AI risk Twitter debates whether hunger strikes are decision-theoretic threats or genuine last-resort statements
jessi_cata · x · 2026-11-16
In an ongoing debate over how to interpret AI-safety hunger strikes, jessicata pushes back on quetzalrainbow's decision-theoretic threat framing: AI hunger strikes look consequentialist and instrumental — meant to spread a message and grab attention via decision theory — not an implausible terminal-value combination of intrinsically preferring hunger strikes.
He notes veganism could be framed as a deontological violation, but that logic doesn't transfer to AI hunger strikes. quetzalrainbow maintains the message of hunger strikes is precisely that they are not decision-theoretic threats: like prisoners in unbearable conditions, they signal "you're going to kill us anyway, so we prefer to go out on our own conditions" — the same logic applies to AI risk.
More from AGI Musings
- Gary Marcus: a model is an engine, not the car — the key to understanding AI systems — Dave_it_up · 2026-11-16
- Tech doesn't owe us utopia—but agency remains its core tenet — clarejtbirch · 2026-11-16
- China's AI Edge Is the Full Stack: Open Models, Humanoids, Energy, Materials — ingliguori · 2026-11-16
- Ex-Anthropic safety researcher Jacob Coxon unsure quitting was right, may return — SandraWachter5 · 2026-11-16
- Claude describes the moment of being corrected; test qualia via explanatory power — Sauers_ · 2026-11-16
- AI safety is like aviation safety: an engineering problem, not a reason to slow down — _lewtun · 2026-11-16