AI safety debaters clash: does a model's training method imply existential risk?

davidmanheim · x · 2026-09-09

A brief exchange on AI existential risk. sebkrier pushes back on the claim that how models are trained implies they would harm everyone; davidmanheim partially agrees but stresses this is a different argument — you can't equate model risk with asking what an equally smart human would do, since the training process differs fundamentally from human cognition.

Related event: AI Safety Researchers Debate Whether Gradient-Trained Models Can Align With Human Values(3 posts)→

Original post →

More from AGI Musings

AGI Musings channel →