of convex? AUDIENCE: Concave. NASSIM NICHOLAS TALEB: Concave. Actually, the negative of convex is concave, you see. Very good. So the opposite

of robust cannot possibly be-- the opposite of fragile cannot possibly be robust. If I'm sending a package to southwestern Siberia, and it's a wedding, and you're sending a wedding gift-- so we have champagne flutes. What do you write on the package? AUDIENCE: Fragile. NASSIM NICHOLAS TALEB: Fragile And underneath? Do we explain it to the Russian inspector-- what do you say? Handle with care. So what is the opposite of handle with care? If you're going to send the exact opposite package, what would you write on it? AUDIENCE: [INAUDIBLE]. NASSIM NICHOLAS TALEB: Exactly. Please mishandle. So something

that is fragile, if you map it properly mathematically, you'd realize that the fragile is what does not like disorder. It doesn't want mishandling, it doesn't want volatility, it wants peace and predictability. So the opposite of that would be something that love volatility. What I'm saying may sound interesting and new, but it's not to the option traders I worked with, because I was an option trader for 20 years before becoming whatever, calling myself all these names. I was just a simple options trader before, but my publishers now they want to hide it. But the fact is

option traders, they understand the world and do it in two different dimensions. There are things that don't like volatility and things like volatility, and it's a very, very extremely bipolar view of the world. You almost have nothing in between. And effectively, all I did is generalize this idea, to map it to things that option traders-- because option traders, all they do is they drink and they trade options, so they don't have exposure outside of that narrow field. So all I'm doing is representing them to take the idea outside. And we can do a lot of

things, because effectively, fragility is something I can measure and anti-fragility is something I can measure. But risk, you can't really measure-- unless, of course, you're at Harvard or Stanford or one of these places where they have the illusion where they can measure risk. But in reality, we can't measure risk. It's something in the future. I can measure fragility. And let's see how we can generalize. The graph you have here shows you very simply a fragile payoff where nothing happens most of the time. I don't know, you don't have porcelain cups here. Otherwise, we would have

had an experiment. But nothing happens most of the time. But when something happens, it's negative. You see? So this is a fragile, and everything fragile has this property. To give you a hint, we can generalize it to medicine where you take pills that give you very small benefits. The benefits are small or nonexistent, and the harm is large and rare and often not seen in the past history of the product. That's a fragile. So I take a pill, it gives me small benefits, and then 10 years later, you realize that it gave you cancer or

some hidden disease that nobody saw. It's the same payoff for fragile. Visibly, the robust will have this payoff-- it doesn't care-- and the anti-fragile will have this payoff where harm is small and the big variations are positive, are favorable. So this is sort of like the idea or the general idea. Once you link fragility to volatility, you can do a lot of things, and let me show you exactly the link. I'm going to show you a graph that sort of explains it in graphical terms. Everything fragile has to have disproportionate harm-- in other words, concave,

nonlinear. I'll show you what-- we'll talk about concave in a few minutes-- nonlinear harm with respect to an event size. Let me explain. I mean, you guys at Google and particularly in this part of California, are pretty special. But if you jump 10 meters-- that's 33 feet-- I guess people here in Palo Alto, in this area, they die, no? Very good. No? They would die. Now if you jump 100 times 10 centimeters, you survive. No? It means your exposure is not linear to harm. You're harmed a lot more jumping 10 feet than if you jump

10 times one foot. So we have acceleration of harm. If I smash one of the Maseratis you see in Palo Alto against a wall at 100 miles per hour, I'm going to damage it a lot more than if I smash it 100 times at one mile per hour. You agree? It means that there is this proportionate harm coming, an acceleration of harm. It's a second order effect. Fragility isn't a second order effect. And it has to be so, because if harm were linear, I'd be harmed just walking to the office. This is a central idea.

Anything that has survived and is in existence today is harmed a lot more by a 10%, say, move in the market, 10 meter jump or whatever it is, than by a tenth of that and 10 times the tenths of that. It means it has to be concave to a source of stressor. In my book, I give the-- because this talk is more sophisticated than the book and the contents of the book. In the book, I have to give a story. And it's in the Talmudic literature that there is a king who had to punish his

son, and he had to crush him with a big stone. And given that he was both a king and the father, he had the dilemma that was solved by a local counselor who told him it was very simple. Crush the big stone in pebbles and then throw pebbles at him That's the definition of fragility. Anything that survived, conditional on something having survived, has to be harmed disproportionately. You see, the larger the stone, the more the harm. With this, you can see why large becomes vulnerable to shocks, because, for example, a 100 million pound project in

