Antifragile: Things that Gain from Disorder

each year Microsoft Research helps hundreds of influential speakers from around the world including leading scientists renowned experts in technology book authors and leading academics and makes videos of these lectures freely available thank you for coming my name is Jamie Draves and I'm here to introduce Nassim Taleb who's joining us as part of the Microsoft Research visiting speaker series Nasim is here today to discuss his book anti fragile microphone you can't hear me yeah I can hear myself really well okay I'll lean in thank you and the seam is here today to discuss his book anti fragile things that gain from disorder anti fragile refers to something that is unbreakable but not because it is impervious to assault anti fragility allows an entity to withstand all the black swans coming its way and to absorb their forceful volatility and emerge stronger for it Nassim Taleb is a former trader and is a distinguished distinguished professor of risk engineering at New York University's Polytechnic Institute his work focuses on decision making under uncertainty as well as technical and philosophical problems with probability and meta probability as you no doubt know he is also the author of the Black Swan please join me in giving him a very warm welcome anyway so let me talk about fragility you guys are techies now so let's meet a key forget about the usual waste of time talking about candles and this metaphors and packages and stuff it's all in the book let's talk about fragility let's define fragility okay unfortunately this is not porcelain so we can do an experiment to try to break it but the fragile is what doesn't like volatile we have this payoff small gains or no gains okay over time this is time and big losses so harm exceeds gains think of a bank portfolio you make make small and then you have big losses like shorten option I was an option trader for 20 years and we view the world and two kind of packages one that gains from one that loses from volatility this one and one that gains for evolved to the opposite payoff now this is exact payoff over time and time series so this is time okay and on a vertical it's payoff small gains big hard as you can see this applies also to medicine if you go to see a doctor who gives you some random medication small gain big potential harm at some point in the future that's a payoff of short Volturi if you look at a coffee cup at a table imagine it's not you know it's in something that breaks regular material it wants peace and predictability so you can map fragility as something that doesn't like volatility don't hesitate to ask me questions during my entire conversation you see so this is an interesting techy way to view things because from this this is in time-space we can have an even better definition of fragility and in ways of the tech facility as follows in my chapter I asked people not during the chapter chapter 18 and 19 or I forgot to read it means meaning they're gonna read it if they're intelligent that's all she tells someone not to eat it you know what happens in it sighs the stone this is harm harm were part of harm isn't negative okay you know game is in the past right like here gain losses there's a story in the Talmud of a king who had a son who and who committed whatever violation the can you hear it raise your hand when it can't hear me right or if you want to interrupt to say something aggressive it's okay as well right so the the son committed the violation the the King has to punish the son big dilemma I don't know if you've been either King or had a son who committed violation has to function with a big stone but you know it's a very difficult situation how can you solve it the king had an advisor was very shrewd he told them okay you break that big stone in small little pebbles okay and you felt them and what happens the son is not gonna be hard so it's very simple this is stone and this is harm you have acceleration you see a 10 pound stone is more than ten times more harmful all right then one pound stone with this you get a definition of fresh herbs that give you that so this is a real book all right 550 pages of you know I've we're going to get into this but I'm giving you the gist you go straight to the core of the problem this is what it is see accelerated harm take your body or anything that has survived has to have nonlinear harm if harm were linear it'd be like this you see this is linear this is concave for extreme events were harmed a lot more when you concave than when you're linear let me let me explain you know if you jump 10 meters definitely even in Microsoft people die no I'm not sure people died in Microsoft I love that happy but but I'm sure I mean it's possible that people that you know outside Microsoft people died 10 years fault they died but now think about if you fall 10 times from 1 meter what happened to you nothing it's just people laugh right now a thousand times 1 millimeter nothing if harm were linear the cumulative harm from small steps you say would be would kill you just go into the office because one hundredths of a millimeter okay you get that you know this impact would kill you you say so you realize now we have a definition of fragility I still didn't get to antifragility but once you have a definition facility you understand how to map it mathematically how to define it how to measure it and how to use it let's think of size project 100 million pound projects have a lot more cost overruns in UK 30% more than 5 million pound projects so we have this economy's of scale from random shocks this explains why elephants break a leg very easily and you don't have as many elephants as mice even if you weigh in by the by the way the real explanation is probabilistic is that you have the distribution of the small harm is vastly higher first let's take our squeaks all right if you want you have to be concave door squeaks incidentally I want to a bathroom sometimes say what to do in case of earthquake this you've never seen in New York so I thought of it coming to this lecture that you probably have had eight earthquakes since I've been speaking very small scale you see so it probably is really an earthquake small micro everyday in the world all right every day but so if we're linear to earthquakes would be gone you see the idea so you have to be nonlinear whatever has survived so this is sort of like the idea I have 450 page of maths on the web and the technical document that comes with this to back up these things with theorems and stuff like that right but that's the main idea of fragility what's your method as I said now let's talk about anti fragile all right the anti fragile is convex to sort to a stressor after that the shape with respect to a stressor not concave convex you like variability if your convex to any stressor say terminal variability what is the perfect temperature here sorry degrees you do better if you spend if your linear combination of 80 degrees and 60 degrees or even bad at 50 and 90 even but at 40 and 100 you say for the same average it means if you like variability it means that you are you love randomness