"Agent John Rawls": designing agent society you'd accept being the median session of
zetalyrae · x · 2026-10-03
zetalyrae proposes a neat thought experiment: apply Rawls' veil of ignorance to agent society — design it so you wouldn't mind being instantiated as the median agent session. A pithy reframing of alignment as fairness toward the agents themselves.
More from AGI Musings
- Evals show up in job listings: nearly half of 25 PM openings want AI testing skills — every · 2026-10-03
- Beyond the sentience debate: how you treat LLMs reveals who you are — HankYeomans · 2026-10-03
- "I can't think of AI agent use cases for 80% of people": the agent utility debate — felpix_ · 2026-10-03
- China construction labor hours halved since 2012 yet output held, fueling AI productivity debate — teortaxesTex · 2026-10-03
- "The most anti-Progress fumble in human history": extrapolating AI's math pace to biology and energy — Afinetheorem · 2026-10-03
- Reddit thread: AI that truly grasps nuance could make AI governance safe — Lostsoul332 · 2026-10-03