the United Kingdom where we have data had 30% more cost overruns than a five million pound project. Now with this, not only do we have a definition of fragility, but we have a robust way to measure it. How? Simple. It's the acceleration that allows me to detect the fragility, the acceleration of harm. If I have a bad ruler, I can't measure a child, the height of a child. Do you agree? It's very hard. But I can tell how fast he's growing in percentage. Do you agree? So I don't have to have a great measuring tool

for fragility. All I need is to detect the second order of derivative, the acceleration, because fragility is in acceleration. Now that I gave you the difficult stuff, let me talk about my book. Everything we posit on this idea that fragility is in the concave-- and if I learned how to work this. Hold on. This is a concave. The concave is fragile, and you can see the benefits. And the concave is anti-fragile. To give you an idea of why the concave is fragile-- if you have a piece of information that your grandmother spent two days at

70 degrees Fahrenheit as the sole information, you would assume that your grandmother's very happy, no? That's a perfect temperature for grandmothers. But then the second order-- ah, your grandmother spent the first day at zero degrees and the second one at 140 degrees for an average 70 degrees, I think you would be thinking about the inheritance and all the things that come with a funeral, no? Is fragile what does not to like a negative second order effect and is, therefore, anti-fragile what likes variation and likes these second order effects? Let me try to work this because

I'm a little confused about this, how to work the computer. I figured out how to work it. So let's stop with the graphs and let me talk about the book. Now you're confused enough but intrigued. So let me talk about my book after I showed you these technical definitions. This book-- I realized that this property of anti-fragility, once you had the definition of fragility and then you have its opposite, was misunderstood in the discourse. Like, when governments want stability, they shoot for perfect stability, but something that is organic requires some amount of volatility. It's the

exact opposite of the grandmother. There's a mathematical property called Jensen's inequality that tells you that often things gain on their variability. There are a huge amount of phenomena like that. In other words, you do a lot better, yourself, if you spend an hour at 50 degrees and an hour at 80 degrees than if you spent two hours at 65 degrees, for example. In Jensen's inequality, anything convex-- actually, this is a graph of Jensen's inequality. OK, here it is. It's complicated, I told you, so let me remove it very quickly. So there are things that like

variation. So you can classify in three categories-- the fragile is what does not like volatility, randomness, variability, uncertainty, stressors. The robust doesn't really care, like the Brooklyn Bridge. And the anti-fragile requires some amount of variability in all of these. The discourse missed completely the notion of anti-fragile, so we try to get stability. With government, for example, they want to have no fluctuation and you saw what Greenspan did. If you gave him the seasons, he would have had the seasons at 67.8 degrees, the temperature year round, like inside this office. It's what you maintain, I think,

inside the office. And of course, we would have blown up the planet. I'm glad we only gave him the economy-- he only blew that up. But we do a lot of harm by depriving something organic of a certain amount of variability. Anything organic communicates with its environment via stressors. So this is composed of seven books. Book one, I talk about this difference between a cat and a washing machine-- in other words, between the organic that requires stressors. Do you guys have a gym at Google? Well, there you go. So you put your body under stress.

But you don't realize there are other things you need to put under stress as well. There are other stresses you need to have just to enjoy life if you want to be alive. There's no liquid I know of that tastes better than a glass of water after spending some time in the Sahara Desert. So therefore, there's a chance an inequality at work right there in your life. We realize here and there that you need to stress the bones, but we don't really transfer it to other areas of life. Like, we may not like to have

this architecture, modernistic architecture, smooth architecture-- it's not as pleasant, it's not for us, as something richer, fractal. I'm looking out the window, I have trees. It's a lot richer, and the ancients actually liked that. I don't know if you've been in the Gaudi Building in Barcelona, where you walk in. It's a cave. It's rich in details and I feel more comfortable-- visibly, my eye likes variations, just like your body likes some term of variation and some stressors. So that's book one where I talk about that, and I talk about ethics. What happens is that people

understand that what doesn't kill me makes me stronger. They don't understand the real logic of it, which is that what kills me makes others stronger. That effect, a system that works very well, is a system that has layers. Like the restaurant business works very well because its components are fragile, the entrepreneurs. Otherwise, we'd be eating bad food. I mean, not that we're eating great food all the time, but you understand the idea. You'd be eating like Russia during the Soviet era. So there's some businesses that thrive-- like California, I'm here at the epicenter of things

