This feels like a discussion that is already out of date. Very early versions Copilot would reproduce Open Source code verbatim and that wasn't great, but in al

This feels like a discussion that is already out of date. Very early versions Copilot would reproduce Open Source code verbatim and that wasn't great, but in all the months of playing around with ChatGPT that never happened to me once. Quite the opposite. ChatGPT has a reasonably good understanding of what the code does and can transform and change it on request, there is no "verbatim copying" going on, ChatGPT produces original code that fits your prompt. There is still some risk that AI is used to circumvent copyright by feeding code you down own into the AI and have it rewrite it in a way that looks original. That however still requires a human with intend, the AI won't clone any substantially large program just by accident. This "risk" also goes both ways, just as AI can be used to "steal" Open Source, Open Source projects can use AI help to automate, reverse engineer and decompile proprietary code and data formats. So I consider that a win. And of course we are still very early days, AI will get a lot smarter and the accusation of copying will get ever more baseless. Even with StableDiffusion, where you can clearly see the impact the training data has on the final result, you'd have a very hard time finding any images that would violate copyright, as it's really just remixing ideas and concept. I really don't see how Open Source licenses can evolve here to address the problem. In the long run AI will make copyright as we know it largely meaningless.