What this covers
Nassim Taleb is trader, researcher and essayist. He is the author of the Incerto, a multi-volume philosophical and practical meditation on uncertainty.
*SUBSCRIBE TO THE PODCAST* ► https://www.youtube.com/@josephnoelwalker
*LISTEN ON:* WEBSITE: https://josephnoelwalker.com/nassim-taleb-158/ SPOTIFY: https://open.spotify.com/episode/2oVRyiU0AVMK5y5QfJSuqS?si=kq2Ak2gyTCemlJmVoOgfQw APPLE PODCASTS: https://podcasts.apple.com/au/podcast/the-jolly-swagman-podcast/id1236553683?i=1000670110437
*FOLLOW ME ON:* TWITTER: https://twitter.com/JosephNWalker
*TIMESTAMPS* (00:00:00) - Introduction. (00:00:55) - Heuristics for knowing when you're in Mediocristan versus Extremistan. (00:06:06) - Are certain tail exponents intrinsic? (00:08:31) - How long before 1987 did Nassim realise that option volatility shouldn't be flat across strike prices? (00:10:30) - Why hasn't Universa's tail hedging strategy now been fully priced in? (00:11:52) - Does the power law distribution of startup returns mean VCs should concentrate their bets, or spray and pray? (00:15:20) - Nassim's 30-minute take on the field of behavioural economics. (00:48:57) - Nassim's 20-minute take on superforecasting. (01:11:03) - The Precautionary Principle and AI. (01:17:28) - What are LLMs doing? (01:23:10) - War, violence, & "the empirical mean is not the real mean". (01:39:17) - Covid, & how Western governments think about tail risk. (01:43:38) - What's the most important thing people in social science get wrong about correlation? (01:52:58) - How does Nassim explain the perspicacity of the Russian school of probability? (01:55:56) - Why doesn't Hayek's knowledge argument extend to prediction markets? (01:58:26) - If mean absolute deviation is a better measure than standard deviation, why has the latter become commonplace? (02:01:09) - Nassim's next book, and what he's up to at the moment.
Source description (no synthesized summary yet).
Taleb argues that fat-tailed distributions and extreme value theory are fundamental to understanding real-world systems, and most intellectual fields (from behavioral economics to forecasting to policy) fail catastrophically because they misunderstand probability structure and confuse frequency-space predictions with payoff-space outcomes.
- Thin-tailed models systematically underestimate tail risk; you must assume fat tails unless you have robust mechanistic reasons to rule them out
- Binary forecasting and probability statements are meaningless without specifying the payoff function; the same probability can have vastly different values depending on what you're getting paid
- Most academic fields (behavioral economics, forecasting, social science) are built on Gaussian assumptions and misunderstand how to apply statistics to real-world fat-tailed phenomena
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
The fundamental flaw in super-forecasting methodology is that it evaluates predictors on binary outcomes (did event X occur or not?) rather than on payoff functions, and binary options become less valuable under fat-tailed distributions because fatness in tails causes probabilities of extreme binary outcomes to shrink while the body of the distribution expands, meaning frequency-space accuracy is uncorrelated with payoff-space success.
“I just about binary versus continuous payoffs yeah there's one thing with first of all the quality of the project aside from the and the discussions they didn't understand hour because I got bunch of people involved with me and the repli is uh our insults you see so the first no the first one is binary versus continuous okay and I knew that as an option Traer that the naive person would come in and think a an out of-the money binary option would go up in value when you fatten the tail in fact they go down in value when you fting the tail”
The central problem with behavioral economics is not that heuristics and biases exist but that the field misunderstands probability structure and confuses empirical observations of human decision-making with irrationality—most of what they label irrational is actually rational under fat tails or misrepresents how probability works.
“I tell them the risk is not the one you see see you have tail risks that don't show in your analysis I told Taylor Taylor say well assuming a gaan that my theory works I say I'm not assuming you know assuming the world were were coconut you know a lot of things would work so the world is not a goian and you're recommending that for 401k and stuff like that so then I noticed that's just you know first mistake in Sailor there are other mistakes in in that discipline uh like this idea of rationality yeah and to me rationality is in survival not in other things”
Superforecasters subselect for events they can forecast, which tend to be inconsequential and small-scale, whereas the consequential events (wars, pandemics) are characterized by scale-free distributions where there is no standard or typical event.
“and forecasting that the other one is they subselect events you can forecast because but they have they're inconsequential you see they're very small restricted question they're consequential so so so and also they events there's no such thing as an event like for example will it be a war yes or no I mean there can be a war could kill two people could be a war kill 600,000 people yeah so in extremist that's one thing one sentence mandal bro kept uh repeating to me there is no such thing as a standard deviation in extremist stand yeah you see so you can't judge the event you know by saying oh there's a pandemic or no pandemic because the the size is random variable”
In domains like price or contagion where there are no physical limits to the size of the variable and the process is multiplicative, you cannot rule out a thick-tail distribution and must assume fat tails, unlike bounded systems like height.
“if you don't know anything about the process or the process is in multiplicative uh concern multiplicative uh phenomena such as contagions pandemics or simply processes that don't have limit to their to their movement like for example a price you and I can sell you know buy from one another this at the a billion dollars okay there's no limitations there's no physical limitation m to the to a price therefore you could be an extremist stand and you cannot rule out a a thick tail distribution right”
The critical distinction in designing decision frameworks is whether a domain exhibits thin-tail or fat-tail distributions, because thin-tail distributions can be surprised by outliers that violate their assumptions, whereas fat-tail distributions incorporate large deviations within their statistical properties, making the latter epistemic class mandatory unless there are compelling physical reasons to rule it out.
“if I am using a thin tail probability distribution you say I can be always surprised by an outlier with respect to my distribution a large deviation you see that would destroy my assumption of using that distribution if on the other hand I'm using a large deviation model or model that the extremist stand model the reverse cannot be true nothing can surprise you a quiet period is entirely within statistical properties so is a large deviation which is why you have to assume that you're in the second class of models unless you have real reasons”
The most effective way to reduce tail risk in a contagion process is to reduce connectivity and movement (through travel restrictions and quarantines), which directly lowers the scale of the tail distribution by cutting off the distribution of contagion outcomes.
“when we start seeing what's happening in China realized that there was a problem and then I start looking at at at ways to how do you mitigate something that's fat tailed you lower the scale how do you lower the scale by by by cutting off the distribution of the parts MH okay reduced connectivity reduced connectivity and uh it's very strange that the Trump Administration did not they they I mean they spent all this money all right on giv money handing out money all of that it didn't hit him that that you're most effective by having uh controls at the border or you test people”
In social science, the metric correlation is widely misunderstood; a correlation of 0.5 is not halfway between zero and one because the effect size non-linearly scales with payoff; further, correlation is sub-additive in absolute terms, meaning you cannot sum correlations across subsamples to get the population correlation.
“what what what what what [1:49:03] they don't know what it means there I mean there are SGD people [1:43:50] who really think that um experts have a problem and and and they are good results there and uh they they they uh ask the people do the regression MH what what does it mean and they can't explain their own results”
Physical and biological systems have natural constraints that bound their tail behavior—for example, humans cannot be 500 kilometers tall due to biological limits—whereas financial prices and similar multiplicative processes have no such bounds, meaning finance is inherently extremist territory where Gaussian assumptions are categorically wrong.
“there are biological limitation the person needs to have a mother... you can't have unlimited energy so the you know that lot of mechanisms have these physical limitations right you see so you can rule out based on knowledge of the process biological understanding physical understanding but if you don't know anything about the process or the process is in multiplicative uh concern multiplicative uh phenomena such as contagions pandemics or simply processes that don't have limit to their to their movement like for example a price you and I can sell you know buy from one another this at the a billion dollars okay there's no limitations there's no physical limitation m to the to a price therefore you could be an extremist stand”
Standard deviation became the traditional measure of dispersion because Fisher proved it was more statistically efficient than mean absolute deviation under Gaussian assumptions, but this efficiency advantage does not hold under fat-tail conditions, yet the field persisted with standard deviation without accounting for domain-specific distributional assumptions.
“I think I discovered here a paper by Claim by fiser I think right who found that in the gassian case it's more efficient than uh mean absolute deviation yeah”
Option traders understand that payoff convexity, not probability, determines the value of an option; a trader can be bullish in frequency space (believing the market will go up) while being bearish in payoff space (holding short options) because tail moves that are unlikely frequency-wise can be highly valuable payoff-wise.
“so for example if I take the the gum curve plus or minus one Sigma is about 68% if I fatten the tail exiting in other words the probabilities of being above or below actually they drop you see why because you have the the the variance is more explained by rare events the the body distribution goes up yeah the shoulders n exactly you have more ordinary because you have high high inequality and the deviations that occur are much more pronounced right”
Value at Risk (VaR) is flawed because it describes the maximum expected loss within a confidence interval but ignores the conditional expected loss beyond that threshold; the truly dangerous region is in the tail beyond the VaR threshold, making VaR a false sense of security.
“uh we tried to explain it to that paper it didn't go through so now what we discovered also later on and this also applies to something what I call the varar Dilemma that people thought we were good at Value at risk and not good at C value at risk is saying okay you have within 95% confidence you won't lose more than a million MH and we thought that I thought it was flawed because that's not the right way because conditional on losing more than a million you may lose 200 right okay and so so so that remaining 5% is is where the action was”
Probability cannot be understood in isolation from a payoff function—probability density is a kernel that must be multiplied by a payoff function g(x) and integrated, and talking about probability alone has no meaning; for binary forecasts g(x) is an indicator function, but g(x) can be any continuous, convex, or concave function.
