April 8th. YesterdayNew YorkerIt published an in-depth survey that lasted 18 months and revealed for the first time many internal documents that had never been made public beforeSam AltmanI'm sorry。
Two core documents were reported:
The first was an internal memorandum of approximately 70 pages, compiled in secret by the former chief scientist, Ilya Sutskever, in the autumn of 2023, which collected a large amount of Slack information and personnel documents by means of a camera, with the first charge being “false”
Second, Dario Amodei, co-founder of Anthropic, has kept over 200 pages of personal notes, written directly, "Ottman's words are almost certainly fullshit" and "OpenAI's biggest problem is Sam himself."。
In terms of security commitments, the report reveals a typical case of "drawncakes": OpenAI announced in 2023 a "super-matched" team, publicly committed to the company 20% for AI security research, with only 1% to 2% in place and using the oldest chip cluster。
The team leader, Jan Leike, complained to the then CTO Mira Murati, and received a response that "this promise is not real." The team was dissolved in 2024。
A former OpenAI board member described two rare features of Ottman's coexistence to The New Yorker: "A strong desire to be loved by all" and "almost antisocial indifferent to the consequences of deceiving others."。
Many interviewees used the term "anti-social" to describe him without being guidedI don't know. Diane Yoon's judgment is slightly different:
He is not a makiavili-like villain; he is only too in his own faith to live in a world of self-constructing。
In the face of all the allegations, Ottman characterized his own transformation as "good-faith adaptation to a rapidly changing environment" and stated that the main flaw of the early years was "too much to avoid conflict". He said, "I can't change my character" during the board crisis, "I don't remember saying it."。