E-Paper
Meta’s Elite AI Unit Sparks Tension With Old Guard
An influx of highly paid researchers has created new status divisions, even as some recent hires have already left their jobs.
Meta Platforms spent millions of dollars this summer recruiting AI stars. Now comes the hard part of making them work with their new colleagues.
Some of Mark Zuckerberg’s expensive new recruits have already defected to other AI labs. Existing employees have jockeyed for new spots within Meta’s restructured AI organization or lobbied for raises in light of the influx of highly paid new colleagues. At least one who was given a grant worth millions of dollars left anyway, saying they believed newcomers were still making multiples more.
The pay packages may be massive, and the company might be worth nearly $2 trillion. But at its core, Meta is facing a quintessential management problem: how to recruit and retain top-tier talent while keeping remaining employees satisfied, and maintaining harmony across the organization.
“If you don’t lay the groundwork culturally for bringing in these stars, you’re going to end up burning a bunch of them out and pissing them off, and a bunch of them are going to quit and you’re going to waste millions of dollars,” said Laszlo Bock, a tech industry adviser and former head of people operations at Google.
The new AI group’s most elite squad of AI researchers—members of the so-called TBD Lab—sit in an area of the company’s Menlo Park headquarters near Zuckerberg’s desk that requires special badge access for entry, according to people familiar with the matter. Their work is closely guarded, and the names of its members aren’t visible on the company’s internal organization chart. Other names in the organization are viewable.
Some employees pointed to the measures as reflecting new status distinctions within the company at a time when teams are competing for compute resources and a hiring freeze has made it difficult to fill roles without personal approval from Meta Chief AI Officer Alexandr Wang.
Aerial view of Meta’s headquarters, where the company’s new TBD Lab team works.
In a written statement, a Meta spokesman called reporting on the company’s AI operation “navel-gazing” and said, “This is yet another series of false, exaggerated or mischaracterized claims.”
Meta isn’t the only AI player dealing with this issue. The all-out war for talent has forced companies large and small to take extraordinary measures to keep their employees happy.
In August, OpenAI gave some of its research and engineering staff a one-time bonus, with some employees receiving millions of dollars, according to people familiar with the matter. Apple this summer lost a string of AI employees, including one of the leaders developing its AI models, to rival labs. And staffers at Safe Superintelligence, a startup co-founded by OpenAI co-founder Ilya Sutskever, are discouraged from mentioning the company on their LinkedIn profiles, in part to prevent other companies from trying to poach them.
Meta has succeeded in recruiting more than 50 people with Zuckerberg’s promise of abundant computing resources, alongside rich compensation. Of those, at least 21 people are from OpenAI, more than a dozen came from Google, a handful came from xAI and others from Apple, according to people familiar with the matter and public LinkedIn profiles.
Some Meta employees have had success leveraging competing offers to get a share of the money and prestige Zuckerberg has been lavishing on outside recruits. In July, a handful of employees from one of Meta’s AI infrastructure teams secured offers from former OpenAI executive Mira Murati’s new AI startup, Thinking Machines Lab.
Former OpenAI executive Mira Murati launched an AI startup called Thinking Machines Lab.
After the employees took the offers to Meta, their compensation was increased and they were moved to the TBD Lab team. A spokesman for Meta said the company had already been planning to move those staffers to the new team and adjust their compensation “regardless of any recruitment offers they may have received” and that the company hasn’t made counteroffers to employees threatening to leave.
While the recruiting frenzy has abated somewhat since Meta imposed its hiring freeze, the intensely competitive talent market has encouraged researchers to be choosy and ping pong around AI labs with ease.
Rishabh Agarwal, a Toronto-based AI researcher who joined Meta’s AI division in April, received an offer to join the superintelligence lab after asking Zuckerberg for a role, according to people familiar with the matter. He decided to leave for a startup called Periodic Labs, founded by a former OpenAI executive, after being told he would have to work in-person at Meta’s Menlo headquarters.
Two other new hires from OpenAI and xAI, Avi Verma and Ethan Knight, recently left for OpenAI, according to people familiar with the matter and their public LinkedIn profiles. Ruben Mayer, an executive who joined from Scale AI as a product manager, also left Meta for what he said were personal reasons. He hasn’t announced a new job.
Shengjia Zhao, a co-creator of ChatGPT, joined Meta in June, but decided within a week to return to OpenAI, resigning from Meta and signing paperwork to rejoin his former employer, according to people familiar with the matter. Meta managed to retain Zhao by offering him the title of chief scientist—and tripling his compensation.
A Meta spokesman said Zhao had been the new team’s scientific lead since Day One and that his role was formalized once the team had taken shape. Wired previously reported some details of Zhao’s near-departure.
Write to Meghan Bobrowsky at meghan.bobrowsky@wsj.com, Keach Hagey at Keach.Hagey@wsj.com and Berber Jin at berber.jin@wsj.com
