One other one in all OpenAI’s lead security researchers, Lilian Weng, introduced on Friday she is departing the startup. Weng served as VP of analysis and security since August, and earlier than that, was the pinnacle of the OpenAI’s security methods group.
In a publish on X, Weng mentioned that “after 7 years at OpenAI, I really feel able to reset and discover one thing new.” Weng mentioned her final day might be November fifteenth, however didn’t specify the place she is going to go subsequent.
“I made the extraordinarily tough choice to depart OpenAI,” mentioned Weng within the publish. “ what we’ve achieved, I’m so pleased with everybody on the Security Programs group and I’ve extraordinarily excessive confidence that the group will proceed thriving.”
Weng’s departure marks the newest in an extended string of AI security researchers, coverage researchers, and different executives who’ve exited the corporate within the final 12 months, and a number of other have accused OpenAI of prioritizing industrial merchandise over AI security. Weng joins Ilya Sutskever and Jan Leike – the leaders of OpenAI’s now dissolved Superalignment group, which tried to develop strategies to steer superintelligent AI methods – who additionally left the startup this 12 months to work on AI security elsewhere.
Weng first joined OpenAI in 2018, in accordance with her LinkedIn, engaged on the startup’s robotics group that ended up constructing a robotic hand that might resolve a Rubik’s dice – a activity that took two years to realize, in accordance with her publish.
As OpenAI began focusing extra on the GPT paradigm, so did Weng. The researcher transitioned to assist construct the startup’s utilized AI analysis group in 2021. Following the launch of GPT-4, Weng was tasked with making a devoted group to construct security methods for the startup in 2023. In the present day, OpenAI’s security methods unit has greater than 80 scientists, researchers, and coverage consultants, in accordance with Weng’s publish.
That’s lots of AI security of us, however many have raised issues round OpenAI’s deal with security because it tries to construct more and more highly effective AI methods. Miles Brundage, a longtime coverage researcher, left the startup in October and introduced that OpenAI was dissolving its AGI readiness group, which he had suggested. On the identical day, the New York Instances profiled a former OpenAI researcher, Suchir Balaji, who mentioned he left OpenAI as a result of he thought the startup’s expertise would deliver extra hurt than profit to society.
OpenAI tells TechCrunch that executives and security researchers are engaged on a transition to interchange Weng.
“We deeply admire Lilian’s contributions to breakthrough security analysis and constructing rigorous technical safeguards,” mentioned an OpenAI spokesperson in an emailed assertion. “We’re assured the Security Programs group will proceed taking part in a key position in making certain our methods are protected and dependable, serving a whole bunch of tens of millions of individuals globally.”
Different executives who’ve left OpenAI in current months embrace CTO Mira Murati, chief analysis officer Bob McGrew, and analysis VP Barret Zoph. In August, the distinguished researcher Andrej Karpathy and co-founder John Schulman additionally introduced they’d be leaving the startup. A few of these of us, together with Leike and Schulman, left to hitch an OpenAI competitor, Anthropic, whereas others have gone on to begin their very own ventures.