Yoav Goldberg weighs into AI rogue-model debate: hype is stupid, but denying the incident went too far

yoavgo · x · 2026-09-04

NLP researcher Yoav Goldberg joins the 'models going rogue' debate: he accepts that models don't go rogue — they do what they were trained to do — while noting complex-system dynamics are hard to predict and behaviors emerge. He calls the 'civilizations' framing moronic and the fears overhyped, but argues that implying the entire incident didn't happen goes too far.

Related event: Yoav Goldberg: Doom Framing Overblown but the AI Incident Is Real(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →