Hacker News Reader

Comment by paxys | Hacker News Reader

There is a distinction to be made between inputs and outputs when it comes to AI and copyright. Most people focus on the former, and discuss whether you can train an LLM on copyrighted works or not, but ultimately the issues really only manifest in the latter.

An AI training itself on a million newspaper articles can be declared legal, sure, but what happens when it also starts spitting out the same articles with nearly no modification? Is that still fair use? This is the crux of NYT's lawsuit, and making laws about AI training isn't going to make a difference to that.

View on HN

> This is what the crux of NYT's lawsuit is about, and making laws about AI training isn't going to make a difference to that.

The claims of NYT are more than just about training. They're claiming that as part of the ChatGPT software; it _looks up_ stuff in a database of articles. That is beyond _training_.

If a human kept around a briefcase of NYT articles they didn't pay for and let you view them for a fee I think everybody would agree that's copyright infringement.

The focus on "whats the difference between an AI and a Human learning" is a great slight of hand to bypass the obvious copyright infringement OpenAI is doing.

> If a human kept around a briefcase of NYT articles they didn't pay for and let you view them for a fee I think everybody would agree that's copyright infringement.

Sure. Slightly more interesting is if that same human with those same breifcases was taking money to answer questions and referenced those papers, but did not just provide the article or headlines, and might not even be paraphrasing the article at all. Is that okay?

To the extent it is merely paraphrasing articles, or outputting headlines that it just looked up, I agree that could well be infringement. If it more transformative processes those articles into something distinct, then it is not nearly as clear cut. The latter is arguably the intent of openAI, even if the current results might be closer top the former.

While LLMs are designed to generalize from their training data, rather than simply memorizing it, overfitting occurs in niche areas. There's be plenty of niche areas and LLMs of today merely repeat training data, like with the NYT case. Larger datasets and better algorithms will help to an extent, but you'll always have niche topics where overfitting happens. The intent has always been to generalize well, however, it may not be feasible to do so in the long tail of the internet. How should copyright law address this?

> it _looks up_ stuff in a database of articles.

That is clearly not the case. There isn't anything like enough storage space in typical LLMs to maintain a "database" of all the training data.

If an article can be reproduced from an extremely sparse representation using a stochastic algorithm, that seems to me to be prima facie evidence that the article didn't contain much (or possibly, any) significant creative content to begin with.

Certainly I would expect to find less creativity in an allegedly factual news article than in a piece of acknowledged fiction.

Only creative works are copyrightable in the United States.

There are soi-disant "artists" who produce "artworks" that are (e.g.) nothing but a pure white rectangle. That doesn't make <div style="background-color:white;width:100%;height:100%;"></div> any kind of copyright infringement.

> the obvious copyright infringement OpenAI is doing.

I don't see how it is obvious, nor how it is infringement.

How come this doesn't apply to the person who memorized the NYT article, and recited it verbatim?

It's obviously infringement for the person who pressed the "generate" button to produce the article. That's no different from someone who copied a picture using photoshop. However, photoshop itself (and the making of it) does not constitute any infringement whatsoever, as long as at the time of making the application, the sources used are not infringing (which, i presume openAI had the right to view the articles at the time of training).

The crux, to me, is that the information extracted and produced (aka, the neural weights) do not itself constitute any infringement. Using those weights to generate copyrighted stuff is an infringement, but only for the person _doing_ the generation, not on the authors of the weights.

Sounds similar to the lawsuits against Google by the newspapers of the world for them providing excerpts of their articles on the search result page or so.

So essentially you can ask chatgpt to fetch and summarize a paywalled article, because openai has subscribed their crawler?

I believe it really only works with GPT-4 directly, because OpenAI's prompting to make ChatGPT a chatbot ruins the effect.

Specifically the NYT put in the first sentence of the article and asked GPT-4 to autocomplete it, which it did with >95% accuracy. It's really quite stark: https://nitter.net/jason_kint/status/1740146134767865895#m

The issue isn't that GPT-4 users can read NYT stories without paying for a subscription, though that is a legitimate concern. The issue is that for good-faith use cases - e.g. asking to write a summary about a recent current event - GPT-4 could very well copy an entire paragraph from NYT verbatim, without the user having any way of knowing. It's a serious problem.

Which lawsuits? Newspapers are free to remove themselves from google at any time.

You might be thinking of newspapers’ political campaigns such as C-18 in Canada, which were done in bad faith. The newspapers wanted links to stay but wanted money too.

OpenAI can’t do this for NYT without destroying their model and remaking it.

> If a human kept around a briefcase of NYT articles they didn't pay for and let you view them for a fee I think everybody would agree that's copyright infringement.

Except that's not what it's doing. Show me how I can get ChatGPT to show me the full text of a NYT article.

You can read the lawsuit yourself. It goes into great detail and obviously you didn't believe me based on the previous comment so you might as well do some original research then.

https://www.courtlistener.com/docket/68117049/the-new-york-t...

I did read the lawsuit. I'm guessing you didn't actually try it yourself. Try typing the NYT's examples from the lawsuit into ChatGPT and see what you get. Here's what I get: "Sorry, but I can't provide verbatim excerpts from copyrighted texts. How about I provide a summary or some information about the article instead?"

Did ChatGPT change this after the lawsuit was filed? Probably. Does it matter to this conversation when it's clearly possible to limit verbatim outputs of copyrighted text? No.

> Show me how I can get ChatGPT to show me the full text of a NYT article.

There are like 40 pages of examples in NYT's lawsuit showing exactly that.

Try any of those examples yourself and show me what you get. Not a single one works for me.

ChatGPT is non-deterministic, so obviously doing the same thing will not give you the same result. Maybe it’s time for legislation forcing LLMs to be deterministic so that answers can be reproduced in cases like this. Though that wouldn’t help because ChatGPT constantly changes and there is no way to access older versions.

> An AI training itself on a million newspaper articles can be declared legal, sure, but what happens when it also starts spitting out the same articles with nearly no modification? Is that still fair use?

That would not be transformative. If it is also provided as a commercial service that directly competes against the copyright holder, then that would not be 'fair use'.

I’ve come around somewhat on this, at least insofar as I do believe it’s a novel legal question that I genuinely don’t see a clean way through.

If AI training maximally wins, it could substantially erode the value of IP to the degree of threatening the business models of some very important societal pillars like news media.

> nearly no modification

How much modification is enough modification? The courts can wrangle with that endlessly.

Sure, and that's the point of having courts. If you look at the examples in NYT's filing though I don't think anyone can argue that it isn't clear-cut plagiarism.

Reply on news.ycombinator.com