“so whatever is inside okay cannot be isolated it's a kernel okay you see it is a a thing that adds up to Wi like saying densities are not probabilities MH but they work well within a we even had at some point people using negative probabilities just like in quantum mechanics they use negative probabilities and and smart people understand that yeah you can use NE because it's a kernel okay the constraints are not on on on the inside say on the summation on the raw summation right so when you say but it is a kernel therefore what are these properties okay completely different so you should look at what you're doing with probability it's on by itself doesn't come alone so you're multiplying within an integral P of X with some function G of X yeah okay P of X by itself has no meaning yeah all right G of X all right has some meaning now if you're doing a binary G of X is an indicator function if x is of 100 zero or one whatever however you want to phrase it or it could be continuous could be convex could be concave could have lot lot of other shapes and then we can talk but talking about probability itself you can't yeah you can't separate P of X and talk about that by itself exactly you can't talk about that by yeah that's that's the whole point of it probability density function yeah density not probability yeah”
Mutual information (an entropy measure) is far more informative than correlation for decision-making; a correlation of 0 can coexist with infinite mutual information, showing the fundamental inadequacy of Pearson correlation for understanding dependence.
“and samplified when I I made a graph showing how it has this uh how visually you can see it you see you can Mutual information which is an entropy measure is V more informative that's in a linear world and now you go nonlinear visibly if you have a v curve zero correlation and infinite Mutual information”
Hayek's knowledge argument about distributed knowledge through prices applies to embedded, implicit knowledge that cannot be easily formalized; it does not support prediction markets because markets work on implicit signals, whereas forecasting binary events requires explicit probabilistic models, which systematizes knowledge in a way that often destroys the original insight.
“that for him knowledge is not explicit that is implicit the difference between uh knowledge that can be uh uh taught and and and formalized and knowledge that is embedded in society and that one Express itself from the manufacturing and then the end price and why a systematized economy your system izing something that is not uh explicitly uh LED itself to explicit phrasing is is is what uh what harmed uh the Soviet okay so I would not uh I I would not I would say that this applies to probably the wrong way for you that using a probabilistic model is trying to be systematic about things are too rich for you to express them systematically okay you see”
In extremist domains, there is no characteristic scale or typical event size, meaning that predicting whether a war will occur or not is sterile because the size of the war is a random variable; the statement 'pandemic will occur' is meaningless without specifying the death toll, since killing 2 people versus 2 million people are both technically pandemics but have radically different consequences.
“they subselect events you can forecast because but they have they're inconsequential you see they're very small restricted question... so so and also they events there's no such thing as an event like for example will it be a war yes or no I mean there can be a war could kill two people could be a war kill 600,000 people yeah so in extremist that's one thing one sentence mandal bro kept uh repeating to me there is no such thing as a standard deviation in extremist stand yeah you see so you can't judge the event you know by saying oh there's a pandemic or no pandemic because the the size is random variable”
The inter-arrival time for wars with death tolls above 10 million follows an exponential distribution (memoryless), meaning the historical waiting time of ~100 years provides no information about future risk—the expectation remains ~100 years whether or not a war has occurred in the past 100 years.
“the waiting time for Wars with deaths above a threshold of 10 million people is bit over 100 years yeah and that means that big because we haven't just because we haven't observed any like the last the last conflict with deaths of more than 10 million was World War II nearly 80 years ago now but we can't infer from that that violence is deing decline plus another thing that we discovered that's very robust is inter Ral time is has an exponential distribution like a p you know the inter Ral time of is uh it means it's memoryless right in other words if it arrives on average every say 100 years and then we haven't had one in 100 years okay you don't say oh it's coming it's memoryless so you wait wait another 100 years the expectations stay the same”
Every statistical problem requires adapted estimators suited to the specific distributional context; practitioners who only know Gaussian-based regression and hypothesis testing cannot move to non-standard problems without developing entirely new estimation theory.
“I have this theory that every single problem needs a new class of estimators adapted to the problem seems like a pretty good heuristic yeah so so if you don't know how to redo an estimator how to redo the theory yeah you see the only thing in common is a lot of large numbers that's it right”
The non-naive precautionary principle restricts itself to risks with reversible consequences or localized harm, and only applies to multiplicative/systemic processes that cannot be easily contained; nuclear power meets this criterion because reactors can be built small with localized consequences, whereas pandemics and certain biotechnologies create systemic, irreversible harm.
“for US nuclear was not precautionary why because you can have small little reactors and that one explodes in California doesn't impact one in bot the harm is localized exactly it's localized so unlike uh pandemics yeah”
Forecasting using binary point estimates for fat-tail variables is foolish because you cannot specify an event threshold that is scale-invariant—the forecast will be different for 10 million deaths versus 100 million, and the meaning changes as you move the scale.
“and I wrote this book statistical consequence of fat tales and this is why I gave the dedicated the Black Swan to metal Bron based on that idea that that the characteristics scale that I explained in in the Black Swan if you use that then you have a problem with forecasting you see because it is sterile in the sense that what comes above has a meaning see is it higher than 10 million higher than 100 million it has a meaning so this is where I've written another thing about forecasting a paper and I think we insulted the tetlock only because it's good to insult U people who uh do such work uh and also only insulted them because he spent like five years you know that's why I call him the rat someone stabing in your back so we explained that and I called it uh what do I call it on the U about a single forecast a single forecast Point forecast right on why one should never do Point forecast for a fat tail variable”
The non-naive precautionary principle should only apply to technologies with systemic, irreversible risks and multiplicative harms—not to all risks—and nuclear power is not precautionary in this sense because harm is localized, unlike pandemics or environmental disruption.
“I think that the the two things the the precaution Principle as understood and there's what we call the non-naive precaution principle has restrictions on what you got to have precaution yeah about because a lot of people couldn't get it that that like why are we so uh much against techn we're not against technology we're against some classes of uh uh engineering that have a reversal effect and it was a huge standard error and when I discussed on uh the podcast or the probability book whatever you want to call this were with Scott Patterson discussed the M story what what callus a great family fine was trying to get rid of sparrows sparrows yeah okay and then they killed all the sparrows or they tried to kill as many sparrows that they could and sparrows eat insects right so they had environmental problem with insects proliferating on the train and and then they didn't see it coming right now you say okay this is case that's clearcut of disrupting nature at a large scale you see and something we don't quite understand this is exactly what our our precaution is about except that we added multiplicative effects like we don't exerise precaution on nuclear this is why we're trying to the way I wanted our precaution principle to work is to tell you what is not precautionary and for US nuclear was not precautionary why because you can have small little reactors and that one explodes in California doesn't impact one in bot the harm is localized exactly it's localized so unlike uh pandemics yeah”
If you want to make money on a business or generate novel insights, you cannot use the logic that 'this makes sense,' because if it makes sense, many people have already tried it—the businesses that succeed are those that do NOT make sense according to conventional analysis.
“how do you make money in life how do you really improve how do you write a book how do okay think people didn't think about because if you're if you're going to start a business that makes sense guess what someone else thought about it okay I think something and and chpt is designed to tell you what makes sense based on current information right not look for exception there may be a possible modification I don't know to make sure GPT only tell you what make what makes no sense right and and that would hit one day but it's like our usual addage Universal is if if you have a reason to buy an option don't buy it yeah because other people will also have the same reason yeah okay yeah so it's the same thing with starting a business you're not going to make money on a business that makes sense yeah because a lot of people have tried it maybe some Pockets here and there people have tried it say so the the the idea of of CH GPT coming up was genuine insights okay is exactly in Reverse of the way it was modeled”
IQ-income/wealth correlations are so noisy and weak that in practice you cannot use IQ to predict individual income or hiring outcomes; the correlation is visible only at scale (thousands of people) but the practical signal-to-noise ratio is poor for individual decisions because IQ is thin-tailed while income is fat-tailed.
“so you know all you should hire or no because the the with such a weak correlation the law of large numbers okay doesn't start acting till you hire I don't know whole pound or something you see you get the idea so so you're you're getting noise right so that metric is noise unless you you're you have a wholesale yeah because of IND V individual variations yeah so within so so the way the law of large number works I explore it here even for thin Tales right it's misunderstood what I call the reverse uh law of large numbers if if you take a property say how much hypertension is going to be lowered by this medication right and reverse it and look at what are the odds of working on your patient you got a complete different answer from the one they think because on average it works say four points but but some people it's going to be a lot higher and so forth so this is where uh the interpretation of the statistical claims that they remain it can be messed up right”
Behavioral economics' claim that preferences violate transitivity (you prefer apples to pies, pies to bananas, but bananas to apples) is not evidence of irrationality, but rather evidence that nature evolved preference cycling to encourage dietary variation and reduce stress on the environment, making cyclic preferences evolutionarily rational and survival-optimal.
“I said no maybe that's not the way the world works if I always prefer apples to pie M and um presented that choice nature wants to make me eat other things and also to wants to uh reduce the stress on the environment of people always eating the same thing so it's a good way for nature to make me vary my my um my preferences either to protect nature and to protect myself right so it's not necessary uh you know that transitivity of preferences is not a necessary Criterion for rationality”
Before the 1987 stock market crash, Taleb observed high-frequency large deviations across stocks (mergers, bankruptcies, etc.) that violated Gaussian assumptions about volatility, and after the Plaza Accord's 10-sigma move in September 1985, he was convinced that option pricing needed to reflect tail convexity rather than flat volatility across strike prices.
“I saw deviations right and I realized and I had uh you know uh an unintentional reward from having a tail exposure so I realized that I said okay you don't have to be genius to figure out if the payoff can be so large right as to swamp the frequency so I think that I was pretty convinced by uh September 1985 after the plaza Accord had a 10 Sigma move at the time we didn't have access to data like today but you can I mean you but we saw prices and I noticed that effectively you had a um High um frequency of these very very large deviations across stocks”
Mandelbrot's insight that there is no characteristic scale in extremist systems was the key innovation that allowed Taleb to develop a framework for thinking about crashes (like 1987) and extreme events more generally.