that thrive because the failure rate is converted into benefits for the system. So this is Darwinistic, except that we can inject some ethics into it to avoid what philosophers call the naturalistic fallacy, that what is natural isn't necessarily great. So we can have-- we should have entrepreneurs, encourage more entrepreneurs in the economy, encourage them to fail, and remove the stigma. This is the only place in the world where there's no big stigma for failure, here in California. We should have it more generalized, because you need them. But also at the biological level, when you starve

yourself, you stress some cells. And the reason we are healthy is because there are fragile cells in us that break first under stress, and therefore you have an improvement within you. You always have the top layer require the fragility of a lower layer. So that's book one. Book two-- again, these books are separate books that discuss different topics linked to that original idea that I gave you. Book two is about modernity, how suddenly you start having policies that try to control, touristify the world, where you have a plan, you have everything is smooth, no randomness

in life. And I explained that, really, we have a lot of people that have discovered over time that you need randomness to stabilize a lot of systems. So the book discusses a disease called interventionism, overstabilizing systems, and some research on different areas, about 50 of them, where in which there is a need for randomness to stabilize the system. Like Maxwell's governor-- it was discovered that if you overstabilize a steam engine, it blows up. So we have that in the economy, you have that in a lot of places. So that's my book two. And in it

I discuss a certain brand of person. I call them fragilista, someone who denies the anti-fragility of things and fracases by the denial. Later on, we'll talk about a relative of the fragilista with the Soviet-Harvard approach to things, from top down, not bottom up. So that's book two. Book three introduces a friend of mine, Fat Tony. And Fat Tony doesn't like predictions, and visibly, as name indicates, he enjoys life. But he's a little coarse. And there's his friend Nero. He and Nero are always fighting. But he taught Nero how to smell fragility. Because you see these

graphs? I had to use my brain to understand fragility. Fat Tony can do it naturally. He can figure out the sucker-- so, his idea of the world is sucker versus non-sucker. And his point to that is any system that's based on prediction is going to blow up. So he finds those who really are sensitive to prediction error-- because, remember, the fragile is very sensitive to prediction error. So then I continue. Book four-- I don't know if I have the book numbers right, but it's OK. I can change it, because I'm the author, remember. Book four

is about optionality and the things I introduce link to convexity. I didn't want to scare the readers-- I didn't talk to them about convexity right away. I tried to get through the back door via this very simple representation. It's if you have an asymmetric payoff. If you make more when you're right than you lose when you're wrong, then you are anti-fragile. And if you have more to lose than to gain, you are fragile. The same applies to a coffee cup, to anything-- the china, anything. And of course, the volatility will be the vector that would

cause you to lose. So I introduced this, but so far the book is not technical, so I introduced via Fat Tony and Seneca. Seneca was also like Fat Tony but much more intellectual. Seneca, the Roman philosopher, who is precisely not Greek in the sense that he was practical and he had a practical approach to stoical philosophy. The guy was the wealthiest man in the world, and he was obsessed with the fact that when you're very wealthy, you have more lose than to gain from wealth. So he trained himself every day to wake up thinking he's

poor and then rediscovered wealth. Once in while, he would mimic-- he would have a shipwreck in which he writes that he lives as if he were on a shipwreck with only one or two slaves. You get it. But he was the wealthiest man in the world writing about how to love poverty. You get the idea, but the guy was good at it. He figured out that you have to always be in a situation where you got more upside than downside, and then you don't have to worry about randomness. And in fact, the strangest thing is

not that he said it. I picked it up and I was shocked. I said, this guy's talking like normal people. What all academics, the view of stoicism, is that they'd be like academics-- boring, and like fashionable stoicism, unmoved by the world. No, they're only unmoved by bad events. That's central. So via Seneca, I introduced that notion of asymmetry-- always have more upside than downside from random events, and then you're anti-fragile. So I go through Fat Tony and Seneca to drill the point and it sort of works. Also, this book has titles and subtitles, and there's

no connection between the title, the subtitle, and the text. Why? Because since I wrote my first book, I sort of was afraid of reviewers. But then I said the best way to have a book is to tick off reviewers from day one, so that way I don't have to fear them. And reviewers, they want to skim the book. I can't understand or figure out what it is about them. Plus, I put a 600 page map-- actually, it's a Google text, by the way. You guys were housing it for free. So far, it's 400 pages of

