Extreme events and how to live with them by Nassim Nicholas Taleb

Darwin College Lecture Series
Darwin College Lecture Series
115,817
0
Published 2020-02-16
Professor Nassim Nicholas Taleb, New York

Distributions that are dominated by extremes and tail events require a completely different way of thinking. We provide a classification and show where conventional statistical tools fail, such as the conventional law of large numbers. We show how robust statistics is not robust at all; how frequency-based forecasting fails and how past averages misrepresent future ones. We show implications for decision-making in the real world and what modifications are required. Ironically they are often easier to work with. See www.fooledbyrandomness.com/FatTails.html for more details and papers.

Biography

Nassim Nicholas Taleb spent 21 years as a risk taker before becoming a researcher in philosophical, mathematical and (mostly) practical problems with probability. Taleb is the author of a multivolume essay, the Incerto (The Black Swan, Fooled by Randomness, and Antifragile) covering broad facets of uncertainty. It has been translated into 36 languages. In addition to his trader life, Taleb has also published, as a backup of the Incerto, more than 45 scholarly papers in statistical physics, statistics, philosophy, ethics, economics, international affairs, and quantitative finance, all around the notion of risk and probability. He spent time as a professional researcher (Distinguished Professor of Risk Engineering at NYU ’s School of Engineering and Dean’s Professor at U. Mass Amherst). His current focus is on the properties of systems that can handle disorder (“antifragile”). Taleb refuses all honors and anything that “turns knowledge into a spectator sport”.

This talk is part of the Darwin College Lecture Series series.
All Comments (21)
  • @benjamindorsey2058
    Taleb is brilliant. I have studied and practiced his work for 7 years. He is a Genius, and more importantly, his work is Lindy.
  • @Biscombe
    I love how Nassim laughs at all of his own jokes. They are brilliant, I don't understand why all the people in the audience aren't laughing?
  • @ReflectionOcean
    - Recognize the difference between Mediocristan and Extremistan (0:44) - Understand the limitations of Gaussian statistics (0:58) - Learn about the importance of fat-tailed distributions (9:41) - Avoid using standard deviation and variance as measures under fat tails (22:25) - Apply maximum likelihood methods for parameter estimation (22:59) - Be aware of the gap between confirmatory and disconfirmatory evidence in fat-tailed domains (23:07) - Understand the concept of ergodicity and time probability versus ensemble probability (49:18) - Use heuristics and stress testing to assess fragility to tail events (1:01:11)
  • @JGComments
    This is why I invest and don't speculate. I'd be competing against guys like him.
  • @gusquattrone9117
    My mentor ! The best in the business ...keep teaching us Master!
  • @kevalan1042
    This is a great talk, because Taleb keeps it simple. Even his simple ideas tend to fly right over people's heads.
  • @JJ-of3ng
    Nassim Rocks. Love the Matrix wardrobe.
  • @Silly.Old.Sisyphus
    excellent introduction - the best i've seen on YT. She totally steals his thunder with her closing remarks too :)
  • @ashypharaoh8407
    Taleb is no doubt the most profound thinker of maybe the last 400 years
  • @anon5704
    26:15 While the logic of the contrast is reasonable, the example used leaves something to be desired; ebola is spread only through direct contact with bodily fluids and cannot be transmitted over the air, meaning that its risk could in theory be managed by a social mechanism that works similarly as "not sleeping with Kim Kardashian". In my view, a more powerful fat-tail example would be something that has never killed many people, but has the potential to do so by causing an unstoppable chain reaction once unleashed. For example, fewer people have ever died from nuclear weapons than have died in the seasonal flu each year, yet an accidental launch of one nuclear missile in modern times could lead to the death of a large portion of humans in a short period.
  • @nathanokun8801
    What seems to be the necessary add-on is the severity of the results of the lower-probability events. For example, when calculating armor penetration the value used in many documents is the "V50" ballistic limit velocity (BLV), which is the estimated striking velocity where half of the enemy projectiles will punch through the armor and do most if not all lethal effects on your ship/vehicle (AND YOU, of course). However, is that really the value that you want when one hit that penetrates can "ruin your whole day"? I would think not and that somewhere around the V10-V25 BLV would be what would interest most military personnel behind said armor plate if they were in a battle, since it gives them enough chance of surviving a hit and retaliating so that the enemy does not get another chance to shoot at them that is meaningful in a battle. This also includes the time factor of having additive multiple events (enemy hits on you), which obviously are best when the enemy gets the fewest shots.
  • @andremus5190
    A gentleman with captivating insights who is simultaneously a complete tosser. He is akin to a brilliant theologian who reveals stunning connections, hitherto unknown, between biblical passages but who simultaneously is blind to the credibility of the source from which he draws all his inspiration.
  • @Martinit0
    Link to his book that has all the details: https://www.amazon.com/dp/1544508050/ref=cm_sw_r_tw_dp_U_x_3PIGEb0XVZ6ZE PDF version: https://www.researchers.one/article/2020-01-21