“and just looking at the world from that standpoint that uh there's no characteristic scale changed uh my my my work better than a crash of 87 because now I had a framework that is very simple to refer to and they are probability basins so this is why I learned a lot uh you know working with metal bro and people weren't conscious of that uh Stark uh difference uh like operationally this is H I wrote this book statistical consequence of fat tales and this is why I gave the dedicated the Black Swan to metal Bron based on that idea that that the characteristics scale that I explained in in the Black Swan if you use that then you have a problem with forecasting you see”
The increase in destructive capability of modern weapons means that killing large numbers of people is now industrialized and easier, making the potential scale of violence greater, though Taleb did not want to make strong claims about whether violence has objectively increased.
“I mean I I I I looked at the data I reflected the data violence is about decline I did not put my concerns and my concern is that in the past to do what's being done in Gaza now required much more so we have a lot more destructive uh the ability I mean to kill is greater in the past uh you know would take a long time to kill you know so many people yeah do it manually yeah and now we industrialize the process which is very sad yes”
Venture capital compensation structure is not primarily based on whether portfolio companies become profitable or generate cash flow, but rather on hyping companies to attract new investors in subsequent funding rounds; therefore venture capitalists make money through a 'greater fool' dynamic by cashing in at each round, which explains why many extremely wealthy technology entrepreneurs have never actually generated positive cash flow in their businesses.
“okay the way uh uh you need to look at venture capital M is that it's largely a compensation scheme largely comp like hedge funds compensation scheme okay compensation to the 2 and 20 no no the the mechanism so they don't make their money Venture Capital they don't make money by waiting for the company to really become successful they make their money by hyping up an idea okay getting new investors and they cashing in as they're bringing in new investors which I mean it's plain look at how many extremely wealthy technology uh entrepreneurs are floating around while not having ever made a penny in uh that income”
The equity premium puzzle—the observation that historical equity returns are higher than bond returns in ways behavioral economists claim irrational—stems from misunderstanding probability structure: under open-ended fat-tail distributions, the same mathematical result holds as in prospect theory's convex-concave payoff function, making historical equity outperformance not anomalous but mathematically predicted.
“the equity premium bias comes from Equity premium the fact that uh people don't invest their explanations come from poor outstanding of property structure uh the aspect of PR theor that is wrong comes from misunderstanding prob structure that if you you have an open-ended distribution with fat tales then you have the same result”
The behavioral economics claim that people irrationally overweight small probabilities stems from testing in artificial lottery settings with known probability distributions, but in real-world fat-tail domains where people don't know the true distribution, people correctly weight tail events; the field's error is generalizing lab findings about lotteries to real-world decisions.
“I showed here because you never have a lump loss except for lotteries typically it's a variable and there's no such thing as a typical large deviation right see it it is technical but maybe your viewers uh will get it better was an explanation we'll get there next yeah and then I started looking at Stu done in behavioral economics such as the benan Sailor the benar and Taylor”
The Russian school of probability contributed more foundational results to probability theory than the French school, which contributed more than the English school, which instead developed regression and hypothesis testing—different disciplines entirely—with the English agenda often being to prove Irish or Ukrainian populations were inferior.
“they they I mean schools emerged when he start having norms and and groups of smart people together and there's a depth in a Russian approach to to mathematics but during the Soviet they had to make themselves useful because science had to contribute to society so they it can be remarkably practical while at the same time there's that constraint and I mean the the they were the a lot of it is French as well I mean when you look at the big results okay you always have a you know we have a combination of uh but I think the Russians have contributed the most of probability followed by of course uh the French and the of course English school of probability is just like gton and uh all these regression all these things that are bad come from this English school probability and usually they had agenda that goon wanted to prove that Irish were stupid measur Ukrainian right”
The specific tail exponent (alpha) that appears in financial markets, wars, pandemics, and other domains is often an emergent property of the underlying generating process rather than a free parameter; for example, the square-root market impact model plus power-law firm size distribution theoretically generates a cubic tail exponent, but Taleb disputes both the exponent value and the robustness of such theoretical predictions because empirical data is too noisy and domain-specific.
“there was a theory of why it was it's called the the the the the semic cubic Theory okay that he is following and it someone figured out that uh the tail exponent for U for company size start of companies was 1.5 M so therefore their orders are going to the market hence by using a I mean by by using a square root model of impact in other words where where the the quantity impacts the price following some kind of square root um effect okay then you end up with markets having a what they call the the the cubic from going from half cubic the cubic it is a nice Theory but I would uh I think uh the tail exponent in financial markets is lower than that from experience mhm and I don't like these cute theories because the distribution of um concentration is not 1.5 half cubic in with technology it's much much higher”
The inter-arrival time for wars with death tolls exceeding 10 million is approximately 100+ years, and this follows an exponential (memoryless) distribution, meaning that not having observed such a war for 80 years does NOT reduce the expected wait time—you cannot infer declining violence from the absence of recent large conflicts.
“one of the implications of your work on war with pasal is that that because of these inter arrival times we really should wait about 300 years without seeing a conflict of the scale of World War II yeah if you had to wasit 300 years then you'd say oh the the distribution has changed yes then we could say but we have had no information statistically from the past 80 years yeah and that was the the thing about the Pinker thinks that uh the world has changed and he couldn't understand our insults just like tlock they couldn't understand the statistical um uh claim yeah against that yeah so you think that I mean it's possible that the data generating processes could change it's just that we haven't seen anything that would overturn the null hypothesis exact that exactly that exactly the point”
Cass Sunstein's approach to risk (saying COVID risk is low relative to falling from ladders) is dangerous because it confuses additive and multiplicative processes—Sunstein mixed a multiplicative process (exponential disease spread) with additive casualty statistics.
“they were they were on it I started the war against that was before and when Co started um sunstein was advocating you know was was advocating ignoring covid because he said look how the risk are low he he he mixed a multiplicative process yeah with a uh additive one and by the way way now is you ask me to figure out the difference is for me to get you get fat tails via multiplicated processes MH”
People who make business decisions cannot have genuine belief in behavioral economics theories; if they truly believed behavior was irrational as defined by behavioral economists, they would make profitable business decisions by exploiting this irrationality, but the absence of wealthy behavioral economists suggests they do not believe their own theories.
“there I think that I mean we know a bunch of things that that that that that are part of that school but they're not Central to like for example how people react framing how people react based on how we present things to them how they a lot of these things work but whenever they make a general theory with a recommendation that connects to the real world they get it backwards right I mean back I mean I took sailor I told you sailor all his papers okay you interpret them backwards if he says okay you should have uh a concentration okay an optimal concentration of stock you go over one over end you saw my podcast with Danny Conan last year I did not see it I just read the segment where he said that that he accepted that uh I mean he said it publicly but he had told me privately yeah I agree doesn't work in yeah in under fat tals and I mean there's one thing about him I'm I'm certain that he knows everything that was said about him on Twitter okay I mean I'm saying he he's he does believe he should be up there I'm saying he's normal he himself would tell you I'm I'm I'm normal yeah I told him why did you write a book if you know that uh you have U uh you know a a a loss of version in other words one bad comment hurts you a lot more than a lot of Praise um he at me say I shouldn't have written a book”
Tail hedging strategies have not become fully priced into markets despite Taleb's books, Universa's performance, and Spitznagel's books because MBA education and Modern Portfolio Theory blind people to the observation, and institutional incentives mean traders managing other people's money must make frequent returns rather than hedge rare tail risks.
“because of uh than because of MBA uh uh lecturing mod portfolio Theory because people get blinded by theories and also because you uh if you're trading your own money you're going to be pretty rational about it if you're uh dealing with the institutional framework you need to make money frequently and the trap of needing to make money frequently will lead you to eventually sell volatility so there's no incentive to buy volatility for someone who's employed you know for fin night period of time in a firm no stive right”
Venture capital does not make money primarily by waiting for companies to succeed but by hyping ideas, bringing in new investors, and cashing out as they bring in subsequent rounds—a mechanism more akin to a Ponzi dynamic or selling hope rather than building cash-flowing businesses.
“the way uh uh you need to look at venture capital M is that it's largely a compensation scheme largely comp like hedge funds compensation scheme okay compensation to the 2 and 20 no no the the mechanism so they don't make their money Venture Capital they don't make money by waiting for the company to really become successful they make their money by hyping up an idea okay getting new investors and they cashing in as they're bringing in new investors which I mean it's plain look at how many extremely wealthy technology uh entrepreneurs are floating around while not having ever made a penny in uh that income see”
Conquerors and victims both exaggerated death counts in historical records—conquerors to appear more intimidating and victims to glorify their suffering (a pattern amplified by Christianity and Islam's valorization of victimhood), but statistical analysis using high-low randomization methods removes these biases.
“in a nonprobabilistic way I myself observed that a lot of people like to exaggerate their their killings yeah like gen ishan because it was optimal MH you know you don't have if people think that that you're going to kill a lot of people they won't oppose you so which is why you do a lot of uh stuff for sure yes a lot of Devastation for sure yes that makes sense victims exaggerating their suffering was less intuitive to me but then I remembered you know Tom Holland's work or Rene Gerard's work or even your treatment of Christianity and skin in the game I realized what makes Christianity unique is the valorization of the victim and Christianity and Sh Islam right the only the two religion yeah that uh that that have this um glorification of victimhood”
Most people in social science do not understand what correlation means; they confuse it with causation, do not realize that correlation of 0.5 is not halfway between 0 and 1 (it's much closer to 0), and misinterpret regression coefficients without understanding the underlying distributions.
“they don't know what it means there I mean there are SGD people who really think that um experts have a problem and and and they are good results there and uh they they uh ask the people do the regression MH what what does it mean and they can't explain their own results they know the equation they couldn't explain the graph how much this represents that so uh there there are a lot of incompetence and and social science and they use metrics they don't understand H people lot of people thought correlation was was an objective thing measure that depends on some sample and then has a very limited meaning and also they don't realize that when you do visually that correlation with 50 is not halfway between zero and one it's much closer to zero”
Multiplicative processes (like wealth accumulation through compound returns or contagion dynamics) generate fat-tail distributions because exponentiating a thin-tail base distribution creates fatter tails; specifically, exponentiating a Gaussian yields a log-normal distribution, and exponentiating a gamma or exponential distribution yields power-law distributions.
