Multi-agent risks go mainstream after AISI/OpenAI/Anthropic incidents, EA circles urged to think for themselves
sebkrier · x · 2026-09-11
- The retweeter recalls discussing multi-agent risks with @sebkrier long before the UK AISI, OpenAI, and Anthropic incidents made the topic hot, when such takes were dismissed as idiosyncratic.
- Now researchers are swarming the area, and the author uses the moment to criticize excessive deference to funders in EA and AI safety circles: convince funders of your position, not the reverse. Think for yourselves.
Related event: Multi-Agent Risks Research Goes Mainstream, Early Warners Vindicated(2 posts)→
More from AGI Musings
- The 1940s carpenter analogy: today's AI skeptics sound exactly like power-tool protesters — WolframRvnwlf · 2026-09-11
- Life Under the Algorithm: how algorithmic wages push workers into predictable poverty — AINowInstitute · 2026-09-11
- Gary Marcus: the AI bubble will pop, but doomers and bubble-skeptics are different groups — GaryMarcus · 2026-09-11
- AI safety debates like p(doom) have existed for a decade, long before IPO hype — Afinetheorem · 2026-09-11
- Mathematician littmath rewrites CMSA talk last minute to discuss AI and the profession's future — littmath · 2026-09-11
- LeClerc mocks doomers: p(doom) inversely correlated with testosterone — robleclerc · 2026-09-11