Purely moral argument but I don't see, how given that all AIs were trained through violating copyright, the weights ought to be anything but public domain

Posters: nosilverv. nosilverv

3 likes recorded by the source corpus.