Key Takeaways
1. Meta is reorganizing its Superintelligence Labs into four groups: research, superintelligence, products, and infrastructure to improve development speed and accountability.
2. The company has made significant investments, including $14.3 billion in Scale AI and plans for $72 billion in capital spending for 2023, focusing on data centers and recruitment.
3. Leadership is rethinking strategies, including potentially moving toward a closed model for their flagship AI, diverging from their traditional open-source approach.
4. Key personnel changes include hiring Nat Friedman and Daniel Gross for new product features, while some former researchers have left for other companies.
5. The restructuring aims to expedite decision-making and product delivery, marking a final major change as the company focuses on advancing superintelligence despite significant costs.
Meta is reorganizing its Superintelligence Labs into four distinct groups: research, superintelligence, products, and infrastructure. This change aims to boost development speed and enhance accountability within its AI framework. According to leadership, this restructuring is designed to simplify operations and refocus efforts after a period of uncertainty regarding the company’s strategy for advanced models. Some executives might depart, and the division could experience additional downsizing or role changes.
Major Investments and Plans
In June, Meta poured billions into Scale AI and brought in its CEO, Alexandr Wang, to spearhead the superintelligence initiative. Reports indicate that this investment was around $14.3 billion, with a potential of $72 billion allocated for capital spending in 2023, primarily for data centers and recruitment.
Rethinking Strategies
The company’s leaders are reconsidering their model strategies. They have been looking into licensing external systems and discussing whether the upcoming flagship model should be “closed,” which diverges from Meta’s traditional open-source approach. The firm has discontinued the previously planned frontier model, Behemoth, due to lackluster results and has begun anew under Wang’s direction. Hiring has surged, with offers exceeding $100 million, creating some tensions between new hires and established researchers.
Nat Friedman, who previously served as GitHub’s CEO, alongside Daniel Gross, ex-leader at Safe Superintelligence, will guide new product features and applied research. Rob Fergus, a co-founder of FAIR, will keep overseeing core research initiatives. Joelle Pineau and Angela Fan, both former Meta researchers, have moved on to roles at Cohere and OpenAI, respectively. Loredana Crisan, who was the vice president of generative AI, is now the chief design officer at Figma. Ahmad Al-Dahle and Amir Frenkel are now under Wang’s leadership for strategic AI efforts.
Future Direction
This restructuring comes after stalled advancements and reflects a renewed focus on superintelligence. Leadership describes it as the final major structural change for the time being, aimed at expediting decision-making, reducing excess, and delivering products more quickly, despite ongoing high expenditures.
Source:
Link
Leave a Reply