The Open Source Initiative, which is responsible for overseeing the open source definition, has embarked on an initiative to define Open Source AI, including a series of workshops, webinars, and extensive discussions. With the increasing use of AI by public administrations for delivering public services, it has become imperative to address the question of what it means for an AI system to be Open Source.
The Open Source Definition describes the rights that users of a piece of software must receive, but the software component of an AI system is quite small. Most of the system is data, and it’s not obvious how the Open Source Definition could be applied. To this end, it might be necessary to look at why we require that software be open source. What are the objectives, and how can we ensure these same benefits are possible with regard to the data component of AI systems.
The global drafting process commenced in June with a call for research papers, followed by the initiation of a webinar series. This set the stage for a kick-off meeting to establish an initial version of a working document. Subsequently, three community review sessions were conducted during conferences, accompanied by an ongoing online consultation that remains accessible.
OSI has been conducting a series of webinars on this subject, with three more happening this week on Tuesday 10, Wednesday 11, and Thursday 12 October. We strongly encourage you to attend the upcoming sessions and engage in broader discussions regarding the definition of open source AI.
We had the opportunity to converse with Stefano Maffulli, the Executive Director of OSI, who shared his insights regarding the importance of defining such systems.
“Defining open source AI is of paramount importance in today's rapidly evolving technological landscape. There's a real sense of urgency around this now, whether you're talking to lawmakers or data scientists. In the past, the open source community has missed the chance to define emerging technologies -- cloud computing is a great example -- and the OSI is determined not to let it happen again.
In relation to the public sector, defining open source AI holds immense significance. The OSI has convened a global conversation with stakeholders from academia, industry, policy makers and non-profit organisations to find a clear, standard definition for Open Source AI.
We hope that the public sector will reap the same benefits from Open Source AI that it has from Open Source in general - increased innovation, improved efficiency, strengthened sovereignty, more cross-sector and cross-country collaboration and reduced costs.”
The next step after these webinars will be the presentation of a draft definition followed by a collective feedback and working session at the All Things Open conference on 17 October. This session is in the USA, but keep an eye out for the upcoming feedback sessions that will be held in 2024.