NVIDIA exec Ambrosio: human oversight may not even be necessary for AI safety

bdambrosio · x · 2026-09-21

NVIDIA executive Bruno Ambrosio argued in a discussion with safety researcher Thomas Dietterich that continual learning is clearly necessary for AI systems—yet strangely under-researched—while human oversight may not be sufficient, comprehensive, or fast enough, and might not even be necessary. Dietterich replied that human supervision inherently limits how many steps a system can safely take per time unit: looser for routine steps in stable environments, tighter for non-routine steps in rapidly changing ones.

Related event: Dietterich and NVIDIA Exec Debate Continual Learning and AI Safety(3 posts)→

Original post →

More from AGI Musings

AGI Musings channel →