I'm absolutely sympathetic to the copyright concerns (the idea that some future of copilot will use my GPL licensed code to at the very least devalue my work is

I'm absolutely sympathetic to the copyright concerns (the idea that some future of copilot will use my GPL licensed code to at the very least devalue my work is very irritating!), but not only would this likely require changes to the way copyright works much more quickly than the law typically operates, I'm not even sure how enforceable it really is. You might be able to force OpenAI to hand over a part of dall-e's revenue, but how would that work with an open source model like stable diffusion that runs on consumer hardware? There's really no way to control how much/in what way its used. Even if you make it illegal to use James Gurney's work to train model without his explicit permission, people could just finetune it in private with essentially no way to determine whether they used works they didn't have permission for. The closest thing to the cassette recorder solution would be to put a tax on every computing sold, but that seems unlikely to me to be palatable to the general public. We definitely have very interesting times ahead of us.