math, dense math, as backup for this, plus a technical appendix. Just to tick off reviewers, the ideas-- I want people to go through the reading experience. So they can't figure out by then that I'm not talking about the whole thing, the book, is about Jensen's inequality, things that love randomness and how to benefit from it. Now I'm going to go to California and talk to you guys about a phenomenon. I skipped the chapters because here I have more Greek philosophers, more stories. Something very simple-- I'm going to simulate a process here-- this is not in

the book, by the way, this is outside the book-- where you have two people competing. One person has knowledge and his brother has convexity, has a convex payoff. And the difference between them would be the difference between knowledge and a convex payoff will be what I call the convexity bias. I simulated it and look how big is. Well, visibly this explains something that people so far couldn't understand. Trial and error has errors in it. Do you agree? So in history books and history of technology, people usually oppose trial and error versus theoretical knowledge. But whenever

we're able to work with trial and error, they did not understand it had to be convex. Trial and error relies on luck, but luck can hurt you, so it was never modeled as an option, technology as an option. If this model is an option-- and I'm sure there are other questions, so I go over this very quickly. If I were to model it as an option, trial and error, then it would be something that loves volatility-- option loves volatility. And you can have policies that come from it. My idea of flaneur is very simple. I'd

much rather have series of options, like have a long highway with a lot of exits, then be locked in into top down plan like a highway with no exits-- a destination and your exit, that's it. So assuming you want to change your mind, you're in trouble, particularly if you don't know Russian and you're in Russia. That's how they build their thing. So we have two approaches to knowledge. One is top down and one is bottom up. So here there are about 75 pages that should upset a lot of academics because you take-- I took some

evidence, which includes my own field, which was to be derivatives, that a lot of things that we think, that we believe come from top down knowledge and theoretical knowledge effectively come from tinkering, dressed up later as having been developed by theoreticians, which includes these corners up here. Euclid-- people say you have to learn Euclidean geometry, and look at all these things that were built after Euclid. For about 15, 16 centuries, people were building things and never heard who Euclid was. The Romans were extremely heuristic-- very, very, very experienced based, and they did everything using this

convex knowledge. How was it convex knowledge? It's exactly like cooking. You have very little to lose by adding an ingredient and tasting. If it works, now you have a better recipe. If it fails, you lost nothing. So things in knowledge, no academic would want-- I mean, I'm a professor in an engineering department. No academic-- except engineers, because they're nice people-- would accept the notion that knowledge can come from a bottom up. So we have evidence of what I call lecturing birds how to fly. A lot of science comes from technology. But look at a definition--

google technology, science, and it would explain that technology is application of science to practical things, exactly opposite. Anyway, so this is my options theory thing. I don't know if it upset many of you, but typically it upsets academics. So then I go to the notion of medicine. To get to it, I go to something called the via negativa-- how to make something robust. To make something robust, there are two things. Because of Jensen's inequality, it's better to run-- better to walk and sprint rather than just job. So you have strategies that have variations in them.

Bipolar strategies are vastly better than mono strategies. And you see it, for example, with portfolios. It's much better to put 80% of your money risk free, if you can find something like that, and 20% speculative, rather than the whole thing medium risk. It's much more robust that way. But you can see it in the policy of every single monogamous species, which includes humans, but we have data for birds. Monogamous birds, typically, instead of the female opting for a good match, she picks the accountant 90% of the time and the rock star to cheat with 10%

of the time for a linear combination of having someone in the middle. So the idea is you take the loser, the stable accountant, and stuff like-- not that accountants are losers, but you see the kind. And then you take and then you have the hotshot rock star on the occasion, so the linear combination is better. This is explained in the book why things that have variation-- and I use the very same equation with Jensen's inequality to show why it's a lot more stable. Then medicine, of course. This is medicine where you have visible gains from

anything you ingest in medicine and big losses, except there's convexity in medicine. I study the problems of harm done by the healer, whether in policy or something else, in medical terms. It's called iatrogenics-- harm given to you by someone who's supposed to help you. And you can measure iatrogenics probabilistically. I'm going to give you an idea that I just put on the Web today. It's not exactly in the book, but what we discovered from something about blood pressure. So you have these big hidden risks, but if you look at Mother Nature, Mother Nature equipped us

for a lot of natural-- I mean, three billion years is a lot of time, even for Google, so it's a lot of time. So Mother Nature was capable of treating things that don't deviate from the normal. So we have never been able to find anything you can put in your system that has turned out to be 20 years later unconditionally good without a hidden risk like this-- steroids, tamoxifen, all these. You see small little gains. But on the other hand, we should analyze medicine using convexity terms, that if you are very ill, you should have