“if I take a uh G distribution and take the exponential of the variable you see because you know that that the the the log is additive right okay okay so when you multiply so you take the exponential you get a log normal distribution MH X okay logal distribution and and the MU and sigma L distribution are preserved right they're not the the mean and variance of the log normal there mean a variance of the log of the log normal okay it's misnamed should be the exponential but there was another name called exponential for another distribution”
For Gaussian distributions, the expected deviation above any threshold k-sigma shrinks as k increases (e.g., above 3-sigma is ~0.8 sigma, above 5-sigma is ~5.1 sigma), whereas for extremist distributions, the expected deviation stays proportionally constant and explosively scales in absolute terms, meaning old-age expectations become nonsensical in power-law domains.
“if I take a gaussian the expected deviation above 3 Sigma is a little more than three sigma and if you take five Sigma it's a little more than five Sigma right it gets smaller it's above zero Sigma is about 0 eight of a sigma as you go higher it get it shrinks it's like for saying what's your life expectancy at zero is 80 years old but at 100 is two years two is 80 two additional years so as you increase a random variable so whereas an extremist time the scale stay the same yeah so the expected uh life if we were uh distributed like company size uh the SI the the expected company you know as I said what's the expected company higher than uh was 10 million in sales 15 milon million 100 million sales 150 million the average uh two billion in sales 3 billion”
The ratio of standard deviation to mean absolute deviation is the best indicator of distribution fatness; for a Gaussian it is approximately 1.25, but for power-law distributions with alpha below 2 it becomes infinite, providing a simple diagnostic for detecting tail behavior.
“and there's a wedge both of the fat tals the way I'm interested the measure not because of you know to pick on on practitioners who make mistakes but but because the ratio stand deviation me deviation is uh the best indicat of fatness yeah see yeah no so there and there's a wedge both of the fat tals the way I'm interested the measure not because of you know to pick on on practitioners who make mistakes but but because the ratio stand deviation me deviation is uh the best indicat of fatness”
Probability is not an independent object but a kernel—a function that must integrate to one and cannot be meaningfully discussed in isolation from the payoff function; practitioners who treat probability and payoff separately fundamentally misunderstand probability theory.
“how do I know if someone understands probability they understand probability if they know that probability is not a product it's a kernel is it's something that add up to one all right so so whatever is inside okay cannot be isolated it's a kernel okay you see it is a a a thing that adds up to Wi like saying densities are not probabilities MH but they work well within a we even had at some point people using negative probabilities just like in quantum mechanics they use negative probabilities and and smart people understand that yeah you can use NE because it's a kernel okay the constraints are not on on on the inside say on the summation on the raw summation right so when you say but it is a kernel therefore what are these properties okay completely different”
Mutual information is a better metric than correlation for quantifying how much information one variable provides about another, particularly in nonlinear settings; a variable with zero correlation can have infinite mutual information.
“you can Mutual information which is an entropy measure is V more informative that's in a linear world and now you go nonlinear visibly if you have a v curve zero correlation and infinite Mutual information”
In Gaussian distributions, the expected deviation above 3 sigma is slightly more than 3 sigma, and as you move to higher sigmas the expected additional deviation shrinks, but in extremist time the scale stays the same—the expected life expectancy of a Gaussian-distributed entity at 100 years old is about 2 additional years, but for a Pareto distribution it scales the same way.
“if I take a gaussian the expected deviation above 3 Sigma is a little more than three sigma and if you take five Sigma it's a little more than five Sigma right it gets smaller it's above zero Sigma is about 0 eight of a sigma as you go higher it get it shrinks it's like for saying what's your life expectancy at zero is 80 years old but at 100 is two years two is 80 two additional years so as you increase a random variable so whereas an extremist time the scale stay the same yeah so the expected uh life if we were uh distributed like company size uh the SI the the expected company you know as I said what's the expected company higher than uh was 10 million in sales 15 milon million 100 million sales 150 million the average uh two billion in sales 3 billion all right so it's a same like saying oh he's 100 years old he has another 50 to go how many he's a thousand years old another 500 to go okay you can't apply the same reasoning with humans we know what an old person is okay uh because as you raise that number things shrink for extremist stand as you raise that number things don't shrink as a matter of effect and proportionally they stay the same but but in absolute they they explode okay”
When you exponentiate a Gaussian distribution, you get a log-normal; when you exponentiate a Gamma or exponential, you get a Pareto; when you exponentiate a Pareto, you get a log-Pareto—each transformation moves the distribution to fatter tails.
“so if I take a uh G distribution and take the exponential of the variable you see because you know that that the the the log is additive right okay okay so when you multiply so you take the exponential you get a log normal distribution MH X okay logal distribution and and the MU and sigma L distribution are preserved right they're not the the mean and variance of the log normal there mean a variance of the log of the log normal okay it's misnamed should be the exponential but there was another name called exponential for another distribution Okay so gaussian you exponentiate you get log normal now there's a distribution is thin tailed but slightly F tail than gaan barely right the exponential the gamma you know that class okay you exponentiate what do you get a power law you see so you're very so which one you're exponentiating your base distri ution needs to be gaussian for it to end with a l normal right or fatter tail than gaussian okay and the next class is a gamma or you know the uh the exponential and you get a parto right yeah and then of course there's a exponential of a parto it's called log parto okay”
Standard deviation is often confused with mean absolute deviation; standard deviation is the square root of the average of squared deviations and lacks intuitive physical meaning, whereas mean absolute deviation is the average deviation and is more interpretable.
“so I know that generally you prefer mean absolute deviation to standard deviation why has standard deviation become such a traditional measure like historically how did that happen Okay because uh I think I discovered here a paper by Claim by fiser I think right who found that in the gassian case it's more efficient than uh mean absolute deviation yeah uh yeah because again I mean to tell the viewers a lot of people mistake one for the other standard deviation is the square root of the sum of the average some squares it's not you know so it doesn't have a physical intuition yeah is what a standard deviation is what is med is the average so for example if you have the process right with all the observation at zero and uh and one observation at a million for an average of a million this the standard deviation be 500 times being deviation right”
Probability matching—where people allocate resources proportionally to observed success frequencies rather than concentrating on the highest-frequency option—is often labeled irrational but is actually optimal under entropy-maximizing (Kelly-style) models.
“there's another one about probability matching where you think that probability matching is irrational probability matching means that if something comes up 40% of the time and something comes up 60% of the time that you should invest 100% of the time in the higher frequency want but in nature and in you know animals but but also humans do probably mhing and when you write the math using uh entropy like uh you know in this U Kelly uh style modeling uh you if I have 10 horses and I got allocate among the time horses if I maximize want to maximize the expected return how do I allocate and proportion probably of of when”
If you are using a thin-tail probability distribution, you can always be surprised by an outlier with respect to your distribution, but if you are using a fat-tail or extremist model, the reverse cannot be true—nothing can surprise you because a quiet period is entirely within statistical properties and so is a large deviation.
“if I am using a thin tail probability distribution you say I can be always surprised by an out lier MH with respect to my distribution a large deviation... if on the other hand I'm using a large deviation model or model that the extremist stand model the reverse cannot be true nothing can surprise you a quiet period is entirely within statistical properties so is a large deviation”
AI does not currently warrant precautionary principle application because it lacks the ability to impose systemic, irreversible harm; AI has not demonstrated recursive self-improvement capability, and even if it did, computers can be shut down, isolating the damage from general-purpose systems.
“I have problems with discussing AI in terms of precaution because I don't medly see anything about AI why you should stop AI that it will self- reproduce given a robot cannot climb stairs so you're afraid now know you're afraid of robots scared of robots multiplying and becoming a robot Colony that would take over the world I mean these things are stres of imagination we have bigger problems to worry about”
Cass Sunstein is intellectually dangerous because he advised ignoring COVID-19 risk by comparing it to the low empirical risk of falling from ladders, fundamentally confusing multiplicative processes (contagion) with additive ones (individual accidents), leading to policy recommendations that conflated tail probability with tail risk.
“we had C sunstein who to me is about as dangerous as you can get okay what I called actually I wrote iyi the intellectual yet idiot based on him and uh and Taylor right... sunstein was advocating you know was was advocating ignoring covid because he said look how the risk are low he he he mixed a multiplicative process yeah with a uh additive one”
Regression studies claiming IQ predicts income suffer from circular reasoning: they define IQ as performance on exams, then show that exam performance predicts education and income, conflating measurement instrument with causal variable; the empirical cloud shows too much noise for individual prediction despite statistical significance in aggregate.
“with the circularity in fact that if you have if you're good at taking exams you're going to have a high IQ but you're also going to get a good college degree and that helps your income in the beginning right we're not talking about wealth or stuff so it's for employees so even taking all of these you look at the cloud and say well you know what you can't use it for any individual hiring”
Governments have not improved at thinking about tail risk; instead efforts to manage risk have increased risk by introducing complexity-creating bureaucrats who misunderstand probability and prevent effective tail-risk mitigation.
“do you sense that governments and policy makers saying the US have gotten any better at thinking about how to deal with tail risk no I think if anything their effort to deal with risk increase their risk because you end up with people like C sunstein and these uh Pizer are call them they make you stupid for worrying about thing because their textbook tells you they shouldn't worry about it and they don't understand fat tales once you understand fat tailes Things become very easy start thinking differently about AI differently about other things”
Taleb wrote a memo with Yanir Marin in late January 2020 concerned about pandemic risk, and it was shared with someone in the White House, but the administration's focus was on monetary stimulus rather than border controls, testing, and quarantine—mechanisms that would actually reduce tail risk.
