In June, Mark Zuckerberg went to the final hail of Mary in the constantly intense artificial intelligence race: he turned a completely novel meta laboratory AI after a scale of $ 14.3 billion-then spent billions more employing some of the outstanding scientists and engineers in the industry.
A few months ahead, and Zuckerberg could recruit talent, but now the question is whether he will be able to stop him from slipping through his fingers.
To this end, we will have to wait and see. Finally, the finish changed the brand of the entire AI division into “Meta Superinteligence Labs”, the division of thousands of people. One team in this department-team, which we have heard of the most, filled by the leading AI researchers in the industry, whose task is achievement Superinteligence – is called “TBD laboratory”. From this team, only one energetic employee, Ethan Knight, so far, was confirmed by Meta spokesman Dave Arnold The Verge. Knight left less than a month after joining.
Of the other two, which allegedly left the TBD laboratory, Avi Verma and Rishabh Agarwal, they never really started in roles from the laboratory, confirmed Arnold. Other employees who have announced Their departures from the finish were not part of the TBD laboratory, but rather a full organization of superintelligence, including Rohan Varma and the director of meta products management in generative artificial intelligence, Chaya Nayak. Nayak left on OpenAi.
Employing the finish line for the Superinteligency team increased a few months ago, and its takeover of 49 percent share in AI Scale – an industry giant, which provided training data for companies such as OpenAI, Google, Microsoft and Meta. As part of the CEO Scale Agreement, Ai Alexandr Wang joined the finish line to direct the laboratory, and Meta employed Opeli, Deepmind, Anthropic and other companies. Most of the employees came from Opelai, from which the finish was supposedly trotted up to 10 employees, throwing some packages and equity, which, According to some reportsAdd to $ 300 million in four years. (Meta questions this number.)
But in recent months the employees of the AI industry said The Verge This money was not enough to attract some talents, and some rejected the scope and meta offers, especially in the case of high visibility TBD laboratory. In the full safety industry and enormous salaries, at least for engineers and researchers at a specific level, these employees prioritize the work that is in line with their values, regardless of these values (think: security AI, social effects, acceleration, etc.).
Last week, finish Headers Headers for sending notes to freeze employment and restructuring for his AI department. The company said that both are natural observations of employment processes in recent months. And although the finish is almost certainly tense in some aspects of its employment budget, this does not necessarily mean that the company is withdrawing or changing its strategy. When the technological giant publicly spends billions of creating novel employees and wants to put all his eggs in a superintelligence basket, he will organize and organize himself-especially if he tries to catch up with the world’s leading laboratories in the constantly intense AI arms race. For example, the company is apparently He no longer works in his AI Behemoth model, which did not meet the expectations of the newer models.
So what happens with freezing and restructuring of employment? Note viewed by The Verge About freezing of employing superintelligence, he states that because the finish line begins to plan its budget for 2026, and “with such a significant change and increase in our orgies in 2025”, the company “temporarily stops employment of all MSL [Meta Superintelligence Labs] Teams, except for business critical roles. “
The note further states that in the last two years there has been a “common practice in many orgs in the finish line and will allow leadership to carefully plan our employment growth in 2026, when we are working on our strategy.” Wang employees were found every week to assess the hired critical business “in individual cases.”
Wang last week Published WX: “We really invest only in Meta Superinteligence laboratory as a company. Every report to this contradictory is clearly wrong.”
Another note viewed by The VergePosted by Wang, he states that the Meta superintellity department will restructure and focus on three areas: research, product and infrastructure.
These three areas will be covered by four teams. First there is a TBD laboratory, which Wang calls “a small team focused on training and scaling large models to achieve superintelligence during initial training, reasoning and after training and examining new directions such as the OMNI model.”
Then there is a fundamental research laboratory AI, i.e. fair, which has been existing for some time, and the company seemed Deprioricjuj in recent months. His leader, Joelle Pineau, left in May. Now, Wang wrote in a note, Fair will become an “innovative engine for MSL, and we will try to integrate and scale many ideas and research projects from Fair to larger models carried out by TBD Lab.”
There is also the “Products & Applied Research” team, which “will bring our product-oriented research efforts to develop product development” and “include teams previously working on the assistant, voice, media, trust, incarnation and programmers in AI Tech”-Wang said. Finally, the “MSL Infra” team will focus on “accelerating AI research and production by building advanced infrastructure, optimized GPU clusters, comprehensive environments, infrastructure of data and programmers to support the latest research, products and development AI in Meta”, wrote Wang. Wang also wrote that the finish will dissolve its organization of the “Agi Foundation” and introduce these employees to one of the above teams.
IN post At X, Meta spokesman, Andy Stone, wrote that freezing and restructuring of employment were “mundane” updates, because “everything that happens is: a temporary break for employment during planning and forecasting, what we do regularly, and the structure of our new super -artistic efforts, after we introduced many new people on board.”
Update August 29: The article has been updated to reflect more AI META plans using specific models.
36 Comments