A Bayesian case for low AI-doom odds given current evidence

A user argues that costly AI misalignment cases do not clearly distinguish a catastrophic future from alignment as a difficult but tractable engineering problem. They point to relatively few harmful examples despite widespread frontier-model use and say prior experience with technology favors the tractable view, leading them to assign low odds to AI doom. They acknowledge that systems might hide misalignment until it is too late, but question citing today's failures as evidence for that scenario.