Meta is shifting its focus on developing “superintelligence” after struggling to catch up with rival tech giants in the field of artificial intelligence. The company’s foray into AI research, led by Yann LeCun, has made notable contributions but lagged behind competitors like OpenAI and Google.
In 2013, Meta (then Facebook) established its AI research division, FAIR, to advance AI tools for core businesses like content moderation and image captioning. However, the company failed to develop AI as a standalone product. Now, with ChatGPT dominating the market, Meta has launched its flagship model, Llama 4, but it still trails behind competitors.
A benchmarking test showed that an experimental version of Llama 4 scored second globally, while the public release ranked only 32nd. Meanwhile, top AI labs have released new “reasoning” models with improved performance in advanced math and coding problems. Meta has yet to develop its own reasoning model using a new training paradigm. The company’s new focus on “superintelligence” suggests a significant revamp of its AI strategy to catch up with the competition.
Source: https://www.theatlantic.com/technology/archive/2025/07/meta-superintelligence-team/683607