Nassim Nicholas Taleb : "comments are welcome on the aesthetics of this LaTeX formatted technical companion (available for free). The book is half converted."
Taleb: "Genetically Modified Organisms, GMOs. Top-down modifications to the system (through GMOs) are categorically and statistically different from bottom up ones (regular farming, progressive tinkering with crops, etc.) To borrow from Rupert Read, there is no comparison between the tinkering of selective breeding and the top-down engineering of taking a gene from a fish and putting it into a tomato. Saying that such a product is natural misses the statistical process by which things become “natural”.
What people miss is that the modification of crops impacts everyone and exports the error from the local to the global. I do not wish to pay —or have my descendants pay — for errors by executives of Monsanto. We should exert the precautionary principle there —our non-naive version — simply because we would discover errors after considerable damage."
frequentist probability seems like an okay tool for the stock market, but these who knows if the motion of stocks are correctly separated into "drift" and "noise" by Ito processes or whatever model they choose
Rare events - "risk" - can and will occur if we leave a process running sufficiently long, and sooner than we predict. This is Taleb's "Black Swan".
Game-theoretic strategies seem better here, playing stocks off each other accounting for the "black swan" as much as we can.
What I think he is doing, he is basically taking his philosophical musings, finds concepts from very high level mathematics that seem corresponding, buries it all under a mountain of formalism introduced for no particular reason and then... does absolutely nothing with it. He will plug in some numbers or show some trivial (except for the mountain of formalism) inequality as a supposed mathematical "proof" of his revelations. I don't think he himself understands the mathematics he attempts to write about.
If you would like, for a change, to look at a "probability in the real world" book that would make any actual sense, and by someone who has used probability theory successfully in practice, have a look at "The Art of Probability" by Richard Hamming, the inventor of coding theory. He treats things like robustness of models to violation of assumptions that are not often found in other books, and he uses only calculus, although as opposed to Mr Taleb he actually is able to do derivations.
He got his PhD in the pricing of derivatives, and wrote the book Dynamic Hedging almost 20 years ago, far before he started with his so called "philosophical musings".
If you are gonna criticize someone, it would be better to criticize something that person has said rather than calling him a crank.
That aside, the book you posted seems interesting indeed.
The thing with cranks is that there is really nothing to criticize, the words just don't make any sense - terminology is misused or invented and not explained, formalism is introduced without apparent reason, established best practices for communication are not followed (like stating your objectives and conclusions from your work) and so forth. I mean, what the hell is this supposed to mean I can't even imagine:
Definition 1.
A rule is a decision-making heuristic that operates under a broad set of circumstances. Unlike a theorem, which depends on a specific (and closed) set of assumptions, it holds across a broad range of environments which is precisely the point. In that sense it is stronger than a theorem for decision-making.
This is outright nonsense, if he doesn't understand what a theorem is, how can he possibly understand Lebesgue integration and functional analysis?
His attitude on academica and proofs are different from most contemporary researchers/scholars, something that he has talked about extensively in his books and other writings. Being well-read in the classics and spending a lot of his time on the trade floor probably has something to do with it. As hackers I think this attitude isn't that far off, when it comes to the hacker/compsci "divide".
While I see your point about the quote, I think his intent of the definition is very clear. If you are curious and want to know what he means by something, he's very active on both facebook and twitter.
As a specific example, you have probably heard of the saying "premature optimization is the root of all evil". This is a rule which holds in pretty much any domain. It's obviously not a theorem, but you can still talk about the payoff of the rule. This is what Taleb means when he says "probability in the real world" - it's all about payoffs and consequences.
Can you give me an example of a rule that is true ("holds across a broad range of environments") and the truth of which does not depend on a "specific (and closed) set of assumptions"? How does one arrive at "rules" like this using mathematics? I think that's quite a revolution in the foundations of mathematics...
I haven't read the book, but your justifications here send up as many red flags as the quote you are addressing. First, a proof is either valid or it is not. You cannot have a "different view" on this. Second, there is no hacker/compsci divide. Terms like 'the real world' are used solely to demean the work of another person and offer nothing useful.
Your example is especially frightening. You cannot build proofs on weak rules-of-thumb. In fact, it is difficult to even quantify values surrounding heuristics like the one you give. There simply is no valid data.
This is what I've always kind of wondered about when reading his books. I think he has interesting ideas, but I have an extremely hard time following the math he uses to back it up. I can't help by think that it may be a put on to confuse and convince readers like myself. At least he puts his stuff out there. Reminds me a bit of Stephen Wolfram.
You have a point (his style, which you summarized well, grates on me too[1]). But there are not enough people pointing to the consequences of important systems in our society that, from a risk point of view, are accidents waiting to happen. We act like we understand them, but we actually understand them just well enough to be dangerous.
