Open Source AI: A New Definition Raises Questions About Big Tech’s Openness
The definition of open source is evolving, and the new standards foropen source AI are shining a light on the practices of tech giants like Meta and Google. While they promote their large language models as open, a closer look revealsa more nuanced reality.
For decades, the Open Source Initiative (OSI) has been the leading authority on open source, defining and upholding its principles.Recognizing the growing importance of open source AI, the OSI embarked on a multi-year effort to define and standardize it. This involved a global consultation with a diverse group of experts, including researchers, lawyers, policymakers, activists, and representatives fromtech giants like Meta, Google, and Amazon.
The result of this collaboration is the first draft of the Open Source AI Definition (OSAID) 1.0, released on October 2nd. While OSI Executive Director Maffulli acknowledges this is a very rough 1.0, it has already sparked debate within the community. Some critics argue that the definition is too lenient and fails to adequately address the complexities of open source AI.
The OSAID 1.0 has been met with skepticism, particularly regarding the practices of Meta andGoogle. While their models are freely available, they are often subject to licensing restrictions that limit user modifications and access to the training data. This raises concerns about the true openness of these models, as researchers cannot independently verify their claims.
Avijit Ghosh, an applied policy researcher at Hugging Face,a platform for building and sharing AI models, highlights the misuse of the term open source for marketing purposes. Companies are known to misapply the term when marketing their models, he says. This can create a false sense of trust, even when researchers lack the ability to independently verify the model’strue open source nature.
The debate extends beyond legal and commercial concerns. Open source is fundamental to developer freedom and rights. True open source licenses empower developers, allowing them to freely use, modify, and share software. Without open source, developers lose this freedom, hindering innovation and progress.
The revised OSAID1.0, scheduled for release on October 28th, will likely address some of these concerns. However, the debate surrounding open source AI is far from over. The future of AI development hinges on a clear and robust definition of open source, ensuring transparency, collaboration, and the preservation of developer freedom.
References:
- Open Source AI Definition 1.0 Draft: https://opensource.org/ai/drafts/the-open-source-ai-definition-1-0-rc1
- InfoQ article: https://www.infoq.com/news/2024/10/open-source-ai-definition
Views: 0