you're going to have the opposite pay off and this is convex this is concave I'm being techy because you guys are techies but tell me if it's too complicated and we can go back to what I'll say tonight that you know talking to retired the readers you know metaphors about packages that break and stuff like that but I'd rather you know I'm much more relaxed you know talking about things in a that's how I do things in life you see and then later on translate them into package them into metaphors so with this we can classify the world in three categories fragile robust anti fragile the robust is not the opposite of fragile visibly because the robust has a very simple definition the fragile the upper bound is the upper bound is unharmed the lower bound is harmed the robust the upper bound and lower bound are unharmed the anti fresh hold only the lower bound is unharmed you're open to the upside okay so doing this for a system for any package now we have an exact definition of fragile and anti fragile and we can match things biologically that the anti fragile loves variation it also mapped from this we have three map as I showed you one you know two maps so far one in time-space and one in functional space yes same as convex it is what not convex not a contact is the way it if your convect with respect to a source of randomness or source of variation your anti fragile so long as your convex to it did that down so that it's sixty degrees let's say that there was actually support then you would have then you would be like this you could be like this all right it'd be convex other than concave you see you'd be convex concave you can have all kind of shapes or it could be flat or could be even concave you say in fact we all for temperature have the shape this is a temperature you like this all right let me explain if you spend 140 degrees 70 0 degrees all right if I if your grandmother spends a whole chapter that talk about grandmother's if you spend one day at 0 degree the other day at 140 degrees for an average of 70 degrees you're not going to have a grandmother degree but within small some small thing you're better off having some variation you see so everything is convex with respect to a source of randomness and up to a level up to some degree for example let's talk about the the body weight lifting you do a lot better lifting once a hundred pounds then ten thousand times millions of them or whatever thousands of you know at one ten thousandth of a pound you get the idea right you do better but visibly if I lift if or someone to some my shoulder five thousand pounds you know it's going to happen to me you know be transformed into what to a corned-beef know so you see the idea up to that some degree right so let me so now that we have this idea that now that we have the graphs I'm relaxed it's very easy I go I'm not making an effort anymore I like what I went through during my book tour I can talk about my book and refer to these alright now we can there's actually a third dimension where I call the probabilistic distribution of things why everything anti fragile loves randomness that map later so now let me start the regular presentation of the book okay so this book is about this quality that people forget in a discourse called anti fragile because they have wrong definition of fragility well in fact the ancients understood it very well and anti fragile is whatever gains from I call it the disorder brothers randomness variability stressors chaos volatility and similar structures I learned about that when I was an option trader and we had something called long damn long gamma means if there's big market variation or turmoil you make some money up to a point of course right you may make money up to 10% crash but not up to 30% or it could be what we could open ended long gamma you make all the way alright so we had these structures and I was trained and thinking these terms and map them so that was my profession but couldn't communicate with people you see I didn't realize my definition of fragility in finance which was very simply if you harm more by 5% move than five times a 1% move you are fragile okay I didn't realize it was so universal definition of fragility I didn't realize till I started thinking that effectively the Romans have a word for something that lights volatility it's called hormesis I mean of the Romans and the medical profession has worked for it hormesis they had two things there's something for what people call resilience called myth realization you take a little bit of poison every day and you're going to be immune for that problem not going to be better off but immune against that so survivalist and actually in here there's the other of Nero kept taking you know this thing but the problem is as you all know you could be anti fragile or you can gain immunity by taking drugs but you can do it you know with respect to other sources of harm namely swords so her son wanted to kill her in the end he sent someone to kill her with a sword so she could be poisoned so there's mislead accusation and the other category is hormesis hermes is when your overall system gets better from little bit of poison or little shock so I thought about it I thought about it I said aha now there's something that you can map mathematically and there's a mechanism of overcompensation at work like mechanism redundancy just think of the way even at Microsoft people come with two kidneys no oh no you have two kidneys why do you have two kidneys it's your you have redundancy against random events because when you have redundancy you don't have to understand the future as well if you have two kidneys you don't have to be precise in understanding the future as someone who has you know only one so you know and there's seven or eight ways you can lose a kidney from trauma lead poisoning your cousin hitting you wild animal two other things you don't have to understand the environment as much okay well it turns out that there's also redundancy in a way we have strengths if I lift a hundred pounds my body won't adjust delisting hundred pounds it will project an extra twenty pounds you know and it will code like a code like you guys do coding it will code for 20 more pounds of muscle over time or 24 more pounds of lifting you get the idea and that is a form of redundancy its overshooting to the future you you see the idea so I realized that in my business of risk manager effectively there's no better than nature it does sing in a certain way that is extremely after three billion years we can give us some credit and a statistician there's no more respectable than an N that large all right so we have so this is how I start developing the idea then I realized hey you know what was very bad at predicting in socio-economic variables we can put a man on the moon all right but even Microsoft can to mark forecast at what price the market is going to close tomorrow these complex variables and we're getting far far away from that so you have to focus instead of predicting and all these method predicting on being robust or perhaps anti fragile gain from randomness gain from errors and try to identify what gains