Altman reportedly pitched three AI safety endgames to OpenAI staff
Original title: Sam Altman reportedly pitched three endgames for AI safety, one involving a single firm melting everyone else's GPUs
A reported third scenario has one lab withhold compute from rivals; Altman raised it only for completeness.
According to The Information, OpenAI CEO Sam Altman told employees last week about three approaches to governing advanced AI. One is an IAEA-style inspectorate; a second, which he called the Demis Plan, would put major labs under a CERN-like entity that supplies models to cooperating countries. A third, the Eliezer Plan, imagines one developer gaining enough power to deny compute to everyone else. The report says he raised that case only for completeness.
Unconfirmed: The remarks are secondhand, and one source read the Eliezer Plan comment as a jab tied to Anthropic.
- The account cites The Information and people familiar with internal remarks.
- He named the CERN-style option the Demis Plan and the extreme case the Eliezer Plan.
- OpenAI, Google, and Anthropic are building a Standards Authority for Frontier AI.
- In June the US imposed export controls on Anthropic over Fable jailbreak fears.
- Organizations
- OpenAIAnthropicGoogle DeepMind
- Countries
- United StatesChina
From the original article
2 min read at the-decoder.com
OpenAI CEO Sam Altman laid out three paths to safe AI in an internal presentation: an international watchdog modeled after the IAEA, a merger of the major labs into one entity, or a single company so powerful it could shut down everyone else's compute. He called the last option "terrible."
By Manuel Uth · the-decoder.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.