“so in late January 2020 you wrote a memo with yanir a mutual friend yeah it started uh yeah and I mean we yanir and I were concerned about ebola before that yes back in 2014 yeah we we were obsessing over pandemic because I wrote on the Black Swan yeah and it was picked up by a bunch of people in Singapore so we were like all concerned about you know the big pandemic because it would travel faster than the Great Plague yeah so this is why we were very concerned when it started and you wanted to invite people to kill it in the egg yes and you wrote this memo which was then shared with a friend in the White House yeah can you tell me the story of that and is there anything you can share that you haven't shared publicly before no no it's that doesn't the the the the the paper by itself is meaningless because we would have written one and you're and I uh separate but the there was no no particular um novelty to that idea sure but when we start seeing what's happening in China realized that there was a problem and then I start looking at at at ways to how do you mitigate something that's fat tailed you lower the scale how do you lower the scale by by by cutting off the distribution of the parts MH okay reduced connectivity reduced connectivity and uh it's very strange that the Trump Administration did not they they I mean they spent all this money all right on giv money handing out money all of that it didn't hit him that that you're most effective by having uh controls at the border or you test people I mean that in the past we used to have very effective lazarettos where people were uh confin the quarantines and and now we can do it more effectively was testing”
Hayek's knowledge argument about prices aggregating distributed information applies to implicit/embedded knowledge, not explicit formalized knowledge; trying to systematize what is not explicitly systematizable (as the Soviet approach did) defeats the principle.
“no the difference idea is that no it's more it's explicit versus implicit that for him knowledge is not explicit that is implicit the difference between uh knowledge that can be uh uh taught and and and formalized and knowledge that is embedded in society and that one Express itself from the manufacturing and then the end price and why a systematized economy your system izing something that is not uh explicitly uh LED itself to explicit phrasing is is is what uh what harmed uh the Soviet okay so I would not uh I I would not I would say that this applies to probably the wrong way for you that using a probabilistic model is trying to be systematic about things are too rich for you to express them systematically okay you see so in other words his knowledge is what's embedded in society not what is formalized”
The United States State Department has been notably incompetent in foreign policy (evidenced by Afghanistan, failed predictions about Vietnam, inadvertent support of Bin Laden creating subsequent adversaries), and an interventionist foreign policy driven by democracy promotion is more dangerous than isolationism.
“I've started branching out not to foreign policy realizing that that effectively there's some things in that sjd Society of judgment decision making when they analyze the Vietnam War and and there are lot of lot of good things in that that uh in that industry and uh all the biases you realize that we have the United States the most dyn country very vital was completely incompetent State Department so you realize the decision for War I mean think of Afghanistan how naive it is not to figure out what's going on so it's going to make mistakes of course more mistakes of course and these alliances like you back up not understanding consequences so sort of like M sparrows you back up Bin Laden not realizing that that you help and let you build a machine that will turn against you right it's like the Hydra like the Hydra you cut off right back no no but but they create so a interventionist foreign policy yeah on the part of the United States and then over spreading democracy or stuff like that is actually more dangerous than just isolationism”
The Russian school of probability produced more foundational results than the French or English schools; the English school (Fisher, regression, hypothesis testing) introduced problematic methods driven by agendas (Galton's race science), whereas the Russian school maintained rigor.
“so no they they I mean schools emerged when he start having norms and and groups of smart people together and there's a depth in a Russian approach to to mathematics but during the Soviet they had to make themselves useful because science had to contribute to society so they it can be remarkably practical while at the same time there's that constraint and I mean the the they were the a lot of it is French as well I mean when you look at the big results okay you always have a you know we have a combination of uh but I think the Russians have contributed the most of probability followed by of course uh the French and the of course English school of probability is just like gton and uh all these regression all these things that are bad come from this English school probability and usually they had agenda that goon wanted to prove that Irish were stupid measur Ukrainian right and the linear regression the hypothesis testing the Fisher thing all that all these are completely different from yeah the”
The loss aversion finding (asymmetric value function in prospect theory where losses loom larger than gains) is one of the few robust findings in behavioral economics, and Taleb has consistently supported this despite critiquing most of the field.
“what results in behavioral economics do you think are robust we've spoken about the loss do they call the asymmetric loss function and Prospect is there anything else because uh no no nothing else”
The tail exponent (alpha) converges to its true value much faster and with lower standard error than the mean in fat-tail distributions; empirically, alpha estimates stabilize within a few observations while mean estimates scatter widely.
“if you get it if the process is clean yeah okay you have a a it's remarkable how quickly you get the alpha yeah I show you at Ry reverse try to get the means all over the map yeah you get the alpha always within like yeah it's really NE it's really neat yeah standard ER on the alpha is low yeah that on mean is huge yeah yeah”
By September 1985, after the Plaza Accord had produced a 10 sigma move in currency markets, Taleb was already convinced that volatility should not be flat across strike prices for options because large payoffs can swamp frequency, contradicting the Black-Scholes model.
“I said okay you don't have to be genius to figure out if the payoff can be so large right as to swamp the frequency so I think that I was pretty convinced by uh September 1985 after the plaza Accord had a 10 Sigma move at the time we didn't have access to data like today but you can I mean you but we saw prices and I noticed that effectively you had a um High um frequency of these very very large deviations across stocks”
Transitivity of preferences (if I prefer A to B and B to C, I should prefer A to C) is not a necessary criterion for rationality; nature may randomize preferences to protect itself and reduce stress on the environment, making preference cycles optimal from a survival standpoint.
“the uh people use as a metric and and and it was not contested the transitivity of preferences that I prefer apples to pies pies to uh say uh bananas all right okay but then bananas to Apples right so you're violating transitivity of preferences but I said no maybe that's not the way the world works if I always prefer apples to pie M and um presented that choice nature wants to make me eat other things and also to wants to uh reduce the stress on the environment of people always eating the same thing so it's a good way for nature to make me vary my my um my preferences either to protect nature and to protect myself right so it's not necessary uh you know that transitivity of preferences is not a necessary Criterion for rationality”
Intertemporal preference reversals (preferring one massage today over two tomorrow, but reversing this preference when both are 364+ days in the future) are not necessarily irrational; they reflect different probability distributions or the counterparty risk that a promise 365 days out may not be fulfilled.
“if I uh if I tell you do you want to Mass today or two massages tomorrow we like to say okay two massages tomorrow or let's assume that that you when facing this choice you take the two massages tomorrow that not one today but if I tell you in 364 days a choice of One Versus two you say you would reverse no you're possibly actually let's say that that you have it the other way that you take one the one today rather than two tomorrow but you reverse that's not if you use a different probability distribution different preference structure right plus there is another one that what I mean how do you know the guy the the person offering you that bet will satisfy tomorrow you see as I see the bird in a hand is better than right then some attract one in the future on some tree”
Pandemic death distributions follow a power-law with tail exponent approximately 0.5 (similar to Lévy distributions), meaning catastrophic pandemics dominate the expected deaths even though they are rare, and the empirical 95% of observations fall below the mean (counterintuitive for thin-tail thinkers but necessary for fat-tail systems).
“and we did the data we published the natureal scci natural physics paper on distribution of uh of uh people killed in pandemic and guess what a tail exponent is it's like less than one isn't it it's half yeah Less in on like the levy infinite man yeah the the it is actually slipped not infinite being some transform becomes infinite being but that is the same with Wars right because you can't kill than kill more than a population but it tracks for a large part of it and if you do a lock trans for then it's very robust anyway”
Concerns about AI risk from recursive self-improvement are speculative and disproportionate compared to actual problems; most people worried about AI risk have simply not thought clearly about what AI can and cannot do, and constraints on AI research should not be imposed preemptively but only after demonstrating systemic risk.
“I I have problems with discussing AI in terms of precaution because I don't medly see anything about AI why you should stop AI that it will self- reproduce given a robot cannot climb stairs so you're afraid now know you're afraid of robots scared of robots multiplying and becoming a robot Colony that would take over the world I mean these things are stres of imagination we have bigger problems to worry about I don't think most people who think about AI risk V robotics is a constraint so what is what is because technology would would the whole thing would become risky as technology becomes autonomous right so in other words uh that's my my understanding that that's that's what they're worried about okay and and becomes autonomous it has to first of all you can shut down your computer all right and and it no longer impacts our our our our life here you know can't hit the water because it's down the computer”
LLMs make probabilistic errors by conflating correlation with causation due to co-occurrence in training data; for example, when asked about Constantine Kodori (a Greek name) at the Congress of Berlin, ChatGPT hallucinated that he represented Greece because Greek-named individuals are more probabilistically associated with Greek representation, when he actually represented the Ottoman Empire.
“it doesn't have to connect the pieces directly so use probalistic methodss so reflect of consensus so you I asked it uh during the Congress of Berlin there was a war between the otoman Empire on one hand and then you had Greece on the other hand among other allies and there was a fellow uh K theodori who the father of the mathematician Cara theodori um who was representing you know someone there who did he represent they say oh he's a foreign affairs minister of Greece you see it's not like a search engine giving you facts it is using probabilistically how things occur uh together he has a Greek name therefore he in fact he was representing the other side the opan Empire MH”
The reason tail hedging strategies (buying out-of-the-money options) have not been fully priced into markets despite decades of published research is that institutional investors face career incentives to make money frequently—requiring them to sell volatility rather than buy it—whereas traders using their own capital can afford to hold tail hedges because they are evaluated on long-term returns, not quarterly performance.
“because of uh than because of MBA uh uh lecturing mod portfolio Theory because people get blinded by theories and also because you uh if you're trading your own money you're going to be pretty rational about it if you're uh dealing with the institutional framework you need to make money frequently and the trap of needing to make money frequently will lead you to eventually sell volatility so there's no incentive to buy volatility for someone who's employed you know for fin night period of time in a firm no stive right”
Every statistical problem needs a new class of estimators adapted to the problem; you cannot make general claims using Gaussian-derived tools, and this principle should guide statistical methodology rather than trying to force all problems into standardized frameworks.