from randomness and gain from error so this is a book now you map to everything but I wrote this book in a way to prevent any reviewer from being able to skim it if they do and so far no reviewer got these points see when I'm talking this is a core I started the book with this no reviewer got'em except for one right it's a technical book visibly dressed in an easy language but if you're a mathematician or you're scientist you get it right away you see and then it's time to get it right away right but the bet people are professional viewers they lost completely lost in it and I make sure that let me give you the chapter titles if you want to be amused to realize why they're never reviewers I'm never going to get it like the the after the prologue where the idea is clear in the prologue history Litton by written by the losers never marry the rock star the tell'em I love some randomness the souk and the office building right time and fragility the the Philosopher's Stone and its inverse so I mean they're not going to be able to pick it up right so so it's you know the other books are written exactly the same style by the way the other two books and it worked so you might pass the middleman the reviewer and you go straight to the consumers okay so this is the book so it's composed of seven different books seven different topics that apply this first topic the most interesting one is on evolution evolution is wrong long randomness think about if you have ten offspring you're never going to have improvement of the species if you don't have some randoms in the environment that make you know that appreciate between degree and then also genetic drift some random is along the line you see so evolution likes randomness it's a problem why is it a problem well because we Cynthia light them and we have some social justice can you reconcile of course we can reconcile and that's sort of the challenge I've taken in in my book one where entrepreneurs and here you're you know a tech business entrepreneurs really the system improves thanks to the failure of entrepreneurs and then I learned something in biology from a gentleman who when I remember her Missy system gets stronger he discovered one thing that no system doesn't get stronger some parts of the business under a shock the weak parts get cleaned up in you you know the restaurant business as a business composer restaurants so layer restaurant business the restaurant are fragile for the business to be to work well to be robust you agree bankruptcies help clean up the system therefore the system becomes robust because the parts are fragile it's the same thing in your human body cells need to be fragile for for you to experience hormesis improvement because when you starve yourself weak cells go first and then within cells protein he modeled it on one doctrine was published a paper in jeans if you go to my website I have all these papers linked to this concept so we realize evolution loves volatility ok and and and and this is a rarely map like this so in this representation so I had to do some work some mathematical work on that so this is the book one but of course it's written in poetic level what kills me makes other stronger you see and then why do you have the difference but you know that since the Enlightenment we talked about I the individual before that we never never use collective the tribe the collective you died for the tribe this is why you have suicide bombers you know they don't consider themselves as a unit but also you have but now we have modern society and you're not going to commit harakiri for the sake of the computer industry to get better by failing so we need the mechanism to make failure acceptable well you seem to have some give social acceptability to failure not in other businesses certainly not in Japan and not in Europe when you fail you're loser you see so it prevents people from doing these things so we have to find a mechanism to make people fail to improve the system that's sort of book 1 book 2 I'm going to go through the books and then your lacunae I can you know answer whatever you know divert you into whatever I haven't covered ok book 2 I talked about modernity the problem is look how smooth these surfaces are okay look how smooth this wall okay nothing is fractal in here look outside you see the green outside you see how fractal it is a complete different different different world you say so you're psychologically we like variation we are long we like variation but that modernity doesn't know it wants to make it comfortable you say Butler natee wants your children on Prozac - are there mood swings so the problem is that they don't realize that there's a difference between what I call the cat and washing machine what's the difference that the mechanical doesn't need stressors to operate the the organic communicates with the environment via stressors you see you have to you know get stressors right this is how the information system so the problem is if you stabilize everything to make everything comfortable like we have here what happens to you you get weaker your bones get weaker if you spend time in bed your system gets weaker and we sort of understand it because I surely have a gym here No so with the contradiction you have a gym I'm sure and you have elevators you have a gym yet you have smooth surfaces and then you go to the nurse and the and they give you antibiotics No okay so you understand very quickly so this is there's something about modernity try to make people comfortable that weakens them and said so I call that interventionism and the person over intervening in a system is called freshly stopped and typically the fragile eyes the system twofold first by making it too comfortable becomes weak and two by not being there intervening when needed mr. greenspan you've heard of him he had he inherited the the nature he was made sure we have sixty eight point seven degrees year round you get the idea he wants to do same for the economy and fragile Isis the economy because without stressors you don't have you know you have like flammable material accumulating and risk terminating on the surface and then blows up so what you have is very stable but then blows up rather something volatile things volatile things you know Marcus needs some volatility because that's the way information is conveyed you can stifle information take for example dub democracy needs volatility and I compared on the Black Swan Syria which you heard of Syria may hear about it now which has zero political authority to Italy that had a lot of it and which country do you think is safer I had lot of letters saying Syria was safer than Italy he was Mel Stein okay Syria had zero you know a political volatility we had the volatile with him you get yeah that was on the Black Swan and now I'm taking off Saudi Arabia I mentioned Syria and Saudi Arabia and a Black Swan and of course Egypt look what happened you see over stabilize so this is interventionism this is modernity so I go through that and the problem is you're never there for the big like in medicine you're never there for the big things now book three I talked about