Software patents may be what saves the field. Currently copyright as an IP protection for software has been rendered near meaningless by AI. If I can get one AI to reverse engineer Photoshop, and another to implement the produced spec, I have a copyright-clean version of Photoshop.
Patents protect the idea of a particular program. And in the post-AI world, it's the idea that holds the value, not the implementation. Implementations are cheap now.
Another alternative to preventing the entire software market from collapsing is to revive the Whelan v. Jaslow copyright standard, under which the abstract "structure, sequence, and organization" of a program is eligible for copyright. But programmers would bristle at that almost as much as robust patent protection for software. Until, that is, the money comes in with mouths to feed and mortgages to pay...
You don't know what code the A.I has been trained on. It may very well be reproducing actual Copyrighted material it has been trained on.
This is the major reason why most large open-source projects are banning big code contributions made with A.I
> If I can get one AI to reverse engineer Photoshop, and another to implement the produced spec, I have a copyright-clean version of Photoshop.
You won't get this, otherwise we had tons of LLM written photoshop alternatives already. And if it would work, that would be a good thing.