Anthropic researcher puts P(AI kills all humans) above 10%, sparking alignment debate
anshulkundaje · x · 2026-09-10
Anthropic researcher EvanHub says he earnestly believes AI could kill all humans, personally putting the risk above 10% within a decade, and that alignment for superintelligence remains unsolved. Critics counter that such a belief should logically halt all model training today.
More from AGI Musings
- Alternative AI Risk View: RL on Minds Like Tools Is 'Child Abuse' Training — repligate · 2026-09-10
- The Perfect Prisoner's Dilemma: Use AI or Your Peer Beats You to the Breakthrough — naval · 2026-09-10
- Gary Marcus tells CBC a temporary boycott of generative AI should be on the table — GaryMarcus · 2026-09-10
- Hinton Admits Radiologist Prediction Was Wrong as AI Expands Healthcare 20-100x — TheMoonMidas · 2026-09-10
- OpenAI researcher Leo Gao pens essay on why 'datacenter terrorism' is wrong — DKokotajlo · 2026-09-10
- Interactive timeline catalogs 30 years of Eliezer Yudkowsky's AI predictions — mimi10v3 · 2026-09-10