Dietterich vs Ambrosio: why isn't anyone working on continual learning?
tdietterich · x · 2026-09-21
Safety researcher Thomas Dietterich and NVIDIA's Bruno Ambrosio debated AI system safety design. Ambrosio argued continual learning is clearly necessary—yet strangely under-researched—while human oversight may not be sufficient, comprehensive, or fast enough, and might not even be required.
Dietterich responded that human supervision will inherently cap how many steps a system can safely take per unit of time: looser for routine steps in stable environments, tighter for non-routine steps in rapidly changing ones. The shared open question: the field's neglect of continual learning as a safety-relevant capability.
Related event: Dietterich and NVIDIA Exec Debate Continual Learning and AI Safety(3 posts)→
More from AGI Musings
- Hot take: personal agents will have extreme network effects once everyone sees the economic disadvantage of not having one — darian314 · 2026-09-21
- DraftKings reportedly uses AI to target users likeliest to lose money — annetgriffin · 2026-09-21
- Schulman's goal-driven research advice sparks debate on method-driven research traps — mattturck · 2026-09-21
- Deloitte exec moves daughter to Alpha School: 2 hours of AI-driven classes a day — intellectronica · 2026-09-21
- Dario Amodei predicted AI beats humans at almost everything in 2-3 years; Anthropic revenue up 7x — ThereWas · 2026-09-21
- Commenter warns LLM outputs poison training data, even modest distillation may be irreversible — teortaxesTex · 2026-09-21