OSI Launches Global Effort to Define Open Source AI Standards
What Open Source AI Standards Definition is and isn’t has no accepted definition. A nonprofit organization is now working to provide a common definition to guarantee the openness and use of AI technology.
Similar to the Open Source Definition, which specifies open source software, the Open Source Initiative (OSI) aims to provide an internationally recognized standard.
According to OSI executive director Stefano Maffulli, “AI is different from regular software and forces all stakeholders to review how the Open Source principles apply to this space.” “Everyone should be able to retain agency and control over technology, according to OSI.” We also understand that markets thrive when there are precise definitions that encourage openness, cooperation, and unrestricted invention.
Prominent AI experts like Yann LeCun think open source software is the way of the future, but the criteria of what constitutes a fully open AI system is still up for dispute.
Prominent IT companies including Google, Microsoft, and Meta have released open-source AI models called Phi-3, Gemma, and Llama 3, in that order. Nevertheless, some of these businesses neglect to release the underlying training data, even when they have made the model weights and underlying code public.
In a recent episode of the AI Business Podcast, Seth Dobrin, the founder and CEO of Qantm AI and a former chief AI officer of IBM, referred to this as pseudo-open source. Additionally, Julia Ferraioli, a machine learning expert at Amazon, stated during a February industry gathering that just because a technology is free.
The OSI seeks to define what open source AI actually is. In order to establish precise criteria about open source AI, the charity has been conducting research with pertinent stakeholders for the past two years.
After compiling what it called “a massive body of work,” OSI is now traveling the world to discuss its findings at a number of seminars before deciding on a possible definition.
“We’re going on a global roadshow to refine and validate the release candidate version of the open source AI definition, after spending almost two years gathering voices from all over the world to identify the principles of Open Source suitable for AI systems,” Maffuli stated.
Over the next months, OSI will be presenting their results at conferences in Paris, Lagos, Hong Kong, and Argentina.
Its U.S. seminars are scheduled for PyCon in Pittsburgh, New York, and Raleigh, North Carolina.