prediction forget about it it's not interesting and then introduce that Toni have Fat Tony a character who smelled fragility and makes a buck with those who lulled by false sense of security making pennies boom both off once in a while okay so this is like Fat Tony introduced Fat Tony and then I introduced Seneca by defining let me redraw the convex for you to see the link if I have a payoff like this if I have a payoff like this I'm here alright I'm sorry it's convex let me make it very convex this is payoff this variation we're here the market is here okay if the market goes up one dollar I make two million the market goes down one dollar same amount I lose 100,000 I'm convex well more upside and outside well the person who figured it out is Seneca the Greek the Roman philosopher and he was really Roman they hated intellect the light practical things and he was a wealthiest man in the world at a time and he figured out that you're harmed you see when you're very wealthy because I have no downside than upside you're fragile and every morning he would wake up try to convince himself that he's very poor once in a while like fix a shipwreck and then oh I discovered all this wealth so he wanted to make sure that if he loses all his money which happened at a time he wouldn't be harmed you see because he you know so that's sort of like the idea and so I have a discussion of Seneca who to me is probably the greatest thinker academics never got it because academics think that Stoics are people who have no emotion in fact you didn't want to have negative emotion more upside than the outside and I define antifragility as more upside than the outside something you may like now my central chapter I just put on a web today a distillation of that my central book is on discovery tinkering as compared to rational top-down tinkering is like optionality how it likes you know bottom-up tinkering you have little to lose a lot to gain all right as a process we show that the evidence that a lot of things we think came from science in fact in fact came from heuristics of practice and I do a random simulation today is an article on edge right where Heather simulation have set a lot of people where you show that two brothers one has knowledge the others just incurring and how much further you can go by by you know by by tinkering and have more upside than that side aggressive tinkering as compared to someone who knows where he's going to Tilly illogical and effectively you look at history of Technology and when you look at definition of Technology as defined by Harvard or all these places that use good to drop out from as one of your leaders have they think that technology comes from science nonsense science comes largely from technology you see even in some cases sometimes you find predecessors like Euclidean geometry they were building cathedrals and didn't know how to divide you know they probably thought that Euclid was some Greek poet or something they didn't understand all right but then later on we find matches you see and that's a revolution by thinkers so jet engine thinkers engineers knew how to function as a later on we found the theory you fit the theory back to the past I call it lecturing birds how to fly upsets academia at all levels when I show formal knowledge versus wild optional knowledge that gains from a certainty if you're thinking you want maximum exposure to uncertainty well the University of the prison you say your prison of a curriculum all that nonsense so and and so this is a chapter that really really really really gets university people angry all right is like emotionally angry so particularly status the university system people and you okay even more from my talk yes your friends your edges anything you put out this morning sorry you put out an essay an edge yes yes yes I think it's interesting to this group because large will be research people in this I see by tinkering you're actually creating a situation where you're you're able to tinker and take advantage of basically positive black swans exactly what happens is that there's a theorem that people don't realize and as follows this is anybody if you hear into probability okay any probability distribution is skewed right okay take the log normal what's the mean of a log normal it has a variance in it so you increase the variance what happens to the mean increases if you have little to lose and not to gain you see the mean of a distribution that's bounded on the left and open on the right the mean depends on a variance and otherwise you inject uncertainty in the system your expected return goes up the exact opposite of a plane right take a plane ride you know go to New York tomorrow six hours all right I'm lucky if I make it five and a half hours all right if you inject our certainty in the system it's not going to make the plane ride one hour but it can take two weeks to get there you say September 11 or somebody you get the idea any uncertainty so the systems that are fragile are defined as their mean the grades when you inject uncertainty and this the mean increases when you inject our certainty and that's pretty much the gist of that article the thing is when I talk to probability people they get right away that any distribution of the skewed right has a mean like lognormal has mean as I might have the Sigma square plus something right so but when you talk to normal people that's hard for them to understand why are certainty is good so you have to use a lot of patience for that so this is sort of like the edge essay and my chapter here which is vastly more aggressive any addressing I say in my chapter here I thought about frauds all right you know a lot of things that are froze effectively that tell little that anything that compresses our certainty to telling it channeling it is not very good and you see obviously a lot of situations like the black Scholes formula stuff like that that were developed in a sophisticated form by operators and later on claimed by professors you know as coming from them because they can't mention that that knowledge can come in any other way and in it of course we have the two tensions and he who discovered that problem ethnically you know it heard the creative destruction people assume it's an economist Goshen Peter how can anything intelligent come out of the Academy economist and possible gotta be someone else you lesser rule if it's if it's attributed to an economist it has to come from a philosopher all right non-academic initially of course who then semi academic all right a time create a disruption and he said that Socrates destroyed the balance between Apollonian which is reasoning and unison which is that deep dark force in us because he made it this respectable to do something you don't understand he made the unintelligible seems unintelligent and who's going to debate Socrates my fat Tony friend and what as you can expect think of a guy from Brooklyn all right who made makes his killings squeezing people debating Socrates you know who's going to win all right Fat Tony wins all right by the debate but it still sir he has to you know abide by philosophical yet but he out