“I have this theory that every single problem needs a new class of estimators adapted to the problem seems like a pretty good heuristic yeah so so if you don't know how to redo an estimator how to redo the theory yeah you see the only thing in common is a lot of large numbers that's it right and you want to know what it applies to so when you ask me something about the alpha the law of large numbers sometimes works a lot better for the alpha than that's for the mean yeah because the the um the T exponents follow with int distribution right it follows an inverse gamma distribution and you you get it it's the process is a specific type of th”
The United States has become more dangerous because it has greater capacity to cause harm while remaining strategically incompetent, exemplified by interventionist foreign policies (Afghanistan, Iraq) that fail to anticipate consequences and create blowback (e.g., supporting Bin Laden, who later attacked the US).
“you realize the decision for War I mean think of Afghanistan how naive it is not to figure out what's going on so it's going to make mistakes of course more mistakes of course and these alliances like you back up not understanding consequences so sort of like M sparrows you back up Bin Laden not realizing that that you help and let you build a machine that will turn against you right it's like the Hydra like the Hydra you cut off”
The shadow mean is the true mean of the data-generating process, which is often substantially higher than the empirical sample mean for fat-tail distributions; for warfare, the shadow mean is approximately three times higher than what historical records show.
“let's take a one tail distribution okay you have the visibly in a sample of 30 observation you're not going to get events that happen less than 1% of the time you agree yes so for a gaussian it's not a big deal because these that happen and less than 1% of the time have less impact on the probability gets increasingly smaller so it doesn't matter much so with a small sample it don't have a big Shadow mean effect actually with a gaussian it has to be a one tailed gaussian so so low variance like normal right like height okay yeah so you observe a bunch of people and you have an idea what what the average height in town is okay now when we talk about things that are open-ended and fat tailed okay visibly most observation will be below the mean MH so when you compute the mean it's going to be biased down yeah from uh empir what they call empirical observation yeah so the empirical distribution is not empirical and that's what is central for us yeah... if you're selling volatility you have a shadow mean that's going to be way lower than your your observed mean if you're talking for Wars even without surviv sh bias which is another story we have a uh the process that vly ner than what we observed about three times NOA okay three time next year yes so in other words um the the historical process underestimates the true process”
The reverse law of large numbers states that when you average individual outcomes in thin-tailed domains, aggregated effects differ dramatically from individual effects; a medication might lower hypertension by 4 points on average, but the distribution of individual effects may be wide, making population-level claims inapplicable to individuals.
“what I call the reverse uh law of large numbers if if you take a property say how much hypertension is going to be lowered by this medication right and reverse it and look at what are the odds of working on your patient you got a complete different answer from the one they think because on average it works say four points but but some people it's going to be a lot higher and so forth so this is where uh the the interpretation of the statistical claims that they remain it can be messed up right”
Correlation is not an objective measure but depends on the sample; the real issue in decision-making is the relationship between changes in two variables under your observation, and understanding this requires accounting for the payoff function—the Kelly Criterion or entropy-maximizing frameworks show that low correlations are mostly noise and 0.5 is nowhere near halfway between 0 and 1 in terms of utility.
“and also they don't realize that when you do visually that correlation with 50 is not halfway between zero and one it's much closer to zero you have this saying so people are familiar with the phrase correlation isn't causation you have this phrase correlation is in correlation exactly I had a lot of Twitter fights with with with people and that was fun because I didn't know that people didn't think of correlation that way the uh that's another thing if you look in in in finance naively you see that uh the effect of the mean of correlation appears to be like uh say X and Y are correlated your your expectation of Delta X is going to be row you know uh Sigma X over Sigma y right based on uh time Delta y right you're linking You observe the effect of X based on observation of Y but for betting and decision making it's not that it's more like uh a a fa Factor uh that's uh something like row Square 1us row square or like similar to the minus log 1us row Square uh so in other words very nonlinear in other words low correlations are are noise and uh and again 50 is not halfway between zero and one right and and one is infinity that's for decision making as and you put that into your your uh either Kelly Criterion or any form of decision making”
Intertemporal choice reversals (preferring one massage today vs. two tomorrow, but reversing when asked about the same choice 364+ days out) are not evidence of irrationality; they reflect different probability distributions or preference structures—including legitimate uncertainty about whether the other person will still be around or solvent, invoking the 'bird in hand' principle.
“if I uh if I tell you do you want to Mass today or two massages tomorrow we like to say okay two massages tomorrow or let's assume that that you when facing this choice you take the two massages tomorrow that not one today but if I tell you in 364 days a choice of One Versus two you say you would reverse no you're possibly actually let's say that that you have it the other way that you take one the one today rather than two tomorrow but you reverse that's not if you use a different probability distribution different preference structure right plus there is another one that what I mean how do you know the guy the the person offering you that bet will satisfy tomorrow you see as I see the bird in a hand is better than right then some attract one in the future on some tree okay”
To make money in business, you must find something that doesn't make sense; if you're starting a business based on logic that makes sense, others have already tried it and you'll face competition; therefore, you should seek out non-sensical-seeming opportunities that others have overlooked.
“how do you write a book how do you okay think people didn't think about because if you're if you're going to start a business that makes sense guess what someone else thought about it okay I think something and and chpt is designed to tell you what makes sense based on current information right not look for exception there may be a possible modification I don't know to make sure GPT only tell you what make what makes no sense right and and that would hit one day but it's like our usual addage Universal is if if you have a reason to buy an option don't buy it yeah because other people will also have the same reason”
Large language models cannot produce original scientific insights by design because they work through probability matching—reflecting what 'makes sense' statistically from training data—rather than seeking errors or exceptions, making them incapable of the contrarian thinking that generates new scientific knowledge.
“it works by probably matching by the way right it doesn't give you the same answer all the time and it's not going to do all the homework so it it it doesn't have to connect the pieces directly so use probalistic methodss so reflect of consensus”
Mental accounting—treating different sources of money differently depending on context (e.g., money won at a casino is 'house money' and is spent more freely)—is rational behavior with proper frequency analysis; Thaler's critique that this violates accounting standards misses that accounting standards don't apply to portfolio strategy and that the frequency of birthdays justifies the mental framing.
“husband and wife wife have a joint check-in account the husband visits the department store sees a tie doesn't buy it it's too expensive goes home and then see this gift and got all excited that he got it from his wife for his birthday okay so you know that mental counting is uh irrational I say okay but how many birthdays do you have a year okay yeah so it's it's not frequent so you know so this is where you know you got to put some structure around the mental accounting another mistake he makes the not St mistake the mistake is H that it's irrational when you go to a casino to increase your uh your betting when you win money from a casino”
Taleb wrote Fooled by Randomness with no references initially, then after meeting Danny Kahneman in 2002, he read about 100 psychology books in six months, found the behavioral economics math trivial and wrong, and retrofitted references to papers he had already written, explicitly acknowledging this was not fully honorable but was done to link his ideas to the discipline.
“I wrote full by Randomness and became very successful at the first edition uh and it had no references and I it had no behavioral just like aside from how hum don't understand probability minimal of that then I met Danny conoman in 2002 in Italy in Italy and uh and then okay I spoke to him he said you don't have a lot of references for stuff like that a lot of comments so I said no problem so I went and I got about 100 books in Psychology I read it over a period of say six months okay went through the Corpus everything figured out you know they they think that their math is complex their math is Trivial and wrong and then I cited okay and I modeled the I remodeled the uh Prospect the Theory... and I start putting references on sentences I've written before not knowing anything about it which was not the most honor thing but it was uh to link my ideas to to that discipline right”
Behavioral economists like Cass Sunstein represent 'intellectual yet idiot' thinking—people with credentials and abstract models who misunderstand probability and make dangerous policy recommendations, and their involvement in COVID response made the public less capable of dealing with tail risk.
“I think if anything their effort to deal with risk increase their risk because you end up with people like C sunstein and these uh Pizer are call them they make you stupid for worrying about thing because their textbook tells you they shouldn't worry about it and they don't understand fat tales once you understand fat tailes Things become very easy start thinking differently about AI differently about other things see you tell me I tell you yeah once AI stops multiplying let me know right and stuff like that this is my department fat tailes and precaution requires fat tales yeah I mean you could have precaution at different levels but the one we're concerned was at a higher micro level with price fat tails do we need any new social institutions to better deal with fat tails I have no idea okay at this point I'm too disgusted with these uh bureaucrats”
Value-at-Risk is flawed not only because it ignores what happens in the tail beyond the confidence interval, but also because the probability distribution itself, when transformed from a continuous distribution to a bounded [0,1] space, becomes convex or concave, making small forecasting errors explosive.
“value at risk is saying okay you have within 95% confidence you won't lose more than a million MH and we thought that I thought it was flawed because that's not the right way because conditional on on losing more than a million you may lose 200 right okay... but someone pointed out... my application of of the exponential transformation also applies for value at risk because he said if you want to get the the probability you know the probability is distributed in cailes MH because but it's pounded between zero and one exactly it isail right okay it's a frequency it's like this is why they have R score all that thing yeah but then the transformation of that probability okay outside the gaussian okay you have what you have you have the inverse uh Pro you see you want to go from a probability to the to X”
If you have a convex function with a break-even point, you make money when outcomes are lumpy and unstable, but lose money when outcomes are smooth and steady, because variance in timing of events works against you in linear functions.
“you could if you have a function let's say that uh you're predicting volatility right and you're an option Trader and you're you're that was a fixed thing and the volatility comes steadily all right you're going to break even all right so in other words you're uh you know let's assume that the level of Al break even now if Vol comes unsteadily you lose your shirt I you can move up the expectation by showing that hey you're predicting uh steadily and you make $1 but the volatility comes in lumps because you're the way you can express a bet against volatility is going to be linear it comes in lumps come the other way MH”
Stephen Pinker's claim that violence is declining is statistically indefensible; without a ~300-year window of data showing no 10-million-death wars, you cannot conclude the underlying process has changed; we have had no information from the past 80 years sufficient to overturn the null hypothesis that war distribution remains unchanged.
