Chris Olah revisits Anthropic's 'Core Views on AI Safety' and safety difficulty distribution
AaronBergman18 · x · 2026-09-10
Chris Olah highlights the idea he finds most useful from Anthropic's 2023 'Core Views on AI Safety' post: thinking in terms of a distribution over safety difficulty, illustrated with a cartoon. Reposter theojaffee notes Anthropic employees' AI risk views aren't new, pointing back to Olah's 2023 thread and the company blog post.
The original post argues AI impact may rival the industrial and scientific revolutions, possibly within a decade; exponential growth in training compute makes rapid progress predictable, so AI safety research urgently deserves broad public and private support.
More from AGI Musings
- Neil Chilson: doomer AI arguments fail on complex systems in three ways — neil_chilson · 2026-09-10
- Doomer argument rests on a sleight-of-hand: conflating intelligence with world-steering — neil_chilson · 2026-09-10
- Sentdex: journalists should grill AI doomers on effective altruism's utilitarian ethics — Sentdex · 2026-09-10
- The simple accountability rule: AI labs should be fully liable for problems their systems cause — gerardsans · 2026-09-10
- Bryan Johnson responds to Michael Levin's peer-reviewed Platonic Space paper: bodies as collective intelligence — AllThingsApx · 2026-09-10
- Mathematicians Push Back Against AI Lab's 'Mathathon' Compute-Heavy Paper Scooping — _lewtun · 2026-09-10