OpenAI CEO Sam Altman don reportedly tell staff say the company get three possible endgames for safe AI, and one of them na say one lab go race ahead and ‘melt everyone else’s GPUs.’ According to The Information, Altman call that kind outcome ‘terrible.’
The internal talk happen last week. The report say Altman describe OpenAI’s current safety playbook as ‘Pacing 1.0’—the approach wey dey ship models once dem pass safety tests and flag any concern. The three options wey him reportedly give staff na a global watchdog, a single super-lab merger, or one lab racing ahead of everybody else.
OpenAI now get one year to choose im own path, according to the report. The discussion come as Anthropic’s Jack Clark also talk about an intelligence cap and say the world get ‘a small number of years’ to act on safety. The report note say Altman’s scenarios seem aligned with im No. 1 rival.
A big team-up of top labs under one roof fit sound like correct plan, but the report add say like every pact so far, getting the rest of the field to agree no easy.
