‘Open-source’ has an updated definition — here’s what it is for now

Open Source AI Definition reaches release candidate stage: The Open Source Initiative (OSI) has released a Release Candidate (RC1) version of the Open Source AI Definition, marking a significant milestone in defining open-source standards for artificial intelligence systems.

• The RC1 version incorporates extensive community feedback gathered through town hall meetings, forum discussions, and in-person conversations across multiple countries.
• This release focuses on refining the definition of the “preferred form to make modifications to a machine learning system,” addressing key aspects of data sharing, code completeness, and legal considerations.

Key updates in the Release Candidate:

1. Data Information requirements:

   • The definition now clarifies that all training data must be shared and disclosed to the extent permitted by law.
   • Four types of data are described: open, public, obtainable, and unshareable, each with different legal requirements for sharing.

2. Code completeness:

   • RC1 emphasizes that the provided code must be comprehensive enough for downstream recipients to understand the training process.
   • This requirement aims to enhance transparency, security, and the ability to meaningfully fork AI systems.

3. Copyleft-like terms:

• The new text explicitly allows for copyleft-like terms to be applied to Code, Data Information, and Parameters, either individually or as bundled combinations.
• This provision anticipates potential scenarios where consortiums might distribute code and data bundles with specific legal terms.

Clarifying the role of Open Source in AI:

Scientific reproducibility and Open Source: The OSI emphasizes that the primary goal of Open Source AI is not scientific reproducibility, but rather to enable meaningful forking of AI systems.

• Open Source aims to provide the ability to study and modify systems without additional permissions, fostering innovation and improvement.
• While Open Source does not impede reproducibility, it does not explicitly require it; additional requirements can be added on top of Open Source principles to achieve reproducibility.

Forking in machine learning context:

• In the AI realm, forking refers to the ability to build systems that behave differently from their original state.
• This capability allows for various improvements, such as fixing security issues, enhancing behavior, and removing bias.

Next steps in the definition process:

Focus on refinement and documentation: With the release candidate phase underway, the OSI will shift its attention to:

• Addressing any major flaws or issues raised during the RC period.
• Refining accompanying documentation, including the Checklist and FAQ.
• Clarifying the basic requirement that data must be shared if legally possible.

Preparation for official release:

• The OSI aims to gather more endorsers for the Definition.
• Continued collection of feedback through various channels.
• Preparation of launch materials for the official release at the All Things Open event on October 28.

Broader implications: The Open Source AI Definition represents a crucial step in establishing standards for transparency and collaboration in AI development. As AI systems become increasingly influential in various sectors, this definition could play a pivotal role in shaping the future of open and accessible AI technologies.