Hacker News Reader

Comment by jacquesm | Hacker News Reader

I don't recall giving them permission and yes, copyright is a thing.

Copyright is grossly overbroad. Corporations should not get to own our collective culture for life of the author + 70 years. It's insane. It's hard to feel pity for them.

I'm still alive.

Copyright is meant to protect against reproducing a work, and something that is transformative should be fair use. If it was free on the internet, they don’t need permission anyway

It was free on the internet, it wasn't free for the taking.

If my bike is out on the street without a lock that does not automatically give you permission to take it.

Copyright isn't so much about protection against reproduction, it is about control over your creation. Literally: "an exclusive legal right to copy, distribute, adapt, display, and perform a creative work".

Note that this is a list of rights that automatically apply to any creative work. You can't train an AI without copying. You can't train an AI without adaptation. AIs have been known to be able to verbatim reproduce parts of the ingested data, they are in effect lossy compressors and automated ones at that. If you take out all of the training data the AI is just a pile of random noise so it is the training data not so much as the training process that adds value to the AI. The training process is trivial compared to the effort that went into the creation of the content used during the training process.

If something was 'free for the taking' and you believe they don't need permission anyway then I guess you're going to have to either come to terms with the balance being that the resulting work product should then also be free for the taking or you will have to explain why you think that that is not the case. The Chinese labs (one in particular) that is being accused here in an evidence free manner at least put their weights out for others to use. So I have less of a problem with that than I do with Anthropic.

Ideas can’t be copyrighted. Paraphrasing some text explaining an idea isn’t a copyright violation. Having trained on something you wrote doesn’t necessarily imply a copyright violation. It depends on what the LLM outputs.

This isn't about ideas. It is about text. Paraphrasing text can definitely be a copyright violation. I've won a couple of copyright related cases and settled one for a substantial sum out of court. In every one of those the other party had some weird idea of what copyright really means.

Okay, paraphrasing text describing an idea isn’t necessarily a copyright violation.

I think the issue is that it’s case-specific, but there are plenty of people claiming copyright violations in circumstances where there’s no specific evidence for it, and a few other people claiming that LLM inference necessarily means it’s transformative use.

Yes, and I fully expect that all of this is going to be settled law in a decade (or two), but right now plenty of this is still up in the air. As new evidence is found about what the residue in the LLMs actually is I expect the law to take note.

Besides all that: no matter how you slice the actual case for infringement, these companies have the authorities in their pocket and even if what they are doing is technically illegal I fully expect them to get away with it because they are very wealthy. Which is one reason they're not seeing any of my money.

Reply on news.ycombinator.com