Some of the world’s largest music publishers filed a blockbuster lawsuit against Anthropic late Friday night, alleging “one of the largest and most blatant ongoing thefts of intellectual property in history.”
Some of the world’s largest music publishers filed a blockbuster lawsuit against Anthropic late Friday night, alleging “one of the largest and most blatant ongoing thefts of intellectual property in history.”
I don’t see how this would go any differently to the book lawsuit they settled.
With that lawsuit, the court found that training the LLM on books is covered under fair use, as long as the books are legally acquired and they do not make illegal copies. (obviously, they can’t use data fron Anna’s Archive any more)
The argument is that them training on the books does not impact sales of the books themselves - someone that was going to buy the book would still buy it even if an LLM was trained on it, just like they’d still buy it even if summaries/CliffNotes and reviews exist. The training is considered transformative as the LLM learns the relationship between the words rather than retaining the text of the books as-is, and transformative use is allowed under fair use.
This is why the AI companies are destroying books - they have to destroy any books they scan in order to comply with US copyright law. Transforming the book from one form to another (eg from print to digital) is OK as long as only one copy remains.
I’m not saying I agree or disagree, just that this is the current state of copyright caselaw in the USA.
Having said that, the music industry is significantly more litigious than the book industry, so we’ll see what happens…
Which is what humans do when learning something using existing works without paying the artist. People tend to forget that the goal of neural networks was to programmatically mimic the way the human brain works.
Of course the mechanisms behind it are different, but they’d have to be because it’s silicon vs carbon based processing.
A couple of things I think are slightly off:
I always thought there was in obvious win where companies doing this could be forced to archive the scan publicly (after some period of time). They’d be able to use it/sell it for training, but we wouldn’t lose rare books forever.
That’s definitely one reason, but the copyright argument is also a part of it. The court explicitly said that their digitization is legal only if does not increase the number of copies of the book.
They have to keep the digital copy of the book because they add it to the training corpus for the LLM. Selling or donating the original physical book after doing that would void the fair use argument.
One of the tests that determines if it’s fair use or not is whether it can serve as a replacement for the original book. Pirated copies can, which is why they’re illegal. A summary like CliffNotes can’t. Even if the LLM can reproduce long passages, you can’t do that reliably (like you said) and it won’t work for all books.
I definitely agree with this. I think copyright law needs to be modernized to handle cases like this. I think the AI companies should be allowed to donate the digital copy to a library (like the Internet Archive) while still being allowed to keep their copy in their training corpus.
I think copyright law already needed to be modernised… and the way I think it could be done well is consistent with AI training needs. So the whole thing doesn’t really bother me. I’m surprised this angle incenses Lemmy so much because I’d have thought we’d on the whole by very anti-copyright in its current form.