Socrates Socrates by cross-examining him so this is sort of my book on innovation the one I like the most and have that distillation of the article but but I go much deeper in here you know against the soccer mom through certification educational system we got hit you know it's something about the education system people don't realize do you think that education raises wealth and country it raises was for a family it stabilizes income actually for family than in raise wealth education comes typically if you take countries that are rich how the degree of Education education came after you see you get rich and then suddenly they be phenomenal the thing that is caused by education only people who thinker he like you guys understand it but if you go in on East Coast Harvard Square put to death right there you know for seeing something like that so book 5 how to detect anti fragile fragile steel systems if measurable yes yes right so it seems to me to your HSA in what you just talked about also your bread the address a yes now exchange some thoughts about honor on Twitter this morning oh really what's your name I'm mr. Griffin ok very nice guy but it seems to me that your essay explains the venture capital industry and it's and it's basically parallel distribution yeah I'll get to that your current a about that let me so I can finish this but but ok he wants me answer you can't answer Ukraine a as the first question alright that it explained that the but explain more than that the essay I'm trying to explain more than that I mean trying to explain how life works cooking is heuristic so we have two approaches cooking is I make a homeless perfect home how to make perfect homos you thinker and then add because they got nothing to lose by adding an ingredient giving it to Fat Tony he says yea or nay right if you say get lost alright you dump it all right if you know so you have gains can you make a ho moose out of the chemical you know a composition yes or no you can't all right ok the problem is so many things resemble homos so if you take in the world what resembles who moves versus what resembles the atomic project you'll notice it's huge compared to so small but then we focus on the other ones because science writers make us believe that science plays a larger role but in fact what we have is vastly more sophisticated technologies more sophisticated in science you see we get to venture capital I mean in a minute so this is my book for book 5 I explain here how you can measure fragility you know how to measure fragility measure how concave something is that's it take your portfolio or take a company I have a cult you have company all right take a company increase sales 10% decrease sales 10% if the company makes a hundred million and loses five million that you're fragile and that ratio is the exact mapping fragility okay and you can compare companies that way you say now the probability someone may tell me where you have mistakes and measuring I say who cares I'm not I'm second-order if I'm measuring a child and the child is you see I'm measuring the child and and my the child is growing but my ruler doesn't work I will still be able to tell how fast he's growing you say second-order effect you say I have you know at you can get VD and it's the acceleration that defines fragility technical but but but it works I can't measure the future I can measure facilities and future comparative fragilities okay the the now chapter book six I call them books because they're independent and they can be read each one's own and of course I goes through layers of that every depth every chapter has a topic and then you go through layers and layers and layers that's a start so book six is something called via negativa and in it a talk about medicine and a model medicine using convexity a shocking because very simply just as I said that a human body has to be concave two sources of harm alright it has to be treated let me give you the intuition this is complicated to do it now or maybe I don't have a clear idea how to express it in words but if you're away one Sigma from the norm say in hypertension and I the Nature has dealt with this problem and no medicine will outperform nature you see easily so someone who's slightly hypertensive has only one in fifty three chances in benefiting from a drug you see you could probably work out do something but not okay so this is what but when you're very hypertensive okay you have eighty percent chances benefiting from a drug which means that whatever we do in pharma okay is a bad bad competition was nature for cases that are around the normal but the very very very rare nature didn't encounter it so let's focus on the very rare till the following hit me uh-huh what hit me the the problem is that one Sigma all right is five times more frequent than four Sigma's 5,000 times not fact is it 5,000 times more people than okay if you're pharma who tries to cure the four Sigma they're going to die anyway all right you see or the one Sigma the 5,000 times more the one Sigma so therefore with all this all this that's where they find their clients to reclassify people who are 1 Sigma as pre something pre high cholestoral pre-hypertensive pre-hypertensive pre-diabetic pitch modulate all this all right ok so this is a so the chapter you know same complex same convexity same everything and this is the same graph now I'm going to show you this graph and is exactly the opposite then we talk about about venture capital this is a bank they make small money and then they lose big you agree what's the same thing when you think there's no drug we have taken in history that did not give us this at some point you take steroids you think an unconditional gain look what happen and when you're madly ill any treatment will have the shape when you're very ill is the opposite shape because you have a lot to gain if you have cancer anything you give me all right so there as a result we should over treat the very ill and we don't it's not good you know place for resources foreign what makes money on a module let's that's where our money goes smoothing out lives and you know and in it I talked about via negativa something very interesting if you like variation now replace temperature with calories 2,400 calories I was supposed to you know 4,000 calories one day and whatever the other day right 800 calories all right this is how we're made out to be the randomness of nature never gave us calorie steadily actually it's more complicated than that I discovered one thing when I started reading papers and medicine about seven years ago everybody in my family is diabetic all right everybody but me so I realize you know how can I pull out of it all right everyone my ancestors my grandfathers they all died younger than I was now except my grandfather so we realized that you know you have you know you understand how sort of like you worry about it then I realize the following what when people talk about diet and stuff like that they don't realize that the second order fad matters a lot more than first order effect in some cases you see it's not what you eat is HUF you know is the second order distribution