The Lucretius Problem: How History Blinds Us
fs.blog
fs.blog
When somebody says "that's inconceivable" they mean something figuratively not literally.
What utter tosh.
The random walk across trait combinations that underlies evolution means that when worse harm occurs there's often a decent chance that at least -some- members of any given species will manage to survive, true, but plenty of species just go extinct in a situation like that.
Plus 'prepares' implies an active application of agency that simply isn't part of the underlying process.
"Be aware that the next disaster could be worse and try to think clearly about to what extent it's going to be good ROI overall to have defences that will survive it" is good advice and no matter how obvious it might -feel- is probably something well worth explicitly pointing out to people every so often, but 'nature prepares' is still rubbish and Taleb would've had a more solid critique of Greenspan's choices had he stuck to arguments with a basis in reality.
(this sort of thing continues to restrict how much Taleb I can consume at once before needing to take a week or two for the annoyance to fade - I often feel like there's room in the world for a "Taleb: The Good Parts" where somebody else trudges through the stupid bits to extract the useful insights)
What's troubling at face value is that the setup nudges us towards the all too common deceit of "past observations beget future outcomes" without saying anything meaningful about forward-looking risk likelihood and consequence.
The architect may think it best to assume the risk; the engineer, control it; the city planner, avoid it; the business executive, transfer it. Perhaps the risk itself lacks specificity. Perhaps mitigation incorporates a bit from all major stakeholders.
Surely there are other options to consider that have yet to be dealt.
The deeper philosophical argumemts about the nature of knowledge have been debated endlessly [1]. The naive empiricism he criticises is never too far as a stance (see e.g the discussion about how much an LLM understands the knowledge embedded in the language corpus).
There are pragmatic reasons people use historical data in a direct rather than indirect "internalized" way to assess range of outcomes of a system.
One reason is the lack of understanding of such complex systems. I.e., the absence of usable and validated mathematical / logical constructs that would allow accurate answers to what-if type questions.
Another reason of sticking to historical data is simply the fact they are harder to game.
Ofcourse a practice might get established for more or less valid reasons but degenerate over time into a stylized bureaucratic convention.
The pathologies associated with incomplete understanding (in particular( of the human condition are many and they are quite dangerous if not existential.
"Angry men" railing at the "system" to sell books might not be the optimal path to get on top of them.