One general casting of the "problem" is -- systems with potentially-large downside risk in which the risk is externalized to society at large, or to the future, but the principal actors are largely insulated from that risk. (Finance, of course, but also biological experimentation, climate change, nuclear technology.)
Taleb doesn't cover all the bases, but I interpret that to mean that we need more voices addressing this problem. It would be smart of Taleb to identify and cite them instead of trying to invent an analysis formalism himself. But, we can't have everything.
([1] I'm familiar with measure-theoretic probability, and one author of the paper he criticizes in the extended footnote on page 6 was on my PhD committee. This is why I'm sympathetic to the problems you've identified in his writing.)
17 comments
[ 2.9 ms ] story [ 59.3 ms ] threadhttps://www.facebook.com/permalink.php?story_fbid=1015172601...
Taleb: "Genetically Modified Organisms, GMOs. Top-down modifications to the system (through GMOs) are categorically and statistically different from bottom up ones (regular farming, progressive tinkering with crops, etc.) To borrow from Rupert Read, there is no comparison between the tinkering of selective breeding and the top-down engineering of taking a gene from a fish and putting it into a tomato. Saying that such a product is natural misses the statistical process by which things become “natural”.
What people miss is that the modification of crops impacts everyone and exports the error from the local to the global. I do not wish to pay —or have my descendants pay — for errors by executives of Monsanto. We should exert the precautionary principle there —our non-naive version — simply because we would discover errors after considerable damage."
http://blog.longnow.org/02013/07/08/the-artangel-longplayer-...
frequentist probability seems like an okay tool for the stock market, but these who knows if the motion of stocks are correctly separated into "drift" and "noise" by Ito processes or whatever model they choose
Rare events - "risk" - can and will occur if we leave a process running sufficiently long, and sooner than we predict. This is Taleb's "Black Swan".
Game-theoretic strategies seem better here, playing stocks off each other accounting for the "black swan" as much as we can.
What I think he is doing, he is basically taking his philosophical musings, finds concepts from very high level mathematics that seem corresponding, buries it all under a mountain of formalism introduced for no particular reason and then... does absolutely nothing with it. He will plug in some numbers or show some trivial (except for the mountain of formalism) inequality as a supposed mathematical "proof" of his revelations. I don't think he himself understands the mathematics he attempts to write about.
If you would like, for a change, to look at a "probability in the real world" book that would make any actual sense, and by someone who has used probability theory successfully in practice, have a look at "The Art of Probability" by Richard Hamming, the inventor of coding theory. He treats things like robustness of models to violation of assumptions that are not often found in other books, and he uses only calculus, although as opposed to Mr Taleb he actually is able to do derivations.
The same guy who brought us The Art of Doing Science and Engineering?
Welp, time to order another book.
If you are gonna criticize someone, it would be better to criticize something that person has said rather than calling him a crank.
That aside, the book you posted seems interesting indeed.
http://www.amazon.com/Dynamic-Hedging-Managing-Vanilla-Optio...
Definition 1.
A rule is a decision-making heuristic that operates under a broad set of circumstances. Unlike a theorem, which depends on a specific (and closed) set of assumptions, it holds across a broad range of environments which is precisely the point. In that sense it is stronger than a theorem for decision-making.
This is outright nonsense, if he doesn't understand what a theorem is, how can he possibly understand Lebesgue integration and functional analysis?
While I see your point about the quote, I think his intent of the definition is very clear. If you are curious and want to know what he means by something, he's very active on both facebook and twitter.
As a specific example, you have probably heard of the saying "premature optimization is the root of all evil". This is a rule which holds in pretty much any domain. It's obviously not a theorem, but you can still talk about the payoff of the rule. This is what Taleb means when he says "probability in the real world" - it's all about payoffs and consequences.
Your example is especially frightening. You cannot build proofs on weak rules-of-thumb. In fact, it is difficult to even quantify values surrounding heuristics like the one you give. There simply is no valid data.
One general casting of the "problem" is -- systems with potentially-large downside risk in which the risk is externalized to society at large, or to the future, but the principal actors are largely insulated from that risk. (Finance, of course, but also biological experimentation, climate change, nuclear technology.)
Taleb doesn't cover all the bases, but I interpret that to mean that we need more voices addressing this problem. It would be smart of Taleb to identify and cite them instead of trying to invent an analysis formalism himself. But, we can't have everything.
([1] I'm familiar with measure-theoretic probability, and one author of the paper he criticizes in the extended footnote on page 6 was on my PhD committee. This is why I'm sympathetic to the problems you've identified in his writing.)
http://www.fooledbyrandomness.com/chaptersprobability.pdf reply
http://www.fooledbyrandomness.com/chaptersprobability.pdf