In June, Mark Zuckerberg made a substantial move in the competitive field of artificial intelligence by launching a new Meta AI lab. This initiative followed his company’s acquisition of Scale AI for $14.3 billion, coupled with a hefty investment in recruiting top-notch researchers and engineers from the industry.
A few months later, although Zuckerberg has managed to bring in a talented workforce, the focus now shifts to whether he can retain them.
Meta’s AI division has undergone a rebranding to become “Meta Superintelligence Labs,” now encompassing thousands of employees. Among the teams, the most prominent is the “TBD Lab,” which comprises leading AI researchers dedicated to reaching the ambitious goal of superintelligence. According to Meta spokesperson Dave Arnold, only one team member, Ethan Knight, has departed from the lab since its inception, leaving just weeks after beginning his role.
Two others, Avi Verma and Rishabh Agarwal, initially rumored to have left the TBD Lab, did not actually assume roles within the team. Departures announced by other team members, such as Rohan Varma and Chaya Nayak, who was Meta’s director of product management for generative AI, were unrelated to TBD Lab, with Nayak opting to join OpenAI instead.
The search for top talent accelerated following Meta’s acquisition of a 49 percent stake in Scale AI, a significant provider of training data to tech giants like OpenAI, Google, Microsoft, and Meta itself. This partnership brought Scale AI CEO Alexandr Wang onboard to lead the lab, while also recruiting professionals from OpenAI, DeepMind, and Anthropic. Reports indicate that Meta made headlines by offering competitive pay packages and equity incentives that could total around $300 million over four years, a figure Meta has challenged.
However, some AI professionals disclosed to Technology News that financial incentives alone have not swayed all potential candidates, particularly for the visible roles within TBD Lab. In a landscape where job security and salaries are high for skilled engineers and researchers, many prioritize aligning their workplace with their values, whether that pertains to AI safety or societal impacts.
Recently, Meta made headlines by issuing memos regarding a hiring freeze and restructuring within its AI division. The company indicated that these changes are a normal progression following recent hiring activity. While Meta appears to be tightening its hiring budget, this does not imply a reversal of its strategic direction. With its substantial investment in talent acquisition aimed at Superintelligence, some pause for reevaluation is expected, especially in comparison to leading AI laboratories. Reports suggest that the firm has ceased efforts on its Behemoth AI model, which did not meet internal expectations, in favor of developing new models.
The memo concerning the hiring freeze highlighted that Meta is in the midst of planning its budget for 2026, hence the pause across all teams in Meta Superintelligence Labs, except for roles deemed critical for business.
This memo stated the hiring pause is a common practice within many Meta divisions over the past two years, allowing leadership to carefully strategize for future growth. Wang’s team has planned to convene weekly to assess essential hires on a case-by-case basis.
Wang recently shared on social media that Meta is committed to increasing investment in its Superintelligence Labs, refuting any contrary reports.
Further details from Wang revealed plans for restructuring within the Superintelligence division, prioritizing three key areas: research, product, and infrastructure. Four teams will cover these areas.
The first is the TBD Lab, characterized as a small team focused on training and scaling models to reach superintelligence across various stages, and exploring innovative approaches, including an omni model.
Meta’s Fundamental AI Research lab, known as FAIR, has existed for some time but has recently been deprioritized by the company. With the departure of its leader in May, the lab is set to evolve into an innovation driver for the Superintelligence Labs, integrating its research into larger modeling efforts managed by TBD Lab.
Additionally, the “Products & Applied Research” team aims to streamline product-oriented research with product development, incorporating teams previously focused on various AI technology areas. Finally, the “MSL Infra” group will enhance AI research capabilities through advanced infrastructure, including optimized GPU clusters and comprehensive environments to support state-of-the-art developments across Meta.
Wang also indicated that Meta will disband its “AGI Foundations” organization, reallocating those personnel into the newly restructured teams.
In a comment shared on social media, Meta spokesperson Andy Stone described the recent changes as routine, clarifying that the updates involve a momentary hiring pause while the company organizes its future plans and structures for its superintelligence objectives.
Update August 29th: The article was revised to include more details regarding Meta’s AI initiatives and specific models.
40 Comments
