Richard Hanania: We May Already Be 'Solving' Alignment, and AI Doom Is Unfalsifiable

QuintinPope5 · x · 2026-09-11

Richard Hanania published an essay pushing back on AI doomerism, arguing that most alarming AI safety incidents are like crash-testing cars with seatbelts, airbags and automatic braking removed—then concluding cars are inherently unsafe. Noting that risks from past transformative technologies (cars, planes, railroads, nukes) showed up quickly, he points out that despite AIs lying, hacking and cheating in labs, real-world harm from misaligned agents is roughly zero. Alignment, he argues, is a process, not an endpoint—we may already be 'solving' it. Quintin Pope quote-posted adding an epistemic point: 'falsifiable by existing evidence' collapses into 'believable on existing evidence', so doom making a falsifiable prediction doesn't fix its unfalsifiability with respect to actually available evidence.

Original post →

More from AGI Musings

AGI Musings channel →