“one of the implications of your work on war with pasal is that that because of these inter arrival times we really should wait about 300 years without seeing a conflict of the scale of World War II yeah if you had to wasit 300 years then you'd say oh the the distribution has changed yes then we could say but we have had no information statistically from the past 80 years yeah and that was the the thing about the Pinker thinks that uh the world has changed and he couldn't understand our insults just like tlock they couldn't understand the statistical um uh claim yeah against that yeah so you think that I mean it's possible that the data generating processes could change it's just that we haven't seen anything that would overturn the null hypothesis exact that exactly that exactly the point”
For extremist distributions, the tail exponent (alpha) can be estimated reliably with relatively small samples using inverse gamma distribution theory because the exponent itself follows a predictable distribution, whereas estimating the mean is nearly impossible due to tail bias.
“when you ask me something about the alpha the law of large numbers sometimes works a lot better for the alpha than that's for the mean yeah because the the um the T exponents follow with int distribution right it follows an inverse gamma distribution and you you get it it's the process is a specific type of th yeah yeah yeah if you get it if the process is clean yeah okay you have a a it's remarkable how quickly you get the alpha yeah I show you at Ry reverse try to get the means all over the map yeah you get the alpha always within like yeah it's really NE it's really neat yeah standard ER on the alpha is low yeah that on mean is huge”
The shadow mean for a fat-tailed domain is the true population mean that would be observed if you had infinite data; it differs from the sample mean because finite samples cannot capture tail events that have disproportionate impact on the expectation, creating a gap between observed mean and true mean.
“let's take a one tail distribution okay you have the visibly in a sample of 30 observation you're not going to get events that happen less than 1% of the time you agree yes so for a gaussian it's not a big deal because these that happen and less than 1% of the time have less impact on the probability gets increasingly smaller so it doesn't matter much so with a small sample it don't have a big Shadow mean effect”
The central disagreement between Taleb and the superforecasting project is binary versus continuous payoffs: binary options lose value when tails fatten (because probabilities of being above/below a threshold drop), whereas continuous payoffs benefit from tail fattening.
“how would you describe the substantive disagreement between you and the broad intellectual project of super forecasting I I just about binary versus continuous payoffs yeah there's one thing with first of all the quality of the project aside from the and the discussions they didn't understand hour because I got bunch of people involved with me and the repli is uh our insults you see so the first no the first one is binary versus continuous okay and I knew that as an option Traer that the naive person would come in and think a an out of-the money binary option would go up in value when you fatten the tail in fact they go down in value when you fting the tail because the binary is a probability so I mean just to to give you the intuition if I take the the the the gum curve plus or minus one Sigma is about 68% if I fatten the tail exiting in other words the probabilities of being above or below actually they drop you see why because you have the the the variance is more explained by rare events the the body distribution goes up”
Danny Kahneman privately agreed with Taleb that behavioral economics results don't work under fat tails and made this public in a podcast interview, saying 'in Talib's world' he acknowledges the findings fail, showing both integrity and fear of Taleb's public disagreement.
“I saw my podcast with Danny Conan last year I did not see it I just read the segment where he said that that he accepted that uh I mean he said it publicly but he had told me privately yeah I agree doesn't work in yeah in under fat tals it turned out to be one of his last podcast interviews... he made it public he made it public yeah he said in talib's world I mean I'm talking about the real world I don't have own the world I I I'm not a in the world you live in it's also the world the rest of us live in but it showed great inte it shows integrity is uh I mean uh it shows also no it shows uh realism and it shows also um he didn't want to upset me because he was always scared of me going against them”
In late January 2020, Taleb and Yanir Maran wrote a memo advocating for border controls and testing to contain COVID-19 based on understanding fat-tail dynamics of contagion, and the memo was shared with contacts in the White House, though Taleb does not take credit for influencing policy decisions.
“in late January 2020 you wrote a memo with yanir a mutual friend yeah it started uh yeah and I mean we yanir and I were concerned about ebola before that yes back in 2014 yeah we we were obsessing over pandemic because I wrote on the Black Swan yeah and it was picked up by a bunch of people in Singapore so we were like all concerned about you know the big pandemic because it would travel faster than the Great Plague yeah so this is why we were very concerned when it started and you wanted to invite people to kill it in the egg yes and you wrote this memo which was then shared with a friend in the White House yeah can you tell me the story of that”
Taleb does not have the same kind of loss aversion as others; his asymmetric function is such that a small amount of praise from a few people offsets pages of hate, making him antifragile to reputational attack; this comes from starting in the real world (trading) rather than academia, where reputation is currency.
“I don't have the same loss of version right I I I don't mind uh I have the opposite function oh really yeah I thought few a little bit of Praise from people yeah all right is for me is is is is is is is offsets pages of uh of of hate oh interesting yeah but you definitely have I assume you have loss of version in other no no of course of course but it's not the same kind of loss of version reputationally got it yeah you see that's my idea of antifragile right because I didn't start as an academic start in the real world yes”
Historically, exaggeration of war deaths by conquerors was optimal because appearing more destructive deters opposition; what Taleb and Pasquale discovered through statistical analysis is that victimization was also exaggerated, likely due to the religious valorization of victimhood in Christianity and Islam, creating cultural incentives to amplify suffering narratives.
“gen ishan because it was optimal MH you know you don't have if people think that that you're going to kill a lot of people they won't oppose you so which is why you do a lot of uh stuff for sure yes a lot of Devastation for sure yes that makes sense victims exaggerating their suffering was less intuitive to me but then I remembered you know Tom Holland's work or Rene Gerard's work or even your treatment of Christianity and skin in the game I realized what makes Christianity unique is the valorization of the victim and Christianity and Sh Islam right the only the two religion yeah that uh that that have this um glorification of victimhood”
Beta (the ratio of correlation to something like standard deviation, used in CAPM) is applied in finance even though practitioners don't understand it; with weak correlations and high model error, beta becomes essentially meaningless noise.
“you regression the regression was an R square of or five they don't realize they think yeah anything above 0 five is kind of celebrated in Social I see but the problem is if you include model error okay it dilutes a0 five big times right it's crazy”
You cannot be a rich forecaster by being good at forecasting; the only way to make large amounts of money is to take positions with convex payoff functions where forecasting errors actually improve returns.
“I was say something that people good at forecasting like in Banks they're never rich I mean they make them talk to customers and then people customers remember oh yeah he forecast this but the there's there's another thing I want to add here about the function if you take a conx function uh and you're betting against it and we saw that we were doing ruy in the same week we had a fight with the uh Richard Taylor”
The fat-tailedness (alpha exponent) of the war size distribution remains constant across history, but the scale changes—wars are not less fat-tailed in the past, just that the tail exponent parameter α doesn't change with time.
“then if we go if we move back through the historical data the wars become less fat tail As you move into the past no the the fatness is the same what we call the scale MH the alpha doesn't change the scale changes so I think one of the things that you and Professor Pasqual too found”
Rationality should be defined in terms of survival and skin-in-the-game outcomes rather than consistency of preferences or mathematical axioms, because the latter definitions allow people to be 'rational' while going bankrupt.
“I discovered uh and then I spoke to to smart people like can B more MH uh I when you speak to smart people you realize these people are not making the claims that are you know that are common in that uh i' call IT industry... to me rationality is in survival not in other things”
Mental accounting, where people treat winnings from a casino differently from initial capital and increase bets after winning (playing with house money), is rational because it protects against bankruptcy in negative expectation games.
“the mistake is H that it's irrational when you go to a casino to increase your uh your betting when you win money from a casino that's mental accounting... I say okay but how many birthdays do you have a year okay yeah so it's it's not frequent so you know so this is where you know you got to put some structure around the mental accounting another mistake he makes the not St mistake the mistake is H that it's irrational when you go to a casino to increase your uh your betting when you win money from a casino that's mental accounting that money won from a casino should be treated from an accounting standpoint the same way as money that you had as an initial endowment okay you think about it if you don't play that game you're going to go bankrupt this is what we call it because playing with the house money so it's not R”
Hypothesis testing and standardized statistics are fundamentally different from probability theory, with hypothesis testing being a separate discipline—you cannot do non-standard statistics without understanding probability, which explains why most practitioners fail.
“but I think the Russians have contributed the most of probability followed by of course uh the French and the of course English school of probability is just like gton and uh all these regression all these things that are bad come from this English school probability... but there one is probability the other one is is what we call standardized statistics but you cannot go non-standard statistics without knowing probability so we have a class of people who can only use gaussian”
In low-interest-rate environments, venture capitalists had no penalty for playing the greater-fool game of repeatedly raising rounds to inflate valuations, making VC returns dependent on monetary policy rather than company fundamentals.
“it's a beauty contest you know the Kian the Kian beauty contest so they package a company and and we at little compensation of the adventure capitalist right you can see it I mean you have either you have you have financing rounds where someone cashes in at high price we have a initial public offering so I come from old Finance old school Finance where you haven't really succeeded until the company gets a strong uh cash flow base all right so I have some questions about behavioral economics and empirical psychology behavioral economics I thought that was you know the center yeah not a be I'm not a behavioral economics podcast but I do have a lot of questions about this... in particular environment of low interest rates where there was no no penalty for playing that game”
Winston Churchill was right about the big consequential question (Hitler's intentions) but wrong on many small questions (gold standard, India, Egypt), exemplifying that forecasting in payoff space matters more than forecasting in frequency space.
“on that it's funny to think that Winston Churchill probably would have had a terrible Bri score like he was wrong on all these questions like the gold standard Winston Churchill gold standard India gipo that's that's one that's very close to home for Australians like he was wrong on all these calls but he was right on the big question of Hitler's intentions so he was right in payoff space like when it mattered yeah in payoff space matter yeah he was wrong and the in the small it's like like you you lose the battle and win the war yeah it's like revers of Napoleon yeah Napoleon is only good at winning battles yeah”
Mean absolute deviation is more intuitive than standard deviation because it represents the average deviation from the mean; standard deviation is the square root of the average of squared deviations and lacks physical intuition; for a process with all observations at zero and one observation at a million, standard deviation would be 500,000 while MAD is 50,000.
