Meta is reportedly forming an ‘AI Superintelligence’ team

Meta Chef Mark Zuckerberg has recruited experts to join a team he assembles to make a superintendent of AI, according to Bloomberg And The New York Times. Zuckerberg would have discussed potential recruits with other senior business leaders in a WhatsApp cat nicknamed “recruitment party”. He would have a personal list of recruits, including AI researchers, infrastructure engineers and other entrepreneurs. Zuckerberg invited them to have lunch and dinner at his home in California to bring them to join his team in the last month.

Currently, the immediate objective of AI societies is to achieve a real artificial general intelligence (AG), in which a machine has an intelligence at the level of man and can accomplish any task that a human can do. Superintelligence is a step beyond that. A AI system With superintelligence is supposed to have intellectual powers far beyond all human people.

Time Said Zuckerberg has already operated Alexandr Wang, the founder of the AI ​​Startup scale to join the new team. Meta plans to invest billions of dollars in Wang’s company, which provides other AI companies with data to train their models. The agreement will also bring the other SCALE employees aboard Meta, although it is not clear if one of them also joins the new team.

Meta also offered dozens of experts in AI from other companies, including Google and I openedCompensation packages worth seven to nine -digit to reach the team. Some, according to Time, have already accepted. In his argument for potential recruits, Zuckerberg apparently declared that his business advertising activities can finance his own AI development, even if it costs dozens of billions of dollars, unlike competitors who must first collect funds.

Zuckerberg, Bloomberg Said, decided to supervise the recruitment itself due to the frustration concerning the quality of the public’s response to the large language model of Meta Llama 4. Llama 4 was not well received, and the criticisms argued that the metal has overvalued but underestimated. The company also had to delay the release of its “Behemoth” model Llama 4that the company has caused outclass “GPT-4.5, Claude Sonnet 3.7 and Gemini 2.0 Pro on several benchmarks STEM.”