The solution is simple. We need an updated GPL license that states that the code cant be used in training ais unless the data model is also open source. A coordinated update of all major open source projects and the issue is sorted as it will force the ai folks to open source their models. Or else they’ll have to stick with generating funny cat pictures.