OSI Proposes New Definition for Open Source AI to Address Ambiguity and Usage Restrictions

The Open Source Initiative (OSI) has released a new draft definition for "open source AI" in an effort to clarify the often ambiguous use of the term. This move comes as tech giants like Meta release large language models with usage restrictions while still labeling them as open source, sparking debates among free-software advocates.

What Changed?

The OSI's latest draft definition aims to establish clear criteria for what constitutes "open source" in the context of AI. This is particularly important as more companies release AI models with varying degrees of openness and usage restrictions.

Why It Matters

For practitioners, this clarification can have significant implications:

• Clarity in Licensing: Understanding the true openness of an AI model helps in making informed decisions about its use.
• Compliance: Ensures that developers and organizations comply with license terms, avoiding potential legal issues.
• Innovation: A clear definition can foster more collaborative and transparent development practices.

Key Points from the Draft

The draft emphasizes "four fundamental freedoms" inspired by the principles of free software:

• Use for Any Purpose: Users have permission to use the AI system without needing additional permissions.
• Study How It Works: The ability to inspect and understand the inner workings of the model.
• Modify for Any Purpose: Freedom to modify the AI system to suit specific needs.
• Share with or Without Modifications: Permission to distribute the modified version, fostering a community-driven development process.

Current Ambiguity

The ambiguity in what constitutes "open source" in AI is highlighted by examples like Meta's Llama 3 model:

• Llama 3: While freely available, it imposes license restrictions based on company size and content produced. This doesn't align with the traditional open source criteria as defined by the OSI.
• Flux: Another "open" AI image generator that is not truly open source due to usage restrictions.

Industry Response

To address these issues formally, the OSI has assembled a group of about 70 participants, including:

• Researchers
• Lawyers
• Policymakers
• Activists

Representatives from major tech companies like Meta, Google, and Amazon are also involved in this effort. The current draft (version 0.0.9) is the result of their collaborative work.

Implications for Practitioners

For developers and organizations using AI models:

• Evaluate Licenses Carefully: Ensure that any AI model you use complies with your project's requirements and legal obligations.
• Contribute to Open Source Projects: Engage in community-driven development by contributing to truly open source AI projects.
• Stay Informed: Keep up-to-date with the OSI's ongoing efforts and future definitions to make informed decisions.

Conclusion

The OSI's new draft definition for open source AI is a step towards clarifying the landscape of AI model licensing. By establishing clear criteria, it aims to promote transparency and foster a more collaborative development environment. As the debate continues, practitioners can benefit from a better understanding of what truly constitutes an open source AI model.