Aral Balkan
@josias Sounds to me like we need new clauses added to “share alike” free/open licenses (including non-code ones like Creative Commons) that specifically disallow use in generating proprietary machine learning models.
Jun 30, 2021 at 3:01pmWeb
Show threadcybercowJun 30, 2021
@aral @josias One could argue that tainingdata is so vital to the functioning of an ki that it has to be seen as source code. So if this argumentation holds, it actually is a GPL violation. But then someone would have to bring this claim through a judge.
Show threadØyvind KolåsJul 1, 2021
@aral it does not look like the people making AI training datasets consider licenses as applicable to their actions, for grave ethics violations involving CC licensed (and not) photos, consider https://exposing.ai/megaface/
Exposing.ai: MegaFace

MegaFace is a dataset of over 4 million faces used benchmarking and developing face recognition technologies