I don't think there should be many (if any at all) copyright restrictions on training. This doesn't give AIs the right to violoate copyright laws in its output.

I don't think there should be many (if any at all) copyright restrictions on training. This doesn't give AIs the right to violoate copyright laws in its output. Plus stopping AIs from training may also stop just gathering metadata on materials like word frequency, genre or protaganist age.