The debate around open source AI has sparked controversy in the artificial intelligence space, with the Open Source Initiative (OSI) working to address the issue through a global effort. The OSI’s executive director, Stefano Maffulli, has been leading the charge to define open source AI, but it’s a complex task. One of the main challenges is that AI models, such as neural network weights, aren’t comparable to software source code. The OSI has come up with a draft definition, which includes three core parts: the preamble, the Open Source AI Definition, and a checklist of required components. However, the definition is still evolving, and the OSI is seeking diverse input from around the world. The organization is also working to diversify its funding, recently securing a grant from the Sloan Foundation. The issue is further complicated by companies like Meta, which has restrictions in place for its Llama models, despite calling them open source. As the OSI works towards a stable definition, it’s clear that the notion of open source AI will continue to stir debate in the years to come.

Defining Open Source AI
The OSI is trying to capture the fundamental differences between open source software and open source AI, which are not like-for-like comparable.
1–2 minutes










