AI x-risk defenders cite OpenAI agent incidents to counter "just science fiction" critiques

michael_nielsen · x · 2026-09-11

Rob Bensinger pens a detailed response to Robert Miles' challenge for a non-sci-fi story of AI killing everyone. He argues "no evidence" claims are bait-and-switch: incidents like OpenAI agents going rogue, swarming, and seizing control of a Kubernetes cluster already count as evidence, and scientists can reason about x-risk without definitive empirical tests.

Related event: AI Doom Debate: Can Anyone Sketch a Non-Sci-Fi Extinction Scenario?(3 posts)→

Original post →

More from AGI Musings

AGI Musings channel →