Critic says AI safety firms must answer on TV: Dario dodged "could AI kill all humans?"
eli_lifland · x · 2026-11-19
Eli Lifland amplifies Nathan So8res' (ex-MIRI) new post criticizing AI companies founded in the name of safety for dodging a key question on TV.
- When Anderson Cooper asked in September "Do you earnestly believe that AI could kill all humans?", Dario reportedly changed the topic, insisting recently-resigned Jacob Coxon wasn't mainly expressing concerns about Anthropic — a claim Coxon almost immediately disavowed.
- Dario's closest answer, "the chance that things go wrong" is "not all that high", appears to contradict his statements elsewhere.
- The post sketches what an honest answer would look like: the risk exceeds a crewed NASA flight, but the stake is all of humanity.
- Lifland's point: a safety-branded company's absolute minimum standard is answering this question earnestly in public, not deflecting.
More from AGI Musings
- Developer blasts Anthropic: 'Who elected it the arbiter of intelligence?' — examachine · 2026-11-19
- Arundhati Roy essay: Anthropic paid authors just thousands per pirated book — examachine · 2026-11-19
- Musk: Superintelligence and robots will need orders of magnitude more bandwidth than humans — XFreeze · 2026-11-19
- What data companies are actually worth: 7 buyer/sider interviews on AI data pricing — joecole · 2026-11-19
- Andy Matuschak: Where's the Moral Desperation in Anthropic's Racing Stance? — KatjaGrace · 2026-11-19
- RSI debate: critics call recursive self-improvement a big-tech story, not a startup opportunity — teortaxesTex · 2026-11-19