Sam Altman on AI consciousness: nobody knows, and it would be 'a very strange form of it'
coherence · x · 2026-10-22
Matthew Berman's new interview with Sam Altman devotes a substantial segment to AI consciousness, and quote-tweeter coherence offers quick takes:
- Sam concedes nobody knows whether today's AIs are conscious — including him — and if so, it is 'a very strange form of it.' Given the professed uncertainty and the stakes, coherence argues OpenAI should devote a tiny fraction of its resources to accelerating the science here.
- Sam dislikes the 'whole framing' of AI consciousness, worrying it could change how humans relate to these systems and how much control we retain. coherence calls this a conflation of patienthood and agency: acknowledging that torturing a dog is wrong doesn't require granting dogs voting rights or property.
Other chapters cover when AI will discover new treatments, why AI hasn't killed coding jobs, personal AI assistants, ChatGPT being too agreeable, OpenAI's bet on chips and cheaper compute, and why 'AI will cure cancer' isn't enough.
More from AGI Musings
- "Name one job AI can never replace" — Reply: "Spoiler: there isn't one" — MattGarciaEth · 2026-10-22
- Dev warns the internet is becoming a giant bot farm run by LLMs — and most of it won't get patched — kristoph · 2026-10-22
- Dan Jeffries: open models are your last line of defense, anyone who takes them is the enemy — Dan_Jeffries1 · 2026-10-22
- AI scientist De Kai: AIs are toddlers, not toasters — and regulation alone won't suffice — dekai123 · 2026-10-22
- European AI models leave cleaner, more auditable trails, but still lack frontier scale — AryHHAry · 2026-10-22
- The 1000+/10000+/100000+ theory: prove you can actually use AI in the agent era — AlchainHust · 2026-10-22