It has been. Bartz v. Anthropic already determined the the act of training an AI is transformative and thus fair use. The only question is under what conditions were the materials acquired. Paid for and downloaded? Ok. Sucked of some random torrent on the high-seas, well... there's a reason Anthropic settled Bartz for $1.5bn, despite having all rulings about how AI is fair use in its favour.

They are also not storing books. The training of AI has been ruled as transformative already in Bartz v. Anthropic. The only question was under what avenue the materials were acquired.

So you are infringing copyright every time you stream a Netflix movie? I mean buffering is copying right? Wrong. The "Normal Operation" standard applies, which is to say if the copyrighted work can't be legally used without buffering on disk or in RAM or doing any kind of memory operation through the course of its use then that operation isn't copyright infringement.

Also the question of whether AI training is copyright infringement has also been ruled on, and so far it's been considered "spectacularly transformative" (judge's words, not mine) and thus fair use.

Except they have. Not just generically either, but specifically on how Anthropic trained AI models. See Bartz, et al. v. Anthropic PBC.

The court ruled here that training an AI model is not copying, and used the words "spectacularly transformative". Ultimately the parties settled because Anthropic used a pirated book data base. We have definitive ruling that training AI is fair use.

The open question is, how was material required. Round Hill said "without permission", but fair use says you don't need permission to do the training. Then the question is, was the music source acquired legally. If it was, Round Hill has no leg to stand on. If Anthropic downloaded a trove on bittorrent then they may have a case.

And yet you are wrong on both accounts, both in the physical analogy and in the legal sense.

a) An AI does not copy and store the original text, but rather trains a neural network with weights that represent the learned material. This is literally modelled on human learning processes and how our brains work, it's why the process was given the name in the first place, it's similar to how neurons work.

b) This has been definitively ruled on directly for the training process of the company in question by a different plaintiff: Bartz, et al. v. Anthropic PBC

"The court held that Anthropic’s use of copyrighted books for LLM training was “spectacularly” transformative and qualified as fair use. The training process involved analyzing statistical relationships between text fragments to enable the LLM to generate new, original text, rather than reproducing or distributing the original works. The court analogized this process to human learning, emphasizing that copyright law does not extend to the methods, concepts, or principles embodied in a work."

Emphasis mine. Both the development of AI systems which modelled them after human learning, as well as the courts which in their ruling analogised it with human learning disagreed with you. And even if your want to argue against the logic, the legal system doesn't care, a precedent that AI training is learning and not copying has been set.

Unfortunately we don't really. The standard for what is a derivative work is substantial recognizable elements. It doesn't specify exactly how you get there.

I'm not trying to convince anyone to change their terminology, but the aptness of the term is all on the conceptual side.

We have a massive shortage why is nobody stepping up to make more supply? And it's not a short-term shortage everyone knew this was coming and we have known for a couple of years and we knew that was going to get worse and the absolute best case scenario is that it starts to get better sometime around 2029 or 2030.

If capitalism can't even solve something this fucking simple what is even the point?

Of course the actual problem is capitalism is doing exactly what it does when it's not regulated. Nobody wants to enter the market because they know the big three can anytime they want drastically increase production within 6 months to a year and then run them out of business. Even if you wanted to enter the memory business no bank would loan you the money because they know you're just going to lose it all.

Capitalism needs a referee and US dumbasses fired the ref.

Drones are really good at handling gusts. No other kind of active aircraft comes close. A bigger problem is market size

Where will I get a SBC that fails after three months now?

Except that they're not bringing in nearly enough revenue to sustain their operations. There's going to be quite a shake up in the AI realm once the venture capital faucet turns off. It was pretty funny to watch that happen with the dockless rental scooters. Downtown used to be full of the damn things from various companies, all hoping to cash in on the gold rush. Now, every once in awhile you might see a dilapidated Bird scooter or two if you really keep your eyes out for them, but it's mostly just Limes now.

For well over a billion dollars should be something you read once or twice in your lifetime not daily news.

Market consolidation is killing us all. And we're too much of a bunch of pussies and or idiots to do anything about it.

So you are bitching about the cost, and when the cost is shown to be outweighed by the benefit over the lifetime of the system, you come back with money doesn't matter ??

Seriously?

Are you expecting me to look into your bank loan opportunities for you?

Go find out.

Are we forgetting gas prices? I've got some Republicans in my family and that's all they've been complaining about lately.

It's actually a nice change from the usual culture war stuff.

The character limit on Twitter (unless you have a paid subscription, yuck) also pretty much guarantees that all debates ultimately devolve into insults, because it's a lot easier to just call someone's mom a ho, rather than systematically attack all the flaws in their argument. Especially if you're typing on a phone.