I will admit this is not a simple case. That being said, if you've lived in the US (and are aware of local mores), but you're not American. you will have a different perspective on the US judicial system.

How is right to learn even relevant here? An LLM by definition cannot learn.

Where did I say analyzing a text should be restricted?

This was a preliminary judgment, he didn't actually rule on the piracy part. That part he deferred to an actual full trial.

The part about training being a copyright violation, though, he ruled against.

But AFAIK they actually didn't acquire the legal rights even to read the stuff they trained from. There were definitely cases of pirated books used to train models.

“I torrented all this music and movies to train my local ai models”

How is right to learn even relevant here? An LLM by definition cannot learn.

I literally quoted a relevant part of the judge's decision:

But Authors cannot rightly exclude anyone from using their works for training or learning as such.

This is an easy case. Using published works to train AI without paying for the right to do so is piracy. The judge making this determination is an idiot.

“I torrented all this music and movies to train my local ai models”

FTA:

Anthropic warned against “[t]he prospect of ruinous statutory damages—$150,000 times 5 million books”: that would mean $750 billion.

So part of their argument is actually that they stole so much that it would be impossible for them/anyone to pay restitution, therefore we should just let them off the hook.

FTA:

Anthropic warned against “[t]he prospect of ruinous statutory damages—$150,000 times 5 million books”: that would mean $750 billion.

So part of their argument is actually that they stole so much that it would be impossible for them/anyone to pay restitution, therefore we should just let them off the hook.

I am training my model on these 100,000 movies your honor.

It's pretty simple as I see it. You treat AI like a person. A person needs to go through legal channels to consume material, so piracy for AI training is as illegal as it would be for personal consumption. Consuming legally possessed copywritten material for "inspiration" or "study" is also fine for a person, so it is fine for AI training as well. Commercializing derivative works that infringes on copyright is illegal for a person, so it should be illegal for an AI as well. All produced materials, even those inspired by another piece of media, are permissible if not monetized, otherwise they need to be suitably transformative. That line can be hard to draw even when AI is not involved, but that is the legal standard for people, so it should be for AI as well. If I browse through Deviant Art and learn to draw similarly my favorite artists from their publically viewable works, and make a legally distinct cartoon mouse by hand in a style that is similar to someone else's and then I sell prints of that work, that is legal. The same should be the case for AI.

But! Scrutiny for AI should be much stricter given the inherent lack of true transformative creativity. And any AI that has used pirated materials should be penalized either by massive fines or by wiping their training and starting over with legally licensed or purchased or otherwise public domain materials only.

But AI is not a person. It's very weird idea to treat it like a person.

“I torrented all this music and movies to train my local ai models”

brb, training a 1-layer neural net so i can ask it to play Pixar films