For each open source AI license evaluated by OSI, which license criteria were either met or not met? How do licenses evaluated so far compare?
For context, the Linux Foundation Model Openness Tool (https://isitopen.ai/) indicates what model components are open or not open.
Is there a point-by-point way to to understand how open source AI licenses compare based on OSI’s criteria?
While there are some discussions about new licenses being created specifically for Open Source AI, the OSAID provides a checklist to identify what components are required to exercise the basic freedoms of Open Source AI.
Each required component should either be made available under an already existing OSI-Approved license or an OSI-Approved term. OSI-Approved term means that the OSI will review licenses and agreements to ensure that all materials are available under terms that conform with the Open Source Definition.
Thanks for the links Nick. Is there a detailed breakdown of each license evaluation, e.g., findings or deliberations leading up to approval decisions?
As of now, new licenses and terms are being evaluated following the standard OSI License Review Process.
Thanks Nick. It sounds like specific details of each license evaluation, including evaluation of licenses that were not approved, are not available. Is that correct?
Everything that is submitted for approval, whether approved or not, can be seen in the archives (which unfortunately aren’t indexed by license).
I think the issue is that because the OSAID is so new, there might be only one license ever submitted for approval against it.
I’m confused: You asked about licenses but mentioned the MOT for context. The MOT is a tool to for evaluating and classifying the completeness and openness of machine learning models. They don’t evaluate licenses.
Can you please clarify what you’re trying to achieve? Maybe describing your use case may help us offer more precise suggestions.
I found this presentation from @lf_matt_white which does present a comparison of 4 licenses (MIT, Apache 2.0, OpenRAIL-M, and OpenMDW) and how well they are suited for Open Source AI.
