Incredibly narrow margin on this poll, with some insightful comments. Personally, it seems very much fair use to me, as long as the training data is not entirely from the same author and the model isn't created for competing with some author.

→ View original post on X — @reza_zadeh