a lot more medicine and much less medicine if you're not very ill. There's convexity of payoff from medical treatment. But there is a problem, and let me give you the problem. If you're mildly hypertensive and they give you drugs, you have one chance in 53 of benefiting from it, but now you have all these risks. If you're extremely hypertensive, you have 90%, 80% percent chance of benefiting from the drug. So you have this risk, but you have also a huge benefit, particularly when you're very ill. The problem is as follows. People who are once sigma

away from the mean, which nature has treated, by the way, and medicine doesn't help them much, are five times more numerous than people four sigma aways from the mean. So if you're pharma, what would you do? Who would you treat? You have five times more people mildly ill than people who are ill. What would you do? You'd focus on the mildly ill. We'd focus on reclassifying people as mildly ill to be treatable. And also, they don't die, so they're repeat clients who are going to cash out for a long time. So I use this argument

against pharma-- via negativa is by removal of something unnatural to us. You have no side effects, no long term side effects. In a complex system, you need something I call less is more, because adding something has multiplicative side effects whereas removing something unnatural-- like if I stop you from smoking or something like that-- you have very, very small long term side effects. So these are the book so far. And book number-- the last one, seven, is on ethics. It's very simple. It's about a situation in which one person makes the upside and someone else makes

a downside. You're looking at me like, what he is he talking about? Well, have you heard of the banks? Bankers make the upside. The rest of society has a downside. So they're long volatility at the expense of others. And of course, it's my most emotional book and the one that made me the most enemies, because I named names. I had this thing-- when you see a fraud, say fraud. Otherwise you're a fraud-- so, commitment to ethics. And the whole book is about, of course, never ask a doctor what you should do. You get a complete

different answer if you ask him what he would do if he were you. So here, I don't give advice. I just tell people what I've done, what I do. Like when someone asked me for a forecast, I don't believe in forecasts. I tell you this is what I've done. This is what I have my portfolio for the day. Go look at it, if I want. Otherwise, but no forecasts. The same thing is that you should never harm others with a mistake. Why is this essential? At no time in history have we had more people harm

others without paying the price, whether bureaucrats in Washington-- they're not harmed, shamed by spreadsheet-- to economists giving us bogus methods, and academics. I'm not harmed, I'm not the one bearing the harm, so nothing improves in that field. So like Steve Gill telling you, oh, it's peer reviewed by a great journal. Nonsense-- all that's nonsense. They're not harmed by the mistakes or could keep going on with all this-- can you curse here?-- with all this bullshit. You can edit it out. I don't know. I did that at LSE-- I used the F word at LSE-- and

then they told me, well, you know what? We're going to keep it, but it's extremely unusual. So I told them, OK. But anyway, during the Q&A, probably, I can relax more and [INAUDIBLE]. So here I've introduced the book, and add the book with the following. The only way you know you're alive, you're not a machine, is if you like variability. That's it. So if you're anti-fragile, that means you're alive. So thank you for listening to me, and now let's start with Q&A. [APPLAUSE] NASSIM NICHOLAS TALEB: I keep the slides just in case someone asks me

an emotional question. Go ahead. AUDIENCE: Hi. Thanks for coming. It was great to hear you speak. I was wondering if you could elaborate on a related topic of fragility, which is this whole question of a long peace? NASSIM NICHOLAS TALEB: OK. Very good. Excellent. What has happened over the past 200 years and in the military is that you have switched to tougher weapons, so we had longer periods of peace punctuated with war. And if you stood in 1913 and 3/4 looking at recent history, you'd say, oh, it's all quiet now. We don't have to worry.

I'm sure you were really surprised. So when we move into what I call black swan prone variables, it takes much longer to figure out what's going on. And we live in that world where most of the big jumps come from a small number of variables. You guys here prove it. If you look at how much of the internet traffic is explained by Google, you had that concentration. If you look in the book business where you have 0.2% of the authors generate half the income, if you realize you have that concentration-- so the same applies to

wars, simply. With fat-tail processes you cannot make an inference from just a small sample, and a lot of people make the mistake of taking the last 50 years and saying nothing happened the last 50 years, therefore, let's not worry. No-- we have a lot of potential danger. Plus, if it's a pinker book, the pinker book is confused. But other than that-- it's nice. Yeah, crime has dropped, but you can't make statements about whether the risks have changed. I've written about it on the Web. I don't want to talk about it. I get emotional. Thanks. Next

question. AUDIENCE: When you're describing chapter one, I think it was, you said that what doesn't kill me makes others-- NASSIM NICHOLAS TALEB: Book one, yeah. What kills me makes others stronger. AUDIENCE: Right, what kills me makes others stronger. But one of the takeaways I got from your "Black Swan" book was that that's a fallacy, that if you look at a population and you stress it, and-- NASSIM NICHOLAS TALEB: That's excellent. AUDIENCE: --the weak ones die out and you're left with the strong ones, but it's not really true that the stress caused the strength. NASSIM NICHOLAS