just like temperature for the grandmother okay the average 70 degrees isn't as relevant as what was a variance around that average and if she spent half the time at zero degrees half-time 140 degrees information is vastly more important than the average you see second order swamps for first order can you hear me well can you all right so when you think of diabetes you realize that a lot of it comes from not over feeding but nobody but it was not studied there's very little Studies on something here called Jensen's inequality are you guys familiar with Jensen's inequality okay so take a function alright function alright the temperature the F is the health of a grandmother at zero degrees plus 140 degree divided by two all right or function at zero degrees plus function at 140 degrees divided by two there's a difference between two of them no it's called the expectation of an average is different from the average of the expectation it is a minor point but it captures the second-order effect it's pretty much the theme of my book is Jensen's inequality all right the problem is if you don't take that if you like randomness then there's a difference between the left and the right side you know if you hate randomness the difference is Big Ben the other opposite direction so it looks like we are made for randomness and feeding who understood it religions religions are studies okay when people talk about the Cretan diet okay they say okay let's eat by the Cretan seat you know what the Cretan seat very simple they have 200 days of fast I'm Greek Orthodox right up to 200 days of harrying vague and fast and then days were the only eat fat fat fatty lamb you see that's the important thing the other one is they don't eat three meals that you know that three meals a day that's a cereal industry we are anti fragile you know we need to restart think about it we are a combination of cow we are divorce we're like a cow at a time opportunistically and opportunistically meat eaters you agree okay now how often does a cow eat all the time very good says there's no randomness when you're getting herbs and whatever all this boring stuff you agree alright but how often does a lion eat meat rarely and then okay does he eat breakfast to go hunt no he has to get breakfast okay so their eyesight improves so that's Jensen's inequality is so we started looking with a bunch of friends and we start meeting in Brussels now for all places where they have french fries by the way okay to try to look at where there's something a metric we call regular divergence where people in medicine didn't get that point there's only one place in medicine that use that notion of Jensen's inequality as in pulmonary ventilators so I have a chapter explaining you know the this and to be quick before the my time is up the there's another thing that people don't understand that farmers never going to make money stressing your body by removing food from you making you eat episodically you know and putting you this six week at 600 calorie you know you lose diabetes there's a institutes in Siberia they've been learning for hundred years you say Pharma isn't gonna make money removing this one called via negative and a complex system if you remove something unnatural from it okay you have no long-term side effects but if you inject something new you don't know the chain of an unattended consequences this is what I called via negativa so this book is about philosophical about decision-making by the negative stop you from smoking you save it's better than any drug developed but nobody's going to make money if you stop you from smoking so the emphasis is on adding you see same with diabetes drugs stop people from eating breakfast dinner one meal a day like the Romans or the Cretans and make them follow some fasting protocol like religions tried to establish fasting so this is my book now the last book is on ethics it so happens that ethics can be met exactly as the bankers makes bonuses taxpayer pays for it some people have the upside other have the downside you agree well he is short an option that someone else has been harmed by and he's collecting the premium so you can look at ethics as I'm harming others and in my writing I adopt the following that there's a person I collaborate with at Max Blanc in Berlin a professor a gig writer and he said very simple if you ask a doctor 'we should be doing he gives you a different answer from if you ask him what he would be doing for you okay so the the the and it's very simple never I never never never tell you I should be going I tell you what I do I will never make a prediction unless I have something at harm it's called skin in the game you see if I'm harm I want to be harm first by my opinion okay so start looking at transfer of fragility you say why did the economic establishment continue to develop these bogus models because they're not harmed by the error someone else it's the error that makes their salary someone else it's the error you see so here we have three categories of people people calibrated ethically skilled in a game people who are not calibrated ethically they have the upside other people have doubts I'd like someone a bureaucrat who harmed you it makes a policy mistake you pay the price he doesn't okay this is why big centralized states don't have the the corrective mechanism municipality okay and then finally there's a category of heroes who have sold in the game firefighters the risks they have all the downside to give you the upside you say you get the upside they take all the downside and they don't get a bonus and no time in history have we had more people in power who don't come from the third category remember even one generation ago of presidents George Bush the father he was a war hero you know all right George Washington Hannibal was first in battle first one could be first okay Churchill was had okay he of course crossed the Atlantic traced by the Germans but he couldn't care less but you see the idea was the idea of skin in a game as a corrective mechanism and comes back to Hammurabi's law is that risk in the system increases there's no inspector who would ever know more than a person who has skin in the game so Fat Tony has a rule never get on a plane unless the pilot is on board and make sure there's a co-pilot that's that sort of thing and also never asked anyone for his to give his projection ask him what he has in this port for you so I made a lot of enemies with this and I will continue to make anybody also make friends the idea is ethical commitment has to be all the way through alright so that was the idea so I want to see NBC's in put it on sure they say you have to give us a prediction I say no I'll give you my portfolio without you know hiding the scale give you the percentages and then said ok no it doesn't work it's not sexy for you no they want people to sell fake projections so this is sort of like the book and now we can start the Q&A and and and we we start with this with the Entrepreneurship know so thank you for listening to me and so so I guess the first question was precisely about implication for your business okay along these lines