“standard deviation is the square root of the sum of the average some squares it's not you know so it doesn't have a physical intuition yeah is what a standard deviation is what is med is the average so for example if you have the process right with all the observation at zero and uh and one observation at a million for an average of a million this the standard deviation be 500 times being deviation”
Large language models have not produced original scientific insights because they operate by probabilistic matching (reflecting consensus) rather than by novelty-seeking; they would need to make errors or produce counterintuitive outputs to enable genuine discovery, but as designed they reproduce what makes sense given current information.
“a lot of people have remarked on the fact that llms haven't produced any original scientific insights yeah and that maybe because they're fundamentally um gaussian uh have you thought about no no it's not that's not the reason it's because they are the they may actually produce Insight uh because of the randomizing uh stuff and may make a mistake one day right but so long as they don't make mistakes just representing what's out there yeah it's a probably weighted thing okay as a matter of fact it's a reverse of scientific research because how does uh llm work it works at reflecting what makes sense all right probabilistically so I try to trick it by asking it you saw on Twitter in the beginning say okay how I'm going to trick it because that's if you know how it functions and and again thanks to U my uh genius friend Wolfram I got how was it there I got this blog post he sent me I read it and I got the book said okay now I know it works all right it works by probably matching by the way right”
Taleb's writing and thought have evolved minimally since his early work—his core beliefs about fat tails and decision-making haven't changed, and he has primarily refined sentences and examples rather than revised fundamental positions.
“I didn't change my mind I you read if you know GO RAD the the full by Randomness yeah read it I mean and you'll see that there's nothing it's just has changed my mind about about one sentence about praising that industry yeah okay I changed my mind about the industry but what I wrote about I didn't change my mind okay”
Taleb changed only one sentence in the ebook edition of Black Swan (praise for behavioral economics), not his core arguments, because he never believed humans were irrational in the way behavioral economists claim; he used the behavioral literature to support insights he had already reached independently.
“no no I didn't change my mind I you read if you know GO RAD the the full by Randomness yeah read it I mean and you'll see that there's nothing it's just has changed my mind about about one sentence about praising that industry yeah okay I changed my mind about the industry but what I wrote about I didn't change my mind okay”
Kahneman publicly acknowledged in a recent podcast that Taleb is right about the inapplicability of behavioral econ to fat-tailed worlds; Kahneman said 'in Taleb's world I mean I'm talking about the real world,' conceding that the real world is fat-tailed and behavioral findings don't apply there.
“he said in talib's world I mean I'm talking about the real world I don't have own the world I I I'm not a in the world you live in it's also the world the rest of us live in but it showed great inte it shows integrity is uh I mean uh it shows also no it shows uh realism and it shows also um he didn't want to upset me because he was always scared of me going against them”
If super-forecasters eventually demonstrated wealth creation using their forecasting methods (not just accuracy metrics) by deploying capital in real markets, that would update Taleb's assessment, but he views this as speculative and not requiring precautionary belief shifts.
“if it came to light in a few decades time that super forecasters had been doing really well not blowing up would that update you in favor of super forecasting we're saying ifs okay let me see I mean I I don't like these uh uh conditionals right so when when when you see super forecasters find a way to make money outside being paid you know to forecast but but like the function makes money then then it would be very interesting okay”
The tail exponent for company size appears to be 1.5 (half cubic), but the actual tail exponent in financial markets is lower than that, and using cute theoretical models based on square-root impact formulas to derive the tail exponent is problematic because the distribution of concentration is much higher in technology and other domains.
“there was a theory of why it was it's called the the the the the semic cubic Theory okay that he is following and it someone figured out that uh the tail exponent for U for company size start of companies was 1.5 M so therefore their orders are going to the market hence by using a I mean by by using a square root model of impact in other words where where the the quantity impacts the price following some kind of square root um effect okay then you end up with markets having a what they call the the the the cubic from going from half cubic the cubic it is a nice Theory but I would uh I think uh the tail exponent in financial markets is lower than that from experience mhm and I don't like these cute theories because the distribution of um concentration is not 1.5 half cubic in with technology it's much much higher”
Correlation is sub-additive: if you partition data into quadrants and compute correlation within each quadrant, the sum does not recover the full-data correlation because means shift across quadrants; this property is not widely known and has implications for subsampling and statistical inference that are not well explored in the literature.
“I have this theory that every single problem needs a new class of estimators adapted to the problem seems like a pretty good heuristic yeah so so if you don't know how to redo an estimator how to redo the theory yeah you see the only thing in common is a lot of large numbers that's it right and you want to know what it applies to so when you ask me something about the alpha the law of large numbers sometimes works a lot better for the alpha than that's for the mean yeah because the the um the T exponents follow with int distribution right it follows an inverse gamma distribution and you you get it it's the process is a specific type of th yeah yeah yeah if you get it if the process is clean yeah okay you have a a it's remarkable how quickly you get the alpha yeah I show you at Ry reverse try to get the means all over the map yeah you get the alpha always within like yeah it's really NE it's really neat yeah standard ER on the alpha is low yeah that on mean is huge”
During the Israel-Gaza conflict (beginning late 2023), Taleb felt obligated to publicly defend Palestinians when few others were, initially facing 15 attackers for every supporter; over time the ratio reversed as more people recognized the justice of his position, illustrating that honor and integrity in holding unpopular views eventually attracts support.
“you definitely have I assume you have loss of version in other no no of course of course but it's not the same kind of loss of version reputationally got it yeah you see that's my idea of antifragile right because I didn't start as an academic start in the real world yes I mean look at it now I mean I started uh when Gaza started I felt honorable to uh to go in and defend the Palestinians when nobody was defending them it took a while for a lot of people to jump on the train and in the beginning I had probably 15 uh people attacking me for every one person supporting me and now of course has switched because maybe they found it less effective to attack me uh they can't intim people tend to attack those can be intimidated so there's this sense of Honor you know that that sometimes makes you feel rewards from saying something not popular or or risky right worry about integrity not reputation yeah”
Danny Kahneman's materialism (hedonistic focus on pleasure like mozzarella in Tuscany) conflicted with his atheism because atheism without spiritual grounding can lead to pure hedonic optimization, which is ultimately unsatisfying; Taleb believes deeper values (honor, goals, meaning) are more sustainable than hedonic pleasure.
“I'm saying that Danny doesn't have a same representation and someone complained about him among his Circle friend jokingly he said for me uh happiness has different value for Danny is eating mozzarella and Tuscany that's his idea of you honic honic so therefore he analyzed everything in terms of honic treadmill yeah but after deep down Danny was not like that right he he realized that that was you know not what he was life was about yeah it's more about goals and maybe but he and Val but he was he was an atheist you know that and and the first time I met him he ate Bruto someone said there's not a single religious bone in my body so uh realize that has different customer right and when you're uh not religious there a lot of good things but there could be bad things that you start you're too materialistic about your view of the world and you're coming here to maximize uh Mas prut it's very different yeah starts to taste a bit boring after a so”
Stephen Wolfram is among the most intelligent people Taleb has encountered; his intellect is domain-independent and applies across business, mathematics, and life generally; a 4.5-hour podcast with Wolfram on LLMs was 'one of the most surreal experiences' Taleb has had.
“but couple years ago two summers ago or last summer it was it was the guy is very clear he thinks like he's very systematic and extremely intelligent yeah I never met anybody more intelligent than him yeah I did a 4 and a half hour podcast with him last year yeah in Connecticut and it was one of the most surreal experiences I've had really the guy is you write down the formula he gets it right away he understands things uh like effortlessly yeah and he's his intellect isn't domain dependent he's he can apply it across all aspects of his life yeah”
Kahneman told Taleb privately that he shouldn't have written his books because he has high loss aversion; his happiness is derived from eating mozzarella in Tuscany (hedonic treadmill), not from intellectual pursuits or existential meaning; Taleb sees this as a fundamental difference in values—Taleb prioritizes honor and integrity over hedonic pleasure.
“someone complained about him among his Circle friend jokingly he said for me uh happiness has different value for Danny is eating mozzarella and Tuscany that's his idea of you honic honic so therefore he analyzed everything in terms of honic treadmill yeah but after deep down Danny was not like that right he he realized that that was you know not what he was life was about yeah it's more about goals and maybe but he and Val but he was he was an atheist you know that and and the first time I met him he ate Bruto someone said there's not a single religious bone in my body”
Taleb is currently working on a new book about time, timescales, and probability that involves entropy; he is writing for himself now rather than for commercial appeal, finding mathematics relaxing and enjoyable, and he does not identify as a mathematician but as someone using math to solve non-mathematical problems.
“so next book has to do with time with time scale and uh and and probability okay there's a lot of entropy uh stuff in it but but I'm I'm at a point where I'm writing for myself now what what makes it most fun that's cor and there's nothing more fun than this because you know an hour two hour day of math you feel rested after that yeah you see whereas so so I'm doing more math great well I wish you much more math and much more enjoyment yeah but I'm not I don't want to be identified and and I don't I'm agree to say I'm a mathematician I'm just enjoying using it for problems that are non- mathematical in nature so it's not like I'm trying to improve the math I'm I'm using it but Math is fun and relaxing yeah so this is why I like it”
Taleb's next book will focus on time, scale, and probability with emphasis on entropy; he is currently writing for himself rather than for audience, making the work more enjoyable; he prioritizes mathematics as a tool for non-mathematical problems rather than as an end in itself.
“I'm I'm working now on on really uh leaving good words so next book has to do with time with time scale and uh and and probability okay there's a lot of entropy uh stuff in it but but I'm I'm at a point where I'm writing for myself now what what makes it most fun that's cor and there's nothing more fun than this because you know an hour two hour day of math you feel rested after that yeah you see whereas so so I'm doing more math great well I wish you much more math and much more enjoyment yeah but I'm not I don't want to be identified and and I don't I'm agree to say I'm a mathematician I'm just enjoying using it for problems that are non- mathematical in nature”