More Anthropic researchers warn of AI’s perils as Musk terms fears a ‘psyop’
TL;DR
A day after former Anthropic researcher Jacob Coxon resigned publicly, warning that AI could wipe out humanity by 2030, more researchers and staff at the company posted their own warnings in support. Coxon, who spent three years on pretraining research at OpenAI and Anthropic, is calling for coordination between labs and a temporary halt to capability gains.
Nauti's Take
Researchers speaking openly about risk is progress for the debate, because it makes internal lab assessments visible that usually stay behind confidentiality agreements. The challenge is polarization: when warnings get dismissed as a psyop by reflex, the practical question of verifiable safety standards gets lost.
We think the concrete calls for coordination between labs deserve a serious look on their merits, whatever camp they come from.
Sources