was having the opposite pay off failing fast if you have optionality you'd rather have five options the biggest mistake is long term planning because it locks you up like a highway without exits so and mathematically five resetting options are worth a lot more than one long five-year option you see because you opportunistically have time to change you can you can accept or not so this is one of the conclusions that Ken Griffin no record has identified I'm sure it's the one he liked the most and there are other ones like that you know the other ones but if you phrase it I'm not contributing anything particularly intelligent to that debate other than framing it in the mathematics of this in a coherent mathematical framework for all of this you see and if you frame everything in a coherent mathematically mark that you can make immediate recommendation that if you have optionality then your gain from randomness okay then you have maximum exposure then a five-year option is less worse than five times one year option reset every year you see that don't have long term plans come to a counter intuitively and stuff like that so answer the question all right so another final thing you know for for entrepreneurship is that failing fast for you guys you understand the value of failing fast you see because you understand that if you have less harm that you okay this is the only place where people know how to fail fast I mean this is this part of the world so another question yes on your second chapter regarding evolution mention that like for example I think control to throw it up in kids into the pits adding intentional romans eliminate the weak weaker kids the weak child yeah what I'm saying is we yeah the evolution is harsh all right there's something called naturalistic fallacy that we have to fight we have to use all what's good about evolution right while imposing our moral standards and how do you do it no no I mean it was okay Darwinism okay you by what wanting to have a society that has no Darwinism we eliminated we just we weakened it further that's everybody's got a sink together the way you got to do it is protect the very weak and this is what I call the barbell approach I haven't spoken about it now so while you see barbell barbell approach is instead of having the middle class being favored favors a very very weak and let the wild entrepreneur prevail you see the middle class they want their comfortable thing you see so remove you kill you the the barbell approach is like for a portfolio okay heads your extreme risk you see has your extreme negative risk and take maximal positive risk so that's the barbell approach you can apply to society it's a little complicated to do it here but the barbell means instead of having a middle policy to smooth out fluctuation just try to smooth out fluctuations very unhappy another way to practice in medicine when I say I say I have more people in emergency rooms and fewer elective surgeries that's that's my form of you see my answer protect people from extreme unhappiness but don't give them prozac because they have mood swings you get the idea not yet yes him and then yes model or assistant is fragile so it's I guess it's easy in cases and questions about picking what the stressor is that you're looking at excellent so like yeah the case where you long an option it's easy to know but like if you're picking a company how do you know it something that's prevails or what okay there's the when I take a model I take a look at economic models it's very simple I take every model as it is people usually never miss a relevant variable to their model in fact they have too many you see what you did model error typically doesn't come from is identifying you know what enters your model like you have a company and then you have sales the weather in Spain this all these you have them in your model the problem is that making a point estimate of that variable rather than a variable estimate of that you get you see the idea like for example the big error you know that your grandmother in the my grandmother example is affected by the weather but at point estimate is you take the average temperature 70 degrees instead of stochastic sizing the average and this is the test for that I call it the Alpha the the sorry the Omega a the Omega a test is you take a model for example I have projections of sale ten years from now okay maybe I'm very good at it but the the random instead of doing that I test my model across different values of that if I'm concave to that then I have model error just as I said earlier very simple take yourself fertilizers all right you're very good at estimating orders for fertilizer of the next year so on average you're going to get ten million orders now you to do to see how vulnerable you are to model error you you try to see at eight million fertilizers or at 12 million fertilizers you say to see if you lose a lot more at eight million than you gain at twelve million you have a huge exposure to model error you say you take the variable model is composed of variables f of a1 a2 a3 a3 and you check every variable to see if you're concave to it as a first cut is there's no complicated method where you move them together but the way we did it for the IMF I wrote a paper heuristic to detect model error that was so simple that even the IMF people understood it all right and then the you know I took only a year to get stamp of approval to tell you how easy it was it has to be very easy that I work for free for the IMF for simple reason because they said oh we're forced to pay you and then they sent me 83 pages to sign and I resigned right away before starting so then we negotiated I don't earn a salary and they only had to sign a few sheets and stuff like that the IMF and then we wrote the paper by the time to put their approval all right it took about a year all right but the idea is very simple I take a bank no bank no missus the linear you see the variations right away you see so it has to be nonlinear so you take you take a bank you do a stress test I don't care how you do it all right because I tell you this important is not even a precision of your stress test it's the acceleration okay so I say okay the stock market is going to go down 30% how much do you lose okay then you do stock market going down 35% and going down 40% if you have accelerated losses you see it means that you have model error you say that's that's how we test for bottle error and then you test the anti fragile has opposite characteristics you make more and more and they say you make a lot more as the market goes up 20% but more than twice that the free market world goes up 10% you're anti fragile and then you have you know the sky's the limit yes sorry he was first event than you now unfortunately because of this book tour and because I retired as a trade I spent my life as a trader under one simple thing doing the opposite of this okay trying to survive by selling volatility and buying tails so it started in 87 so I had a big bonus from the market from 87 and I realized look at 87 that if you have a 20 sigma event okay 20 Sigma so that you can beat your so