- OpenAI researchers Daniel Kokotajlo and William Saunders recently left the company for ChatGPT.
- Kokotajlo said on a forum that he didn’t think OpenAI would “behave responsibly in the AGI era.”
- Kokotajlo was part of the governance team and Saunders worked on OpenAI’s Superalignment team.
Two OpenAI employees who worked on security and governance recently resigned from the company behind ChatGPT.
Daniel Kokotajlo left last month and William Saunders left OpenAI in February. The timing of their departures was confirmed by two people familiar with the matter. The people asked to remain anonymous to discuss the departures, but their identities are known to Business Insider.
Kokotajlo, who worked on the governance team, is listed as an adversarial tester of GPT-4, which launched in March last year. Saunders has been with the Alignment team, now the Superalignment team, since 2021.
Kokotajlo wrote on his profile page on the online forum LessWrong that he resigned “because he lost confidence that he would behave responsibly during the AGI era.”
In another post on the platform in April, he partially explained one of the reasons for his decision to leave. He also participated in a discussion about suspending AGI development.
“I think most of those advocating a pause are trying to argue against a ‘selective pause’ and a true pause that would apply to large labs that are at the forefront of progress,” Kokotajlo wrote .
He added: “The current overton window unfortunately appears focused on a combination of assessments and mitigations which poses a high risk of regulatory capture by IMO (i.e. resulting in a selective pause which does not does not apply to large companies that most need a break). !) My disillusionment with this is part of why I left OpenAI.
Saunders said in a comment on his LessWrong profile page that he resigned that month after three years at maker ChatGPT.
The Superalignment team, initially led by Ilya Sutskever and Jan Leike, is responsible for implementing protective measures against artificial general intelligence. (AGI) goes rogue.
Sutskever and Leike have already predicted that AGI could happen within a decade. It’s unclear whether Sutskever is still at OpenAI after his part in Sam Altman’s brief ouster as CEO last year.
Saunders was also responsible for the interpretability team, which researches how to secure AGI and examines how and why models behave the way they do. He is co-author of several articles on AI models.
The resignations of Kokotajlo and Saunders come amid other departures at OpenAI. Two executives, Diane Yoon and Chris Clark, resigned last week, The Information reported. Yoon was vice president of personnel and Clark was responsible for strategic and nonprofit initiatives.
OpenAI also parted ways with researchers Leopold Aschenbrenner and Pavel Izmailov, according to another report from The Information last month.
OpenAI, Kokotajlo and Saunders did not respond to Business Insider’s requests for comment.
Do you work for OpenAI? Do you have any advice? Contact this journalist at jmann@businessinsider.com for a non-professional device.
businessinsider
