From the news of the Open Source Initiative or OSI providing a definition of what conditions AI open source must have, the outcome is that this definition makes many AI models that developers release claim to be popular open source ones, not fall under it. This includes Meta’s Llama as well.
Faith Eischen, a spokesperson for Meta, pointed out that the company agrees with OSI and has been partners in many issues, but does not agree with this definition of AI open source because there is still no single definition that can be used for the complex and rapidly evolving world of AI.
This may become a significant issue in the future, as in the past, the definition of what is open source by OSI has been used as a reference in legal cases, and this clarity helps developers who want to further develop open source projects to be confident that they can use them correctly. Clément Delangue, CEO of Hugging Face, also came out in support of OSI’s definition, stating that it helps make various frameworks clearer.
The condition that causes Meta’s Llama to not fall under OSI’s definition is that the AI model must provide enough training data, even if it does not have to share the data. This is a concern for major technology companies as they often do not disclose where they obtain data for training.
Source: The Verge
**TLDR:** The Open Source Initiative’s definition of AI open source conditions is causing debate with companies like Meta’s Llama not meeting the criteria due to the need for detailed training data without having to share it.
Leave a Comment