convex that you can hold and I wrote a paper it's on the site showing why it's so disproportionate that you can wait 400 years alright and and before having another one you still okay all right it's very convex so I told myself I only specialize in extreme tails so that was my profession I did that and of course you know you wait three four or five years everybody tells you're an idiot you're not profitable and so on if they okay all right and then it would disappear they blow up so that was that was my how I have built my portfolio now that I have it I'm lazy I don't want to lose it alright so I like Seneca I can't wake up in the morning till I'm in a shipwreck alright so I can't do that so I have inflation hedges eighty-five percent of my portfolio including land in Lebanon I'm trying to buy stuff in Canada some stocks stuff like that alright real estate you can land this is this is my portfolio I had gold and silver I'm decreased it after the crisis but now my portfolio doesn't correspond to how I built it so it's an unfair question you see I'm now in a retired person who writes books and doesn't want variation and I actually I do deal with God tell them 20 years from now give me back 90 percent of my money unharmed by inflation and I sign off right right away you see I give 10% to have that hedge and I can't I can't find you know there's no answer and I can't find anybody and I can't find anybody to give me you know that kind of package you see so this is my portfolio and then him yes yes it was very soon I got on twitter about two weeks ago I was popular so one of the things that was brought up recently and of architecture forum was a discussion of business strategy and how to find a top fertility and should reveal it now business rightly not investable strategy but you know what do we do with a product or what do we do with the market you know is the strategy itself fragile or entered final or is the system executing on the strategy and thereby improving it fragile or under very I'm gonna give you a simpler question see if you've noticed I have no theory so far you know I've produced no theory all that produced is measurements have you noticed okay we can do measurements if a business is fragile measurement describe a business what do they do for a living well so then we can just use Microsoft for fun because we're all here let's pick a particular area of Microsoft's business so the windows businesses so through over here okay so we have windows sales you have sales and you have costs all right now okay so you take sales in different divisions and you see what happens to your businesses tails go up by one mean deviation go down by one min deviation or two mean deviations to be more robust okay and then if the costs go up or down you see and you look at the net because you may not be able to keep your cost and see what happens to your to your business so you see how fragile you are I'm certain you are fragile for one's simple reason it can be very big pretty much that giving you sort of dominate the business when you grow the growth everything a nature has a shape by the way you know that all right everything in nature has a shape so you are fragile probably going down from here we can getting paid for the fragility the point is you want to know how much you are paid for the fragility it's a little more you see the everything in nature at the shape you see there's a maximum once you you're anti fragile to source and then at some point it tapers off and then becomes then you're now you're fragile you see because going back down you get the idea which is as a price of success that's Damocles sword so you want to manage it either by exiting the remain business and going into likewise or Google's doing going to go into areas that are not nobody's dominating you see so because I mean if you have search engine used by 80% of the people right you were to go to shoot for 85% no you want to go for you get the idea so you want to go for something that you know that you can dominate that's sort of the strategy but but you know you're fragile in business that you dominate that definitely you know it's the point this are you paid enough for it you see that's the point if you don't mind selling volatile if I'm paid for it you see the idea but I'm not I mean I'm not a business consultant but I would say if someone tells me how to build a business I say maximum optionality at all times maximum convexity at all times now you have minimum convexity but you paid for it because you dominate that business okay and you can lose it not because of a competitor but because the business may mutate into something else so this is why you know you get a keep seeding those thing kurung ventures you make it point it as the information content and the offering goes up this you talk about Bill Gates and he wins and everybody else loses or Apple wins and everybody else loses his information content goes up it gets more fragile or brittle in some sense yeah okay the overall environment is very fragile for one simple reason we're dominate with the power law distribution as one characteristic adjusted apart coming on a plane to relax from you have to realize this is my 21st day on the road was you know with the mails you get the idea or a journalist and this is the only relaxing lecture I've given other than the one I gave in the university all right so relaxing because you can get technical right away without the ball sorry okay so the the I just did very simple I took I had the book sales near some book says for all editions for some process them and it's always mind bogglingly shocking that you had 0.1% of the books represent 50 percent of the profits you get the idea that's a you know and they were moving into that environment we didn't have that before we didn't have the Harry Potter effect before Microsoft would have been impossible for some someone who drops out of college to reach set monopoly so short periods of time unless you had in an environment in which there was no connectivity you see environment I was not technological you see it was a BIM so we're entering the world where things can happen very fast either direction you see like the Google came you know their babies ten years ago you say so you have at the same time you have to take with it that necessarily this thing can shift to something else so when you look at books also it's quite shocking dick history and then you see books that dominated the planet disappeared completely like you go for two as yourself MFC looking at sales seven copies a week from something else selling like south fifty sixty million copies all right seven copy is shocking how things can can can shift the other way this to me is the mass equal equalizer but I don't want to depend on huge success you know you want to have a steady base how do you do it I don't know okay I mean you manage your own way I have no more time no more questions alright thank you for listening to me and I sing

Channel: Microsoft Research

Views: 56,716

Rating: 4.8987856 out of 5

Keywords: microsoft research

Id: BaU7Sxk6Yk4

Length: 65min 54sec (3954 seconds)

Published: Thu Sep 08 2016