As AI companies get closer to ‘recursive self-improvement,’ this physics professor calls full autonomy ‘the worst idea in the history of humanity’

The prospect of artificial intelligence models teaching themselves autonomously to be more efficient and capable appears ever closer to reality.

≈