Meta restructures AI division to pursue Superintelligence

Meta has once again reorganised its artificial intelligence division, dissolving the recently formed Meta Superintelligence Labs (MSL) and creating four new, distinct teams. The move comes as the company intensifies its effort to compete with rivals like OpenAI and Google in the race to build AI that can surpass human capabilities.

The internal changes aim to streamline operations and bring more focus to the company's AI initiatives. This restructuring is also a response to internal tensions and the need to show a clear path forward after significant investment in talent and infrastructure.

Alexandr Wang, the former CEO of Scale AI, recently joined Meta as its chief AI officer. In an internal memo reported by Bloomberg, he stated that the company must organise around key areas to take superintelligence seriously. The new structure divides the AI group into specialised units.

The four teams

• A new lab dedicated to foundational models, including the company’s Llama series, led by Wang.

• The established Fundamental AI Research (FAIR) team, led by co-founder Robert Fergus, will continue its long-term research.

• A "Products and Applied Research" team, reportedly led by former GitHub CEO Nat Friedman, focused on integrating AI models into consumer products.

• An infrastructure group, MSL Infra, led by Aparna Ramani, manages data centres and computing hardware.

This shake-up follows a period of aggressive hiring. Meta spent millions to attract top researchers from rival companies, offering lucrative compensation. While the company has not announced formal layoffs, internal discussions suggest a potential downsizing of the AI division, which has grown to thousands of employees. Some executives, including Loredana Crisan, a former vice president in the generative AI group, are leaving the company.

Beyond the organisational changes, Meta's strategy is evolving. The company is now considering using third-party AI models to complement its in-house Llama technology. This approach could give Meta more flexibility and a broader set of tools to develop new products. The constant changes highlight the pressure Meta is under to translate its heavy investment in AI into tangible, competitive products.