Embed this noticeAether ??? (aether@poa.st)'s status on Friday, 14-Mar-2025 20:13:26 JST
Aether ???OpenAI warns that China will take the lead in stochastic garbage generators, colloquially called "AI", if they are not allowed free reign to ingest all copyrighted data regardless of the wishes of the rights holders, which curiously is not as one-sided and self-serving as it might seem at first glance.
Copyright law prevents you from making copies of protected works. It doesn't prevent you from reading them or learning from them. It doesn't mean you can't cite them, use facts from them, remake the ideas from them.
Anyone who has reached Uni will know this. That's the whole point of books, after all.
There are particularly egregious cases such as when Meta torrented 82TB of books and then not having the manners to seed afterwards. Strictly speaking, if you paid for the books, or borrowed them from a source that did, and you don't reproduce copies, you are complying with copyright law.
What OpenAI is asking for is federal clarification of what the law is due to a flood of varying state laws and district court decisions currently all differing views on the question.
The Ars commentariat helpfully clarifies this issue by being so stridently and consistently wrong about everything.
This is not the purpose of fair use. Even Weird Al acquired permission from copyright holders, and despite parody often being considered "commentary", though depending on how the parody relates to the work.
AI is not commenting on the work at all. It's just derivative, which isn't protected. These guys are so full of shit.
Think of it this way, a professor commenting on a work to educate his class is fair use, but making copies and handing it out to the class in order to "compete" with China by training students is not. He or his students have to buy the book. I don't see how this is different.