Hacker News Reader

Comment by jeffgreco | Hacker News Reader

I mean I’m no lawyer but this doesn’t strike me as a great example for infringement? Detailed summaries of books sounds like textbook transformative use. Especially in Silverman’s case, reducing her book to “facts” while eliminating artistic elements of her prose make it that much less of a direct substitute for the original work.

View on HN

I can see a good argument in the complaint. The provenance of the training data leads back to it being acquired illegally. Illegally acquired materials were then used in a commercial venture. That the venture was an AI model is perhaps beside the point. You can’t use illegally acquired materials when doing business.

It seems like a weak argument, in that it is just as likely it saw any number of things about it, from book reviews to sales listings to interviews.

Unless OpenAI can prove that the outputs are derived from legally vs illegally-obtained outputs, not sure that’s going to matter. And as far as I understand about their models, that’s effectively impossible.

Isn’t the burden of proof on the other side?

Not when OpenAI publicly declared they trained on pirated works. I can’t imagine “we can’t tell if this is the result of the illegal thing we did or not” is going to stand up very well, nor does it bode well for any refutation of the plaintiff’s depiction of their intent. Part of fair use consideration is commercial impact and when you steal a bunch of books to train your AI model, it’s hard to refute that the impact is not negative or that you didn’t intend commercial harm.

Please read more carefully. OpenAI never “declared they trained on pirated works.”

However, the root of this thread quotes Meta admitting exactly that.

… it’s hard to refute that the impact is not negative or that you didn’t intend commercial harm.

So this is another thing I don’t understand. Is the claim that fewer people will buy Silverman’s book because ChatGPT is able to provide a summary? If so, call me skeptical.

That's not a claim Silverman has to make. "You copied my book without license to do so" is sufficient.

Did they?

Do you think scraping huge swathes of the internet contains pirated works or not?

Does that mean "no"? I guess so.

Where does it say they scraped huge swathes of the internet and didn't look at the results?

> it is just as likely it saw any number of things about it

Is this based on inside information, or just the law of averages? Doesn't the fact that they openly admitted to having been trained on pirated books affect your priors?

They didn't, more conjecture

It seems to me that they are indeed admitting to using the pile/books3 dataset, which seems to contain Silverman’s book, at least.

See post https://news.ycombinator.com/item?id=36659041 for a summary.

>You can’t use illegally acquired materials when doing business.

This vague sentence conjures images of a company building products from stolen parts, but this situation seems different. IANAL, but if I looked at a stolen painting that nobody had ever seen, and sold handwritten descriptions of the painting to whoever wanted to buy one, I'm pretty sure what I've sold is not illegal.

Piracy of content is against the law. All other analogies such as looking at paintings are not at issue here. The content was pirated and there are laws against that, whether we agree with it or not.

So, if the plaintiff can prove the content was pirated, then the use of that content downstream is tainted.

>then the use of that content downstream is tainted

What does that mean exactly? That's why I used the "looking at a stolen painting" example.

Sure, pirating materials is illegal. But I don't think that's the big implication that people are getting at here. Is it legal to sell original works derived from perceiving stolen materials? Seems to me that it is.

In this case the correct analogy would be you brought a stolen painting into your house, looked at it for a while, and then produced your derivative work.

Surely you see the issue here? Receiving stolen property?

Yes, I acknowledged that piracy is illegal in my previous post. That's not what the lawsuit is about, according to The Verge:

>In the OpenAI suit, the trio offers exhibits showing that when prompted, ChatGPT will summarize their books, infringing on their copyrights.

That serves as evidence that the model has seen the material, and the only way the model could have seen the material is if it was pirated.

That assumption is the reason why it's a shady argument for infringement. There's other ways to get a summary of a book than reading it, like asking a friend who's read it to give a summary

You're missing the bigger picture[1]:

  Numerous questions of law or fact common to each Class arise from Defendants’ conduct:
    whether ChatGPT itself is an infringing derivative work based on Plaintiffs’ copyrighted books;
    whether the text outputs of ChatGPT are infringing derivative works based on Plaintiffs’ copyrighted books;

It's not a simple case of "you used our copyrighted materials", it's "you're infringing on our copyright by producing works derived from materials that you used."

1. https://llmlitigation.com/pdf/03223/tremblay-openai-complain...

> So, if the plaintiff can prove the content was pirated, then the use of that content downstream is tainted.

Has that been tested in court?

This is quite an interesting case.

Obtaining the book in the first place[0] appears to be quite a clear case of copyright infrigement.

The question of whether a work derived from the book is infringment is pretty complex, and there's a wide range of tests that get applied to determine that.

But is it necessarily true that if you obtained the original work via copyright infringement and then created an otherwise non-infringing derived work, your derived work is nevertheless infringing due to the provenance of your copy of the original work?

[0] https://news.ycombinator.com/item?id=36659041

The more I think about it, I think it will (and should) turn on the extent to which "the law" considers the AI's to be more like "people" or more like "machines?" People can read and do research and then spit out something different.

But "feeding the data into a machine" seems like obvious infringement, even if the thing that comes out on the other end isn't exactly the same?

Perhaps not, I thought one of the claims is interesting though, that they illegally acquired some of the dataset. What would be the damages from that, the retail price of the hardcopy?

Wouldn't they first need to prove that OpenAI didn't ingest summaries of the book, and not the book itself?

I think this can to some extent be determined in the discovery phase of the lawsuit. We probably could have some interesting outputs from this process.

The remedies under Title 17 are an injunction against further distribution, disgorgement or statutory damages, and potentially attorneys fees. The injunction part is why these cases usually settle if the defendant is actually in the wrong.

Could a suitably prompted LLM repeat, verbatim, the book in its entirety?

Perhaps? But certainly not what’s shown here.

Haven't read the complaint, but there might be an argument that OpenAI used stolen works to train their data, and as such fair use doesn't apply.

Reply on news.ycombinator.com