On the contrary - the part that is problemmatic is the verbatim reproduction of copyrighted code. If that's fixed by a "minor refactoring" then there's no hill to die on. It's not AI code generation per se that's problemmatic - it's when it does things that are break current IP law.

If you want to debate expanding IP law - that's a different discussion and one I would be rather sceptical about. I'd prefer the that IP law in general was rolled back - not forward.

Which is a fine point of view. But it’s not the one that many (most?) detractors actually hold.

It would imply that you cannot offend open source licenses by doing something as simple as recreating a codebase in another language, thus eliminating the exact matches.

> Doomed to fall to a layer of minor refactoring

Not quite - there is a reason why https://wikipedia.org/wiki/Clean_room_design exists as a concept to workaround copyright, and the same concept could hold for ML models.

Clean room design? Did the models train on art the trainers drew themselves?