On Friday, 18 November, we organized an in-person AI_Commons roundtable together with Creative Commons. The event took place at the Internet Archive in San Francisco. This one-day event brought together experts on open licensing and AI research, as well as open activists. We were joined, among others, by representatives of the Wikimedia Foundation, MLCommons, RAIL initiative, Github, and Flickr.
In the first part of the workshop, we discussed the case of the use of openly licensed photographs for face recognition training (based on the white paper that we recently published). We were particularly interested in understanding the role that open licensing frameworks play, and how these frameworks might need to be adjusted to mitigate potential downstream risks and harms related to the development of AI systems.
In the second part of the workshop, we discussed best practices in governing AI datasets, focusing on commons-based approaches.