V. Anantha Nageswaran: AI does not know what it doesn’t know—and that’s reason enough for abundant caution
Perrow’s argument is that in systems with high complexity and tight coupling, catastrophic failures are not aberrative, but normal. They are what the system produces, eventually, by virtue of what it is. The 2008 financial crisis is the canonical recent illustration: every individual institution was regulated and every instrument technically compliant, but catastrophe emerged from the interaction of compliant components at a speed and scale that made intervention impossible before the damage was done.
Post Comment