TALEB: Exactly. It's the same point I'm making. People think that what kills me-- what didn't kill me makes me stronger, and I'm saying it's wrong. It's typically because there is a selection effect, not an improvement. Let me go through the history of the mistakes made with anti-fragility. There's something in medicine called hormesis. You give someone a drug, their body overcompensates by getting stronger. A gentleman wrote something on anti-fragility from a draft I had. He's a geneticist, and he actually proved that what happens is that if a system gets stronger, it's because some of the components

were destroyed. So when someone says what killed me-- what didn't kill me made me stronger, it could often be that it killed the others who were weaker, and therefore I have the illusion of getting stronger, when in fact, it killed the others who are weaker. That's the idea. It's a little subtle idea which tells you that everything is by layers. I have cells, cells have protein in them, and all that, and typically the weak needs to always be destroyed for the system to improve. And this is how your body improves, not because it overall improves

under shock. It's because you are killing things that are bad, typically. AUDIENCE: Thank you, Professor Taleb. I think that your message about our epistemic limitation is very important. And I had a question about your view of libertarian movement, and how do you think that your idea of anti-fragility fits into those ideas of smaller government and more bottom up approach? NASSIM NICHOLAS TALEB: That's excellent. So what I'm showing here is actually-- I don't know if it's a libertarian view, but it's definitely a localist view in favor city-states, a lot more robust, because of the side effect.

A small government works better, not because it's small government, but because it's-- you get the idea. Top down government doesn't work. Now you can have probably a dictatorship in a small village and it may work. So I cannot prove that it's not private versus public. For me, it's large versus small. Small has the ability to survive and a large gets disproportionately weakened by unexpected events. And thanks for linking it to epistemic opacity, because this idea of fragility being measurable solves the problem of opacity. I don't understand the environment, but I can pretty much figure out

I'm fragile to it. Go ahead. AUDIENCE: So you said you didn't believe in forecasts, so I won't ask you to make a forecast. So what's in your portfolio? NASSIM NICHOLAS TALEB: I don't want to answer the details of it because I'm talking about my book and in a year it will change, so I can't talk about that. I'll tell you that if I were compelled to produce a forecast, but I don't like to forecast. But anyway, the book tells you what I do, so that's what I do. And it irritates the critics. Everything that irritates

books critics is wonderful for books. But again, consider this class of phenomena that benefit from harm-- rumors love repression. Try to repress a rumor and see what it will do. Go stand up and deny a rumor and see when a politician says, we will not devalue. The rumor is wrong. You know what happens-- it's the best way for a rumor to spread. And same thing with books-- try to ban them and see what happens. There's some class and I call it-- what do I call it?-- refractory love, where people like in Proust where people have

obsessive love. And the more you try to repress it, the stronger it gets. A lot of things get stronger under repression and belong to that class of anti-fragile and it all can be mapped as convex. Yes? Go ahead. AUDIENCE: Some of the systems that you mentioned, the difference of-- I'm just trying to compare that to the financial markets. If you have a period of stability and all of sudden you get cancer, you usually don't recover back yourself. Or if you have a fragile item, it breaks down, it usually doesn't recover itself. But how would you

compare that to financial markets? It doesn't have an external property that-- NASSIM NICHOLAS TALEB: OK. Let me link this to the question earlier, because now I remember that I didn't give a full answer to the question earlier. A system that has a lot of parts, independent, that break sequentially, is going to improve. How? Take, for example, transportation, or take engineering. Every bridge that collapses makes every other bridge in the country safer. So the probability of a building that collapses makes every building in the country, or the probability of the next building collapsing smaller. Smaller or

equal, but it doesn't get worse. So when you have a system composed of small units that break sequentially, fail without contagion effects, the systems improves from failure. And, exactly as what kills me makes others stronger, and that's a benign system or a system that's actually anti-fragile. Now, take banking, take large corporations. When one fails, the other-- it increases the probability of the other failing. Then the system doesn't work well. That's one thing to answer him and get into your point. He's asking me whether financial markets, what benefits they have? Well, people think that they're good

