I don't know if even copying small pieces of code verbatim should mean anything.
Another example is the photo generation ML algorithms that exist. They generate photos of random "people" (imaginary AI-generated people) by using actual photos of real people. If one eye or nose is verbatim copied from the actual photo to the generated photo, is the entire output now illegal or plagiarism? One might argue it's just an eye, the rest of the picture is completely different, the original photographer doesn't need to grant permission for that use.
Any analogies we make with this, be it text generation, image generation, even video generation, seems like it falls under the same conclusion: so far we've thought all of this was perfectly fine. I don't see why code-generation is any different. A function is just a tiny part of a project. It's not necessarily more important than the composition of a photograph, or a phrase in a book. We as programmers assign meaning to it, we know it takes time to craft it and it might be unique, but likewise a novelist may have spent weeks on a specific 10 word phrase that was reproduced verbatim, in a text of 500 pages.
The more I look at this the more it seems copyright, and IP law in general, is the main problem. Copyleft and OS licenses wouldn't be needed if it wasn't for the aggressive nature of IP law. I don't see the need to defend far more strict interpretations of it because it has now touched our field.
Another example is the photo generation ML algorithms that exist. They generate photos of random "people" (imaginary AI-generated people) by using actual photos of real people. If one eye or nose is verbatim copied from the actual photo to the generated photo, is the entire output now illegal or plagiarism? One might argue it's just an eye, the rest of the picture is completely different, the original photographer doesn't need to grant permission for that use.
Any analogies we make with this, be it text generation, image generation, even video generation, seems like it falls under the same conclusion: so far we've thought all of this was perfectly fine. I don't see why code-generation is any different. A function is just a tiny part of a project. It's not necessarily more important than the composition of a photograph, or a phrase in a book. We as programmers assign meaning to it, we know it takes time to craft it and it might be unique, but likewise a novelist may have spent weeks on a specific 10 word phrase that was reproduced verbatim, in a text of 500 pages.
The more I look at this the more it seems copyright, and IP law in general, is the main problem. Copyleft and OS licenses wouldn't be needed if it wasn't for the aggressive nature of IP law. I don't see the need to defend far more strict interpretations of it because it has now touched our field.