Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

NY Times copyright suit wants OpenAI to delete all GPT instances

arstechnica.com

Post reply on HN
Print view View original

121–130 of 921 posts

Re: NY Times copyright suit wants OpenAI to delete all GPT instances

#121

Post by aurareturn »

fsckboy wrote: post #110

in my head I like to think of web crawler search engines/search engine databases and LLMs as being somewhat similar. Search engines are ok if they just provide snippets with citations (urls), and they would be unacceptable if they provided large block quotes that removed the need to go to the original source to read the original expression of more complex ideas. A web-crawled LLM that lived within the same constraint…

I think it's different. LLMs can solve problems. Part of that problem-solving ability comes from training completely unrelated content such as NYT articles. GPT4 doesn't have to spit out NYT articles verbatim to have benefited from NYT articles. It uses NYT articles for every query.

Re: NY Times copyright suit wants OpenAI to delete all GPT instances

#122

Post by groceryheist »

visarga wrote: post #102
groceryheist wrote: post #90

Earlier quoted context omitted.

Maybe the bloom filter solution is enough, but I wonder. - Paraphrasing n=7 words (and quite a few more) within a sentence can easily be fair use. - As n gets big, the bloom filter has to also. If/when attribution is solved for LLMs (and not fake attribution like from Bing or Perplexity) then creators can be compensated when their works are used in AI outputs. If compensation is high enough this can greatly incentivi…

As n-gram length grows, we are still going to have the same number of ngrams, they go through a hashing function and indexed in the bloom filter as usual. The number of n-grams size n in a text is text_length - ngram_length + 1.

The number of unique values in the bloom filter will go up ~exponentially with n. So to control the false positive rate the bloom filter has to grow.

Re: NY Times copyright suit wants OpenAI to delete all GPT instances

#123

Post by Dalewyn »

__loam wrote: post #84
tarruda wrote: post #72

Would be funny if NT Times won this and all commercial LLMs were shut down. Then LLMs would be distributed only via torrents, like most copyright infringing media.

Making these things anathema to commercial interests and making training them at scale legally perilous would be a huge win.

>making training them at scale legally perilous

Loading data to which you have no rights over into your software is legally perilous, yes.

It's as easy as simply asking for and receiving permission from the data's rightsholders (which might require exchange of coin) to make it not legally perilous.

Re: NY Times copyright suit wants OpenAI to delete all GPT instances

#124

Post by malwrar »

MisterBastahrd wrote: post #73
tucnak wrote: post #34

Earlier quoted context omitted.

>Why someone work with full time writing articles should give the work for free They are not giving it out "for free", in fact they're being paid by their employer to write these articles. Moreover, the writers themselves stand noth' to gain from their past writings financially as they don't belong to the ownership structure of the business.

Their ability to make money in the future is directly tied to their employers' ability to make money with their content. This is a closed financial loop. If OpenAI or any other AI company wants in, they should pay a licensing fee or get the laws changed, not just assume that they can take what they want and pretend like there are no negative consequences for the creator or the rights-holder.

In this limited example, are there such consequences? Are people dropping NYT subscriptions because they trust chatgpt to inform them of current events? I don’t buy it.

Re: NY Times copyright suit wants OpenAI to delete all GPT instances

#125

Post by type_Ben_struct »

logicchains wrote: post #115

NYT's perspective is going to look so stupid in future when we put LLMs into mechanical bodies with the ability to interact with the physical world, and to learn/update their weights live. It would make it completely illegal for such a robot to read/watch/listen to any copyrighted material; no watching TV, no reading library books, no browsing the internet, because in doing so it could memorise some copyrighted conte…

I disagree. The verbatim part is the problem. You’re drawing a comparison to how humans operate except we’re not allowed to operate like that.

While harder to do as a human, if memorised a copyrighted book and then did a live reading on TV, or produced replicas from memory and sold them (the most comparable example), I’d be sued.

Humans produce derivative work all the time, and it’s fine for LLM’s to do that, but you can’t do it verbatim.

Re: NY Times copyright suit wants OpenAI to delete all GPT instances

#126

Post by jrockway »

I read about this in the Times today (and am surprised that it wasn't on HN already).