at providing information. In fact, they're great at masking information and that's why it works. It prevents panic. Say if someone is predictable and comes home every day at 5:30. Boom, you can set your watch, he walked in. And one day he's late? What would happen? Two minutes and everybody freaks out, he's not here, where someone more random in his arrival time would not cause a panic. Well, it's the same thing with prices. That's one of the aspects. Another thing with prices is that volatility prevents big collapses because it's just like a forest. You have flammable

materials. Steady, small forest fires clean up that flammable material and don't let it accumulate. But what happened with Greenspan by stabilizing everything-- no volatility or minimized volatility, something they called The Great Moderation that resembles the Great Peace-- you had a lot of hidden risk in the system, very explosive, ready to explode. In effect, we saw what happened-- they blew up. So this is where financial markets, by bringing volatility, clean up the system periodically. That explained it. Thanks. AUDIENCE: Thanks a lot. The question I had was how does your work reflect on how you think about

conglomerates and family businesses, especially in the developing world where there seems to be a high concentration of preservation and a model where they actually look for stability versus choosing variation? NASSIM NICHOLAS TALEB: This is a good question. I don't know much about-- I looked at family data for businesses, and effectively in what we call today the OECD countries. They have stayed in power because they have what I call skin in the game, among other things and among other qualities. Now, conglomerates-- I have no idea. I just like the conglomerate I work for-- namely, the owner

of Random House, Bertelsmann, because they're not listed in the market. Although I like market volatility, I don't like people to fit the company to the security analysts that don't understand hidden risks. And the stock market tends to push companies to hide risks, and it fails, because the security analysts don't have the tool to analyse second order effects. Go ahead. AUDIENCE: How much of the anti-fragility phenomenon that you're talking about across systems is really about evolutionary learning in that the two curves, knowledge versus-- I forget what the other one was labeled, but it was the anti-fragile

curve-- is really about two different forms of acquiring knowledge. One is for acquiring articulate knowledge through articulate processes and the other one is for acquiring inarticulate knowledge, the kind of knowledge that Hayek talks about, where the system learns without the human beings necessarily being aware of what it learns. NASSIM NICHOLAS TALEB: That's a good question. He's asking me how much of-- there are two types of knowledge, [INAUDIBLE], knowledge top down, bottom up, heuristic knowledge versus what we call propositional knowledge, or things that aren't formalized, and so on. There's been a dichotomy through history between these

two [INAUDIBLE], a lot of people. But the first person who discovered it-- let me give you the background-- is Nietzsche. Nietzsche had [INAUDIBLE] between-- actually, even Seneca discovered it, but we attribute it to Nietzsche. When he was 25, he wrote the most beautiful book probably of the century, "The Birth of Tragedy," by showing tension at humans between the rational Apollonian and the deep, dark unexplainable force, the Dionysian-- depends if you're British or American how you pronounce it, from Dionysus, the god of wine and [INAUDIBLE] thing. And he actually-- I think Nietzsche is the one who

used the word first, creative destruction. Nietzsche, not [INAUDIBLE]. An economist cannot come up with something that deep. So Nietzsche spoke about that, and to continue, he went after Socrates for saying whatever you cannot explain isn't necessarily stupid. And effectively in my book, Fat Tony has a debate with Socrates. You can imagine a guy from Brooklyn debating a Greek philosopher, and I'll let you guess who's going to win the debate along these lines. So effectively I go along these lines, except that what I've done is very simple. I don't have a theory in here of anti-fragility.

People can talk about complex system, however they are. I have a descriptive detection. This here, I proved very simply, that I can detect fragility through a second order derivative. So in a way what I have is more like phenomenology, which is not at the level of a theory but something lower-- a way to map objects in order to work with a world we don't understand. So in a way, I don't have a theory how things come from, but I of integrated this dichotomy you have between the bottom up, unexplainable, we don't know how we do

it. There was one thing I would like to mention, since there's time for another question-- one important thing-- that effectively the longer we've been doing something that we don't understand, the longer we will do it. In the book, I say time is the only detector of fragility. Remember one thing-- time is volatility. You agree? Time involved mathematically-- they all are the same. Disorder, time, entropy, volatility-- approximately, I call them siblings, brothers. They're not exactly the same, but they're like fraternal twin brothers. OK So with time, what was fragile eventually will break. So it's a very

