LLMs (OpenAi models included) are happy to reproduce books word by word, page by page. Just try it out yourself. And even if some words were reproduced wrong, it still would be copyright violation.
As I said in the comment you’re replying to, there’s case law proving you wrong.
I suspect that most of the large AI companies relevant to this discussion will remain based in the US.
Most of the money is in the US, China, and the EU. China won't allow any LLM that accidentally says mean things about their government, the EU is worried about AI that may harm individuals by libelling them.
The Chinese models may well completely ignore western laws, but if they're on the other side of the Great Firewall, or indeed just have Chinese-language UIs and a focus on Chinese-language tokens in the training… well, I'm not 100% confident, but I would be somewhat surprised if, say, JK Rowling was upset upon discovering that western users attempting to pirate her works via a Chinese chatbot were getting a version of Harry Potter that begins with the title literally being "赫奇帕奇巫师石（哈利·波特与魔法石）" (as ChatGPT just told me the Chinese version starts. Google Translate claims the first three characters are "Hufflepuff").
Even if the rules aren't any harder (as I'm not a lawyer, I can't tell if the differences in copyright rules will or won't make a huge difference in compliance costs), it's likely easier for American companies to lobby the American government for what they want done to make business easier.
AI models will get broad federal immunity is my prediction for 2025.
I'll bet DOGE coins on it.
This statement is a figment of the commenters imagination with no basis in reality. All they would have to do is try it to realize they just spouted a lie.
At most LLMs can produce partial excerpts.
LLMs don’t store the data that it’s trained on. That would be infeasible, the models would be too large. Instead, it stores semantic representations which often uses entirely different words and sentence structures than the source content. And of course most of the data is lost entirely during this lossy compression.
Size difference meaning that people often share complete copies of articles to get around pay walls — including here. As I understand it, this is already copyright infringement.
I suspect that those copies are how and why it's possible in cases such as NYT.
It seems like that would be a fact that couldn't be argued with.
Glad you agree that LLMs infringe copyrights.