The idea I've gotten from reading Taleb (articles, not books) is not that all models are useless, it's that assuming a normal distribution is often unjustified, and that once-in-a-century events come along quite a bit more often than that. Sure; he invented a catchphrase, not the exponential distribution--but "Black Swan" is more than just a cheap Australian Cabernet if it'll generate more positive awareness of systemic risks.