Meta is reorganizing to chase the "superintelligence"

Mark Zuckerberg has put a lot of money on the table to poach top AI talent from companies like Apple to OpenAI, passing through startups such as Scale AI. Meta now employs dozens of researchers and AI specialists, whose mission is to develop the « superintelligence ».

After the metaverse, the new chimera of Meta

The « superintelligence » is a concept at least as vague as general artificial intelligence (AGI). The latter is supposed to represent an AI capable of understanding and performing any intellectual task that a human can accomplish. After much hype about AGI to raise the billions of dollars needed for developing its data centers, Sam Altman has become much more cautious. Probably because GPT-5 is still far off, despite the promises.

Read AI as smart as humans, it's not tomorrow's tomorrow

The industry has not yet managed to develop AGI, but that's not a problem. Now it's time for the new chimera: superintelligence, which refers to a hypothetical stage beyond AGI. In other words, an AI that would surpass human intelligence in almost all fields... This is the new obsession of Meta's CEO, whose latest Llama 4 models have not convinced.

« Superintelligence is here, and to take it seriously, we must organize around the key areas that will be essential to achieve it — research, product, and infrastructure », explained Alexandr Wang, former CEO of Scale AI, which Meta invested $1.4 billion in, and now head of AI within the social media giant. In the memo intercepted by Bloomberg, he details the new structure of this group, now named Meta Superintelligence Labs (MSL).

He will lead the TBD Lab, responsible for overseeing Meta's large language models, including Llama. The research and development lab FAIR remains in place to handle fundamental research. The Products and Applied Research team will be responsible for integrating the technologies developed by the TBD Lab into consumer products. Finally, MSL Infra will focus on the infrastructure necessary to train AI models.

This restructuring comes after several months of turmoil. Meta has abandoned its old Behemoth model, which was deemed too disappointing, and is starting from scratch. The company is even considering abandoning its open source creed, according to sources from the New York Times: some executives have mentioned the possibility of developing a future closed model, similar to OpenAI.

Behind the scenes, the restructuring is causing upheaval. Several historical figures have left, and Meta is also looking to reduce its workforce: its AI department now employs several thousand people, and some positions might be cut or reallocated.

Mark Zuckerberg is not planning to slow down. The company expects to spend up to 72 billion dollars this year, mainly to build new data centers and hire researchers. To justify these massive bets, Meta's CEO relies on a promise: that superintelligence will open « a new era of individual emancipation » and ensure the future of his digital empire.

👉🏻 Follow tech news in real time: add 01net to your Google sources, and subscribe to our WhatsApp channel.