A long-time member of the Debian community known as “Samj” has expressed opposition to the new definition of Open Source AI. Samj believes that the new definition is a departure from the original principles of Open Source, which were established based on the DEBIAN manuals for free software developed by Bruce Perevese. Critics argue that Open Source should adhere to the previous version 1.9, with any changes requiring full community consent through open and transparent processes.
The dissatisfaction with the new definition surfaced following discussions on the DEBIAN forums. Many developers are concerned that the new definition could allow for the concealment of data used to train AI, creating a “data barrier” that could benefit monopolies and hinder the development of alternative AI versions. Some view this as a significant setback for the broader free software community.
The issue of closed data is increasingly significant. Research indicates that many Open Source models, while claiming to be open, actually have inaccessible training data, expensive resources, and complex methodologies. Such projects are seen as potentially strengthening the position of major corporations rather than fostering technological accessibility.
Samj argues that training data is equivalent to source code for AI, warning that changing the definition of Open Source could have adverse effects on real projects. However, the Open Source Initiative (OSI) disagrees. In an official statement, OSI representatives clarify that training data is distinct from source code, highlighting instances where regulations prohibit the open sharing of certain data, such as healthcare information or indigenous knowledge protected by legal frameworks.
While some advocates for full data accessibility argue that transparency and replicability of AI models could suffer without it, OSI proposes a more flexible approach. They suggest that users should have the choice of using open or closed data for training AI, allowing for the development of Open Source AI even in sectors with heightened confidentiality requirements, such as healthcare.