IT之家 · 智能时代T2·68 pts
- Former OpenAI safety lead highlights a contradiction between warning about risks and maintaining full-speed development internally.
- Despite CEO Altman's statements supporting controlled pacing of frontier AI, R&D teams remain under high-pressure, high-speed conditions.
- The article underscores dangers of 'recursive self-improvement,' echoing previous cautions from OpenAI's Chief Scientist.