His theories are largely a rejection of placing too much faith in the scientific method (without rejecting it entirely, of course). The essence of his entire philosophy is that naive empiricism causes you to ignore things that happen rarely, and the more you build the assumption that rare events do not happen into your models, the more it hurts when they do happen.

This is, by definition, a hypothesis that is both obviously true and also almost impossible to test. The very problem in question is that there is a fundamental lack of data on these rare events because they happen rarely.

He has a 450 page mathematical explanation of his theories in his book Silent Risk.

http://www.fooledbyrandomness.com/FatTails.html

I'll read this thanks.

I don't particularly care about his position on GMOs. In fact I sort of agree with it; his point can be reduced to "we're not exactly sure what we're doing here, perhaps we should try to avoid scalable disasters" which is sound.

I was ambivalent towards Taleb a few years ago, but ever since he's upped his online presence he's become actively annoying

what's stopping you from ignoring him?

You don't seem to know what scientific means, do you?

Are his theories formulated in a rigidly defined and testable manner, and empirically tested?

No, they're "rules of thumb". Rules of thumbs can be great to make complex decisions (like in trading, as he does), but coming up with a few rules of thumb and doing well at your job does not warrant his level of confidence in his intellect.

Aren't rules of thumb in his case the outcome of inductive reasoning based on many observations? Sounds pretty empirical to me. It is a fundamental strength of science that such formulations accept falsifiability - so the fact that his hypotheses/theories/laws don't hold in all cases just means they need to be refined for those cases.

It also common scientific practice to accept hypotheses within bounds of probability.

The wildest conspiracy theories you've heard are also "the outcome of inductive reasoning based on many observations". Empirical testing is supposed to be statistical, not subject to human bias.

Well no, because they are neither falsifiable nor independently verifiable. Taleb proposes hypotheses based on openly available information and he provides enough information to replicate his findings.