From social network to superintelligence: the evolution of Meta's massive AI research engine

From social network to superintelligence: the evolution of Meta's massive AI research engine
Image: Intellect (generated)

In short: Born in 2013 as a small research group, Meta AI has transformed into a global powerhouse driving everything from open-source frameworks to controversial news-summarizing chatbots. As the company consolidates its efforts under a new superintelligence lab, its influence on how we interact with machines is expanding rapidly.

Meta AI began as Facebook AI Research (FAIR) in 2013, initially directed by Yann LeCun. The division quickly established itself as a major contributor to the field, releasing influential tools like the fastText library in 2016 and the PyTorch framework in 2017. These innovations helped standardize deep learning research across the industry. Following the 2021 rebranding of Facebook, Inc. to Meta Platforms, the team continued its work under the Fundamental AI Research banner, eventually expanding its footprint to include workspaces in cities like London, Paris, and Tel Aviv by 2025.

The scope of Meta's AI ambition is vast, spanning from large language models like Llama—which reached its fourth iteration by January 2026—to specialized hardware. To support these massive computational demands, Meta introduced the AI Research SuperCluster in 2022 and developed its own custom AI accelerators, the MTIA family. However, this rapid growth has not been without friction. The company has faced scrutiny over its training data sources, including reports that it utilized works from the pirate site LibGen in 2022, and has drawn legal and ethical concerns for its chatbot's practice of summarizing news without direct attribution or compensation.

In 2025, Meta reorganized its various AI research and product groups into the Meta Superintelligence Labs. This new structure oversees the Meta AI assistant, which is now integrated into social networking products and hardware like Ray-Ban Meta smartglasses. While the technology powers sophisticated applications—such as the Seamless models for real-time voice translation—it also faces challenges regarding accuracy and safety. For instance, the scientific language model Galactica was withdrawn just three days after its 2022 release due to the generation of inaccurate and racist content, highlighting the ongoing tension between rapid deployment and responsible AI development.

Opens the Intellect+ feed with this card first. Swipe up for the next one.

Sources

  • Summary by Intellect · source: Meta AI

To cite this card, link to https://intellectplus.co.uk/app/technology-meta-ai.

How we source, check and correct cards: Editorial standards.

General information, not professional advice. Spotted an error? Flag the card or see corrections and complaints.

Intellect Plus is a visual educational discovery platform for exploring fascinating facts and ideas. Brain Food, Not Brain Rot. What is Intellect Plus?

Related

Finding something worth knowing…