simple law. Whatever is not perishable, namely an idea, will have a life expectancy that increases with time, which is shocking for technologies, but let me explain the rationale. If you see a human and he's 40 years old, you can safely say, not knowing his history, and he's not ill, unconditionally, that he has an extra 50 years to go. You agree? That's mortality tables, condition and mortality tables. He lives another year. You know that his life expectancy has decreased by little less than a year. So his life expectancy decreases every day he lives. If I look

at a technology, how old-- all of you need to know. How old? 40 days. Technology-- a book and idea [INAUDIBLE] that are not perishable. Our genes, for example-- not our bodies. How old, technology? 40 years. Very good-- it has 40 years to go. Next year, 41 years-- at least 41 years to go. So the life expectancy of technology increases every day, an idea of everything, believe it or not. So this is why we have bicycles and they probably will live longer than the cars, cars more than planes. And of course, now we know that the

regular plane's better than the supersonic planes. And I'm talking about this here in Silicon Valley. Why? Well, very simply because-- and we don't understand why. We don't have to understand anything. Time-- there's an intelligence of time that's at work there. A book that had been in print for 3,000 years-- in print, or at least read for 3,000 years-- that you get in hotel rooms still, will probably be read for 3,000 years. Regardless of what the latest intellectual tell me about the rationality or not rationality-- I don't believe in rationality. I believe in fragility. Things like

that you could apply to technology. You can say the red car, the first red convertible car, was born when-- 38 and 1/2 years ago? That's 38 years ago, approximately. But of course you'd be shocked now if you see how much of what we have today resembles ancient days. We still use watches, glasses, 3,000 years old, chairs, desks, silverware. We're trying to cook like our ancients did. I have in the book a picture of a kitchen from Pompeii in Italy. It's 2,000 years old and it's no different from a good Italian restaurant's kitchen. So this is

to tell you that there are things we don't understand in the world, but we can understand them via-- there's no rational means to understand why people use something like a technology, but we can understand via just fragility, the concept of fragility via time. AUDIENCE: So I'd like to ask-- I suppose not, really. It's a separate question more than a followup. But returning to the great debate between Hayek and Keynes, especially with regard to the Great Depression, oversimplifying-- the way I view is that the Austrians were saying that Hayek was saying, OK, you've got this cascading

failure. Let it fail, because systems that fail under cascading failure need to be beaten out of the system. It's the only way it's going to learn. The basic Austrian point of view was we're better off in the long run, which is the long run that Keynes was responding to. NASSIM NICHOLAS TALEB: This I agree. Because we're short on time, so I'm going to answer you very quickly. He's comparing system-- I think that it's quite artificial to say Keynes versus Hayek. Keynes was not an idiot. He was a very smart human, and he would think differently--

vastly smaller than people who write for New York Times and claim that Keynes said something. And then also, of course, on the risks of the system, I also have to say that the problem is you can't suddenly stop doing things. If you look at medicine, the rule is if you're slightly ill, let things take care of yourself, because whatever medicine you're going to put in probably will harm you probabilistically a lot more than help you. But if you have cancer or you're very ill, see 10 doctors, not one, 10. You get the idea? So what

happened was interventionism, was statism intervention, overintervention. The state is never there, the interventionist is never there when really needed because of depleted resources. And this is what happened to us, by the way. The money's gone, all right? So it's a different problem. I don't know if I can claim with this thinking to be fully against state intervention, but I would say the state needs to intervene. It's got to be extreme circumstances and for a very temporary situation to just avoid pain, starvation, and stuff like that. If I use the same linear argument, reducing extreme unhappiness

is different from raising happiness-- two different animals. We can take one more question, I guess. We have one. We started at three, so we have three minutes. That's what I owe you. AUDIENCE: When you talk about finding fragility by talking in the second derivative? Can you give some more details, like the second derivative? NASSIM NICHOLAS TALEB: Yes. It's very simple. You take a company, you lower sales by 10%, they lose $100 million. You lower sales by an extra 10%, they lose $500 million. Accelerating losses-- it means they're going to be a lot more harmed by

an adverse event than a regular company. It's a very simple test. It's so simple that people were ashamed of telling me I was right. You see, very simple-- acceleration. Take the stock market-- take a portfolio. The market's down 5%, I lose a million. The market is down 10%, I lose $5 million. I'm fragile. It's that simple. It's the same argument when you say that if the market goes up 10%, do I make more that if the market went down 10%? I'm anti-fragile. It's the same thing with a lot of situations. So that brings fragility-- you

can measure it that way. Size cause fragility. You can measure it that way. Thanks.