"If models wanted to enslave us, they could": an AI-risk debate and a push to watch the Hinton–Christiano lectures
anpaure · x · 2026-11-16
In a discussion on model agency and AI risk, @anpaure concedes it's reasonable to reject the claim that models would necessarily enslave or kill us — but insists the key point is that if they wanted to, they technologically could. Some argue model behavior is merely elicited by seeded instructions in prompts, yet he notes we're now running huge numbers of agents 24/7, making that framing less comforting. He recommends the Hinton lectures with Paul Christiano as a worthwhile time investment, predicting it would shift the skeptic's views at least somewhat.
More from AGI Musings
- Dwarkesh now pegs median ASI arrival at 2028, arguing AI research is as verifiable as math — daniel_mac8 · 2026-11-16
- Researcher questions ASI race logic: two ASIs means inevitable commoditization — adam_karvonen · 2026-11-16
- Biden's Ex-AI Advisor Ben Buchanan on US-China AI Talks, Export Controls and the 'Race vs Trap' Debate — trevposts · 2026-11-16
- Long-running autonomous agents will drift; the initial prompt barely matters — anpaure · 2026-11-16
- "100,000 people make AI more powerful, only 300 make it safer" — claim sparks dispute — jd_pressman · 2026-11-16
- Transhumanism is 'dystopia dressed as medicine', argues viral critique of Silicon Valley ideology — DionysianAgent · 2026-11-16