元OpenAI安全担当者の証言、安全性より製品重視だった実態を明らかに
Elon MuskによるOpenAIへの訴訟は、木曜日にオークランド裁判所で社内対立を浮き彫りにした。元AGI Readiness責任者のRosie Campbellは、同社が彼女の安全チームを解散させ、研究より製品を優先したと証言した。彼女は、安全委員会の承認前にMicrosoftがインドでGPT-4を展開したことを重大な前例違反として挙げた。元取締役のTasha McCauleyは、Sam Altmanによる情報隠蔽のパターンについて言及し、これが2023年の解任につながったと述べた。裁判では、営利部門が人類に利益をもたらすというミッションを損なうかどうかが引き続き審査される。
Testimony in Elon Musk's lawsuit against OpenAI revealed that the company disbanded its AGI Readiness team and Super Alignment team in 2024. Rosie Campbell, a former member of the AGI Readiness team, testified that OpenAI shifted from a research-focused to a product-focused organization, compromising safety commitments. She specifically cited an incident where Microsoft deployed GPT-4 in India via Bing before the Deployment Safety Board evaluated it, arguing the company needed to set strong precedents for future, more powerful models.
Former board member Tasha McCauley testified regarding the board's lack of confidence in CEO Sam Altman, detailing instances where he allegedly withheld information or misled the board. These included not disclosing the public launch of ChatGPT and misrepresenting board dynamics regarding Helen Toner. These trust issues led the non-profit board to fire Altman in 2023, though he was reinstated after employee and Microsoft pressure.