The Open Source Initiative (OSI) is facing a challenge from tech giants like Meta, which is refusing to adhere to the OSI's definition of "open source" in the context of artificial intelligence (AI). According to OSI chief Stefano Maffulli, Meta is "bullying" the industry on the concept of open source, while others see it as a necessary pushback against the OSI's outdated definition.

The dispute centers around Meta's Llama large language model (LLM), which the company has labeled as "open source" despite restrictions that fall short of the OSI's Open Source Definition. While some open-source advocates are crying foul, the industry's response has been largely indifferent. As one developer noted, "If you're the average developer, what you really care about is capability: Does this [software] offer something unique and differentiated that's awesome that I need in my application."

Meta's stance is significant, given its history of contributing to open-source projects like Apache Cassandra, React, GraphQL, and PyTorch. The company's refusal to bow to the OSI's definition is seen by some as a necessary challenge to the organization's authority in the AI space. As one commentator noted, "Meta, through the Llama models, has done more for open source LLMs than just about anyone else."

The OSI's definition of open source in AI, released as the Open Source AI Definition 1.0, has been criticized for not requiring training data to be open. This omission has disappointed some of the OSI's most ardent supporters. Meanwhile, companies like Meta and OpenAI are releasing new code and calling it "open source" or "open weights" without much concern for the OSI's definition.

The question remains: does any of this matter? For developers, the answer is clear: what matters is software that is easily accessible and works. Sometimes that means open source, but in the age of cloud computing, it has just as often meant a solid API and a free or low-cost cloud service. As long as Meta's Llama offers a fast track to generative AI applications, developers will continue to use it, regardless of its license.

Still, the fact that Meta and OpenAI are nodding to open source suggests that openness does matter. It may be seen as marketing, but these companies recognize the value of being associated with open-source principles. And while Meta's license may not meet the OSI's definition, it is giving more than lip service to open source – at least for the 99.99% of developers who don't have more than 700 million monthly active users.

In the end, the dispute highlights the need for a more nuanced understanding of what "open source" means in the context of AI. As the industry continues to evolve, it's clear that the OSI's definition will need to adapt to keep pace. And if Meta isn't willing to let that happen, it's hard to blame them.