ivan wrote: ↑Sat Jul 29, 2023 8:31 amChatGPT may be good at "transforming" (or plagiarizing) existing creative works
As you pointed out earlier, chatgpt is a tool, so it cannot plagierize. I don't mean to split hairs, but that sort of statement is what gives people the wrong idea about what AI is and what it can do. Also, it's still perfectly legal to summarize or restate, whether you use a computer or not.
ivan wrote: ↑Sat Jul 29, 2023 8:31 amHow do we know if its "training data" was obtained legally? We do not know if it was, because ChatGPT does not provide sources for its claims.
That doesn't matter. The burden of proof is on anyone who claims that openai or any chatgpt user broke the law. The latter don't have to prove that they're innocent.
As far as I know, no one has passed any laws specifically referring to large language models, so we're still dealing with the same copyright law I grew up with, which means that someone using chatgpt has the same rights as someone not using it. It's irrelevant. You still have to prove that a specific work violates a specific copyright, and considering how LLMs work, that's usually going to be impossible, unless you get a really friendly judge.
Of course new laws get passed every day, and someone could make using LLMs illegal or severely limit them. I think that would be foolish, and that country would be crippling themselves by doing so, but I'm sure a lot of people disagree.
Regardless, I can't imagine anyone being able to make a successful claim against the original poster's code. Unless someone can prove it's *very* similar to theirs, Giovifav or glitchapp can publish their results under any license.
As a wise man once said, never confuse morality with the law.
P.S. I would also argue that it's perfectly moral to publish works created by LLM, and I'm quite willing to discuss it if you like, but don't expect me to convince you if you're deeply against it.