Meet the Non-Profit Trying to Create a Definition for Open Source AI

There is no standard definition for what constitutes open source AI. One nonprofit is now trying to develop a universal definition to ensure transparency and accessibility of AI technologies.

The Open Source Initiative (OSI) hopes to create a globally accepted standard similar to the Open Source Definition which defines open source software.

“AI is different from regular software and forces all stakeholders to review how the Open Source principles apply to this space,” said Stefano Maffulli, OSI’s executive director. “OSI believes that everybody deserves to maintain agency and control of the technology. We also recognize that markets flourish when clear definitions promote transparency, collaboration and permissionless innovation.”

Leading AI figures like Yann LeCun believe the future of AI must be open source but the lack of a definition leaves the concept of what makes an AI system truly open up for debate.

Leading technology vendors including Microsoft, Google and Meta have open sourced AI models like Phi-3, Gemma and Llama 3, respectively. However, while these companies have published the model weights and underlying code, some fail to publish the underlying training data

Seth Dobrin, founder and CEO of Qantm AI and former chief AI officer of IBM, described this as pseudo-open source on a recent AI Business Podcast episode. And Julia Ferraioli, an Amazon machine learning strategist said at an industry event in February that just because a system is free, it does not make it open.

Related:Open Source vs. Closed Models: The True Cost of Running AI

OSI wants to clarify what truly constitutes open source AI. The nonprofit has spent the last two years conducting research with relevant stakeholders to create clear guidelines around open source AI. 

Having gathered what it described as “a massive body of work,” OSI is now taking its findings on the road, hosting a series of workshops around the globe to present its findings before finalizing the potential definition.

“After spending almost two years gathering voices from all over the world to identify the principles of Open Source suitable for AI systems, we’re embarking on a worldwide roadshow to refine and validate the release candidate version of the open source AI definition,” said Maffuli.

OSI will be bringing its findings to events in Paris, Lagos, Hong Kong and Argentina over the next few months.

Its U.S. workshops will take place in Pittsburgh at PyCon, as well as in New York and Raleigh, North Carolina.