Dev jab: AI safety debate ignores that bad RL environments are the real root cause
snwy_me · x · 2026-09-13
A pointed AI-circle take: safety preachers are "philosophers" who assume models are perfectly-developed agents, ignoring that substandard or plainly bad RL is the true root cause. The quoted post adds: vet your training environments, and where is OpenAI's promised CoT oversight?
Related event: Debate: Bad RL Training, Not Evil Models, Drives AI Risk(2 posts)→
More from AGI Musings
- Jessica Taylor revives 2023 essay: non-superintelligent paperclip maximizers are normal — jessi_cata · 2026-09-13
- Chamath accuses Dario of wanting to kill open source under guise of 'slowing AI' — markjeffrey · 2026-09-13
- David Patterson Counters ASI Doom Argument: Intelligence Doesn't Desire Power — davidpattersonx · 2026-09-13
- Cambridge's David Krueger recommends Katja Grace podcast as AI extinction-risk primer — DavidSKrueger · 2026-09-13
- David Krueger recommends KatjaGrace podcast as intro to AI extinction risk — DavidSKrueger · 2026-09-13
- A Prediction Without Uncertainty Is Only Half a Prediction in AI for Science — bravo_abad · 2026-09-13