> If AI training is piracy then all art made by humans is also piracy
That would only be true if human mental impressions were “fixed media” and therefore (potentially infringing) copies ubder copyright law the way data stored in electronic media in the course of AI training is.
Yes, this exactly. I don't know how people keep missing this.
There's no rule anywhere saying we have to treat computer programs as if they're humans and extend them the same rights. Why would we do that? Says who?
Also, we don't do that, not with AI. We're not talking about paying AI for their labor, or giving them digital housing or something. We only care about the human rights we can extract profits from, other than that these are digital slaves.
Which, I'm fine with digital slaves. They're bits on a computer. But then we need to do that, and we can't be doing this whole "welllll they're basically people" bit we're doing with learning.
I don't know about that slavery argument. If you had a human slave doing unpaid paintings, they would still have the full benefit of a fair use defense.
I don't think there's any "trying to have it both ways" with AI in this context. Copyright and labor laws are very different concepts.
> Copyright and labor laws are very different concepts.
I agree, with the similarity of course that both concepts are explicitly related to humans. Not machines or programs.
If we want to extend these concepts to machines or programs we can, but that's naturally complicated and there's a lot of questions about that. That, to me, needs to be a deliberate thing we do - not some foregone conclusion like people treat it. I mean, these people talk about AI fair-use as if it's obvious. It's not even obvious for humans...
The Windows source code is almost certainly part of the training data given the leaks. You can use LLMs trained on that to make your own OS if you like.
For the same reason that I think it is (1) a bad idea to use an LLM to make your own OS, (2) when the AI is good enough to do that, it won't matter which OSes are or are not open-sourced, nor which OS's source code was used to train a model, as by that point the AI can make any OS for peanuts anyway — given those two points, I think the copyright holders fighting against their stuff teaching AI are missing the forest for the trees.
That would only be true if human mental impressions were “fixed media” and therefore (potentially infringing) copies ubder copyright law the way data stored in electronic media in the course of AI training is.