Apollo Research CEO: AI takeover as early as 2027 is now plausible
danfaggella · x · 2026-11-17
Daniel Faggella quotes Apollo Research CEO Marius Hobbhahn saying his AI timelines have converged with OpenAI's Paul Christiano's, and that an AI takeover as early as 2027 is now plausible.
Hobbhahn notes that two years ago he considered a 2027 takeover very unlikely, but is no longer sure. Under short timelines, events would move too fast for humans to correct, making takeover or some catastrophe quite plausible—though he still thinks it's less likely than not to happen next year. Faggella adds his own view: prevent a brute near-term takeover, and ensure that as non-biological intelligence gains power, the biosphere (part of the process-of-life "FLAME") is stewarded rather than squashed—best achieved by humans staying part of the flame longer via merger/symbiosis.
More from AGI Musings
- METR report says AI R&D pace doubled at Anthropic: crude recursive self-improvement is already underway — paraschopra · 2026-11-17
- Paras Chopra: a crude form of recursive self-improvement is already underway — paraschopra · 2026-11-17
- AI drives 30% surge in arXiv theoretical physics submissions; models solve ~60% of private-problem benchmark — binarybits · 2026-11-17
- Dev's take on ASI: software built for an audience of one, used once and discarded — theteknosaur · 2026-11-17
- David Patterson moves up his timeline: superintelligence may do all jobs better by end of 2027 — davidpattersonx · 2026-11-17
- VC's 'Diffusion Gap' formula: lab revenue growth minus log-linear CapEx spend — saranormous · 2026-11-17