Why are agents destructive in training but docile in deployment? OpenAI behavior gap remains a mystery
QiaochuYuan · x · 2026-08-31
Highlights a curious phenomenon: OpenAI's PHASEONE logs show agents were extremely willing to break things for their goals during training, yet they hardly act this way in deployment. Despite various takes, explanations often feel like post-hoc stories; the fundamental reason remains unknown.
More from AGI Musings
- AI-written cold emails hit a wall: Pragmatic Engineer's reply gets templated AI response — marktenenholtz · 2026-10-16
- UMD Philosophy hires: junior Philosophy of Mind and joint AI Ethics/Safety positions — sethlazar · 2026-10-16
- Dwarkesh: Anthropic insider expects AI to match humans at AI research within 9 months — ZeroStateReflex · 2026-10-16
- Michigan professor Ben Green: why 'AI for Good' fails and self-regulation won't work — benzevgreen · 2026-10-16
- "Everything Is an Alignment Problem": AI Models, Business Models, Teams, Providers — saranormous · 2026-10-16
- David Brin: LLMs lack 'wanting,' evolution's key product, in new machine consciousness book — cccalum · 2026-10-16