Shengjia Zhao, co-creator of OpenAI’s ChatGPT, threatened to quit days after he was hired at Facebook parent Meta, while a handful of other new AI hires at Meta’s new Superintelligence Labs division have decided to leave after short tenures at the company, the Financial Times reported.
Zhao went so far as to sign paperwork to leave Meta and return to OpenAI, the report said. Shortly afterward he was given the title of chief scientist of the Superintelligence Labs effort, four unnamed sources said.
Wired earlier reported some details of Zhao’s threatened exit and other staff departures.
Short tenures
Ethan Knight, a machine-learning scientist, is leaving Meta after weeks in his post, while former OpenAI researcher Avi Verma went through Meta’s employee intake process but never showed up for his first day on the job, the report said.
Research scientist Rishabh Agarwal, who started at Meta in April, announced he was leaving the company in a social media post last week.
He said the pitch by Meta chief executive Mark Zuckerberg and AI executive Alexandr Wang was “compelling” but he “felt the pull to take on a different kind of risk”, without giving details.
A number of longer-serving Meta employees have also announced departures in recent days, including generative AI staff Chaya Nayak and Loredana Crisan, who had been at Meta for respectively nine and 10 years.
The lab is additionally considering cuts to the AI staff, a source told the FT, while the paper also cited a memo as stating that AI hiring would be frozen for the time being.
AI hires are paused except for business-critical roles, with Wang’s staff considering requested hires on a case-by-case basis as “we work through our strategy”, the memo reportedly said.
Turmoil
The staff turmoil is evidence of the rapid pace of change as Meta pumps vast amounts of cash into artificial intelligence, as it seeks to catch up to rivals including OpenAI and Google.
The initiative saw Zuckerberg personally approach AI staffers from competitors with offers nine-figure sign-on bonuses, while Meta hired Wang as part of a $14.3 billion (£10.3bn) investment into Wang’s data-labelling start-up Scale AI and also brought in former GitHub chief executive Nat Friedman.
In August, Meta announced a reorganisation of the new AI lab into four distinct teams, in the fourth major overhaul of its AI plans in the past six months.
Under the latest structure, Wang heads a cutting-edge unit called TBD, or To Be Determined, which includes the highest-profile hires.
Zuckerberg, known as a hands-on chief executive, and Wang have struggled to reach a timeline for reaching the AI division’s stated goal of attaining better-than-human “superintelligence”, with Zuckerberg urging the team to move faster, the FT reported.
Executives sidelined
The restructures have led to some veteran executives being sidelined, with Meta’s chief AI scientist Yann LeCun now reporting to Wang, while Ahmad Al-Dahle, who led Meta’s Llama and generative AI efforts earlier this year, not being named as head of any teams.
Chief product officer Chris Cox remains in his role, but no longer oversees generative AI, as Wang, who now heads the overall AI effort, reports directly to Zuckerberg.
Meta said that “some attrition” is normal for any large organisation. The company said the reported tension around efforts to achieve superintelligence was “manufactured”. It added that Cox remains “heavily involved” with its AI efforts, including overseeing recommendation systems.
Meta said Zhao was the scientific lead of the company’s superintelligence effort from the beginning, but it waited until the team was in place before formalising the title.
