Paraphrasing is also known as cloning and is often a copyright violation
In US copyright law facts cannot be copyrighted, so copyright on factual content like newspaper articles is limited. Simply replacing a few words wouldn't work, but I am certain that GPT-4 is capable of paraphrasing factual content at a level that would not be considered infringement if a human did it.
Genuinely - what are you talking about besides your own assumptions? you just assume everything google does is legal and therefore any one else doing anything arguably similar must also be legal? Without regard for factual details that do matter to copyright law? Such as license?? Your own description of copyright law here is very stunted - you can't paraphrase articles of the NYTimes and call it a fair use. You can report on what the NYtimes reports on... because that's what news is.
Not an assumption. This is well established. They've been doing it for twenty years!
> Without regard for factual details that do matter to copyright law? Such as license??
What license? Google doesn't in general have or need an explicit license to crawl websites and neither does OpenAI.
It's not at all well-established. How many anti-trust suits is Google facing now? Your proposition defies common sense.
>What license? Google doesn't in general have or need an explicit license to crawl websites and neither does OpenAI.
It's not the crawling the website that OpenAI did that it needs a license for... why bother conversing if you are going to be this obtuse?
Seems like the legal answer is unclear but, like Napster, such a system seems like it would lose in court.
One site clones fox news. One clones news max. And so on, cloning many news sites, sports sites, any news site. Automated, massive scale content farming. Think of the websites recommended by Taboola but, realistically, a whole lot worse.
Nobody is seriously going to ChatGPT and trying to trick it into regurgitating old NYT articles as an alternative to paying for access to NYT's archives. Meanwhile, newspapers went as far as getting the laws changed in several countries because they felt Google was competing with them too much and didn't like the fact that it was legal.
Can they? Here's reference to a legal fight where Google scraped song lyrics from a lyrics website, and presented the lyrics verbatim directly to users (bypassing the original site and the ads that allowed that site to operate)
https://www.rollingstone.com/music/music-features/genius-law...
Google has a wide range of products and shakedowns. Not all of them are "perfectly" legal: Google is being challenged in court over some of their shakedowns and products practices.
The whole point of a search engine (as we've classically known them) is to index the web and respond to queries with a list of links that you will inspect and click through on. The whole point of an LLM chatbot tool is to eliminate those inspecting and clicking-through steps, becoming a one-stop shop for content whose substance was created by someone else. That's also the whole point of GP's hypothetical, which is why it works as an analogy.
---
There are substantially better arguments for search engines being legitimate fair use. Consider, for example, transformation. AI defenders will argue that these systems are transformative because they reshuffle elements of their input in their output, but that's clearly a much weaker form of transformation than one in which the transformed work has an entirely different nature and purpose, i.e. search engines vs. the results they return. Ultimately these technicality-based "nuh uh" arguments aren't going to save the practice of training AI on unlicensed data, because they are incompatible with the spirit of copyright law even if the novel nature of these technologies means the letter of said law can't quite nail them down yet.
If these arguments do succeed, it will be because the judicial/regulatory environment in which they were applied has been corrupted by capital.
An LLM takes an input string a corpus of text, and returns a series of text that best comes after the next input string.
To get a paragraph of output, you run the search over and over again
Both the search and LLM reshuffle the inputs to the outputs.
If I'm describing the purpose of the LLM, it's got a wide number of usages. "Making my resume look more professional" or "be a crud api" or "reformat my ask into a api call to X service" or "give me a timeline of events surrounding Y with source links"
If I'm describing the purpose of an 18 wheeler, it's got a wide number of usages. "Carry my chicken" or "carry my lettuce" or "carry my Cheetos". Or, simply, "carry my groceries".
And how did the training data contribute to the content in any meaningful way? Inspiration isn't substance.
You think all fantasy writers gotta pay Tolkien estate bc so much of fantasy draws from his tropes? Lmao no.
If training data is so unimportant, why not simply not use it and avoid the controversy? At the very least that would certainly fix the issue where the model demonstrates how "inspired" it is by NYT articles by reproducing them verbatim.
:)
Possession is not a crime when it comes to copyright. It's not like physical things (e.g. drugs or guns) at all. This is why comparing copyright violations to theft is silly.
ChatGPT can absolutely keep verbatim copies of the entire works of basically anything and not run afoul of the law. When it regurgitates a small part of an article that's covered by fair use in theory but the truth is that fair use can only be determined by a judge in a court of law when someone is sued. It cannot be determined with any sort of certainty ahead of time. It's a legal defense, nothing more.
Summarizing content has been legal forever as well (see the other posts here talking about Cliff Notes and some similar products). That's not even fair use that's just like, people's opinions, man (legally speaking).
I don't think the NYT will get what they want out of this at all.
Man thinks piracy is legal
You just described Google. When you think about it, it's surprising that Google is legal. However, it is well established that what Google does is perfectly legal. Remember that internally Google keeps and uses complete verbatim copies of every web page they index.
Yes, Google offers a link to the source. If OpenAI did the same, even if only 0.1% of people clicked on the links and NYTimes hardly got any revenue from it, would that make it legal in your eyes? What if they implemented a system that detected when it was outputting a verbatim copy of something and simply paraphrased it? NYTimes clearly doesn't have copyright on paraphrased versions of their articles. I think it would be pretty silly if the government forced them to do that as it wouldn't make any practical difference to anyone.