Meta Platforms Criticized for Using Its Brand in Open-Source AI

Meta Platforms has faced criticism for its use of branding in open-source AI. How will this controversy affect the company's future?

Meta Platforms Faces Criticism Over Mislabeling AI as Open Source

The Open Source Initiative (OSI), a pivotal software standards organization, has sharply criticized Meta Platforms for misleading the public by labeling its Llama models as open source. OSI argues that this designation is deceptive and muddles the true meaning of open source.

The OSI, claiming to have coined the term “open source” about 25 years ago and currently its steward, told the Financial Times that Meta Platforms is distorting the definition and confusing the public about what truly constitutes open source.

Open Source Requirements

To qualify as open source, the OSI insists that companies must:

• Disclose the training algorithms and other software used in their development,
• Be transparent about the “model weights,” a numeric parameter that defines the strength of the signal between neural networks.

OSI plans to issue guidelines on this matter next week.

Stefano Maffulli, OSI’s Executive Director, described Meta’s open source labeling of its AI as “extremely damaging” at a time when regulators, including the European Commission, are advocating for “genuine open source technologies.”

Google, Microsoft, and IBM and the Open Source Dilemma

Maffulli pointed out that Google and Microsoft have stopped promoting their models as open source, but discussions with Meta Platforms “have not led to a similar outcome.”

Dario Gil, IBM’s Research Director, also criticized Meta Platforms for a lack of transparency in their technical publications. However, he noted that Meta’s models provide a better alternative to “black box models,” which do not publicly disclose their internal workings.

Meta Platforms’ Stance and the Future of Llama

Moreover, the license used by Meta Platforms also prevents competitors from using its family of models, contradicting the open source ethos, despite the fact that anyone can download the systems for free, according to the FT.

In response to the allegations, Meta Platforms told the FT: “Existing open source definitions for software do not cover the complexities of today’s rapidly evolving AI models. We are committed to continuing to work with the industry on new definitions to serve everyone safely and responsibly within the AI community.”

Meta Platforms made its Llama 2 model accessible in 2023, noting at the time that it would enable “a generation of developers and researchers” to test its models. Today, its Llama models have been downloaded over 400 million times.