1mo ago

To skirt copyright law, Anthropic bought and destroyed millions of physical books to train its "Claude" LLM

arstechnica.com

Anthropic destroyed millions of print books to build its AI models

12 comments

What a waste.
- Very on brand for AI companies. In a decade they'll be allowed to give AI agents legal personhood and the right to vote, but only if they first euthanize an equal number of orphans.
- I blame copyright law for requiring this.
I FEEL BAD WHEN BUYING PHYSICAL BOOKS BECAUSE IT FEELS WASTEFUL (usually use e-reader) AND THESE FUCKERS DO THIS.....
- I mean, using energy to display text on an eReader is wasteful too. Just being alive is wasteful. Everything is wasteful with that mindset. Guess what? The universe doesn’t care either way.
  
  No one should feel guilty about using the amount of energy it takes to power an eReader.
- Ultimately, Judge William Alsup ruled that this destructive scanning operation qualified as fair use—but only because Anthropic had legally purchased the books first, destroyed each print copy after scanning, and kept the digital files internally rather than distributing them. The judge compared the process to "conserv[ing] space" through format conversion and found it transformative.
  
  It's literally the process that allows digitized media to be safe to possess. Someone read the FBI warnings before movies on VHS. This is some corporate malicious compliance and what the law looks like when taken to an absurd extreme.
- That depends on whether you consider an LLM to be reading the text, or reproducing it.
  Outside of the kind of malfunctions caused by overfitting, like when the same text appears again and again in the training data, it's not difficult to construct an argument that an LLM does the former, not the latter.
  
  It’s rare a person on social media understands they turn the input into predictive weights, and do not selectively copy and paste out of them.
  
  models can and do sometimes produce verbatim copies of individual items in their training data, and more frequently produce outputs that are close enough to them that they would clearly constitute copyright infringement if a human produced them.
  the argument that models are not derivative works of their training data is absurd, and the fact that it is being accepted by courts is yet another confirmation that the "justice system" is anything but just and the law simply doesn't apply when there is enough money at stake.
- According to the article, Judge William Alsup disagrees

12 comments