My guess is that the court will likely find in the Times favor, because the legal system won't be able to understand how training works and because people are "scared" of AI. To me, reading a book, putting it in some storage system, and then recalling it to form future thoughts is fair use. It's what we all do all the time, and I think that's exactly what training is. I might say something like "I, for one, welcome our new LLM overlords". Am I infringing the copyright of The Simpsons? No.

I am guessing some technicality like a terms-of-use violation of the website (avoidable if you go to the library and type in back issues of the Times), or storing the text between training sessions is what will do OpenAI in here. The legal system has never been particularly comfortable with how computers work; for example, the only reason EULAs work is because you "copy" software when your OS reads the program off of disk into memory (and from memory into cache, and from cache into registers). That would be copyright infringement according to courts, so you have to agree to a license to get that permission.

I think the precedent on copyright law is way off base, granting too much power to authors and too little to user. But because it's so favorable towards "rightsholders", I expect the Times to prevail here.

Re: NY Times copyright suit wants OpenAI to delete all GPT instances

#127

Post by heavyset_go »

rich_sasha wrote: post #111

If you forget about the LLM aspect, and simply build a product out of (legally) scraped NYT articles, is that fair use? Let's say I host these, offer some indexing on it, and rewrite articles. Something like, summarise all articles on US-UK relationships over past 5 years. I charge money for it, and all I pay NYT is a monthly subscription fee. To keep things simple, let's say I never regurgitate chunks of verbatim NY…

Another factor to consider is that neural nets can function as lossy compression, which becomes extremely evident when using models that are overfit.

Sometimes they're so overfit that the compression isn't even lossy, and the data is encoded verbatim in the NN.

Re: NY Times copyright suit wants OpenAI to delete all GPT instances

#128

Post by geysersam »

groceryheist wrote: post #90
visarga wrote: post #82

Earlier quoted context omitted.

> Just learn to recognize and punish plagiarism via RLHF. This is not a RLHF problem. What I was expecting them to do is to keep a bloom filter of ngrams for known copyrighted content, such as enumerating all sets of n=7 consecutive words in an article, and validate against it. The model would only output at maximum n-1 words that look verbatim from the source. But this will blow up in their face. Let's see: - AI com…

Maybe the bloom filter solution is enough, but I wonder. - Paraphrasing n=7 words (and quite a few more) within a sentence can easily be fair use. - As n gets big, the bloom filter has to also. If/when attribution is solved for LLMs (and not fake attribution like from Bing or Perplexity) then creators can be compensated when their works are used in AI outputs. If compensation is high enough this can greatly incentivi…

> if compensation is high enough

Who pays the compensation? If it's the user, why wouldn't they just buy the authors work directly? Why go through the LLM middleman?

Re: NY Times copyright suit wants OpenAI to delete all GPT instances

#130

Post by rich_sasha »

logicchains wrote: post #112
rich_sasha wrote: post #111

If you forget about the LLM aspect, and simply build a product out of (legally) scraped NYT articles, is that fair use? Let's say I host these, offer some indexing on it, and rewrite articles. Something like, summarise all articles on US-UK relationships over past 5 years. I charge money for it, and all I pay NYT is a monthly subscription fee. To keep things simple, let's say I never regurgitate chunks of verbatim NY…

> Something like, summarise all articles on US-UK relationships over past 5 years. I charge money for it, and all I pay NYT is a monthly subscription fee. >Is that fair use? IANAL, but doesn't sound like it. If you pay someone to do the summarisation for you, then you publish the content and charge a fee for it, you're the one liable, not the person you paid to summarise it for you. Similarly if you ask GPT to do it…

That's not the example. Here I proactively scrape NYT, summarise articles for a fee and sell that as a service. It's not people coming to me with some articles to summarise, and maybe then publishing it online.

At some level it becomes a subversion of NYTs fees. First, say I subscribe and simply host the articles verbatim, for a fee. Clearly, that's not right.

Suppose I change some spelling or word order, or use a synonym or two. That's still not ok.

And if I substantially paraphrase the articles? I guess this is the relevant case. This is kind of what LLMs do. And also feels like not fair use.

Post reply on HN