4 ms·

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, i

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.

throw5959 2y ago
It's not really some words, it's more like you won't be able to get more than a page out of it and even that is going to be so wrong it's basically a parody and thus allowed.
- angoragoats 2y ago
  I’d love to see you try to defend this notion in court. Parody requires deliberate intent to be humorous. And courts have repeatedly held that changing the words of a copyrighted work while keeping the same general meaning can still be copyright infringement.
  - throw5959 2y ago
    It's not just "changing some words". The majority of words will be different, sentences will be different. The general meaning might be generally the same, but I don't think that's enough to claim copyright protection.
    - angoragoats 2y ago
      I didn’t use the word “some.” Please don’t misquote me. As I said in the comment you’re replying to, there’s case law proving you wrong.
      - throw5959 2y ago
        Good thing that's not global, right? I'm not in the US, our courts work differently.
        angoragoats 2y ago
        I’m sorry for assuming that you were (though you could have mentioned this in your reply). Most of the large AI companies relevant to our discussion are based in the US.
        throw5959 2y ago
        Will that remain true if they have copyright issues in the US?
        ben_w 2y ago
        (I'm not the other commenter) 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.
  - CaffeineLD50 2y ago
    You could argue intent. The model has no intent to infringe. No mens rea. AI models will get broad federal immunity is my prediction for 2025. I'll bet DOGE coins on it.
    - angoragoats 2y ago
      AI models are not entities that can be sued, accused of a crime, or given legal immunity, to my knowledge. But I get the gist of what you’re saying.
realusername 2y ago
I tried and I could not make it work. And even if you could, that has to be the most inefficient way to pirate books on earth.
menzoic 2y ago
> reproduce books word by word, page by page 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.
- rhubarbtree 2y ago
  The NYT has extracted long articles from ChatGPT and submitted the evidence in court.
  - ben_w 2y ago
    Crucial size difference between an article and a book. 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.
  - Lerc 2y ago
    Given that it has been submitted in court, does that mean you can say what the longest verbatim extract was? It seems like that would be a fact that couldn't be argued with.
- angoragoats 2y ago
  > At most LLMs can produce partial excerpts. Glad you agree that LLMs infringe copyrights.

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.

throw5959 2y ago

It's not really some words, it's more like you won't be able to get more than a page out of it and even that is going to be so wrong it's basically a parody and thus allowed.

angoragoats 2y ago
I’d love to see you try to defend this notion in court. Parody requires deliberate intent to be humorous. And courts have repeatedly held that changing the words of a copyrighted work while keeping the same general meaning can still be copyright infringement.
- throw5959 2y ago
  It's not just "changing some words". The majority of words will be different, sentences will be different. The general meaning might be generally the same, but I don't think that's enough to claim copyright protection.
  - angoragoats 2y ago
    I didn’t use the word “some.” Please don’t misquote me. As I said in the comment you’re replying to, there’s case law proving you wrong.
    - throw5959 2y ago
      Good thing that's not global, right? I'm not in the US, our courts work differently.
      - angoragoats 2y ago
        I’m sorry for assuming that you were (though you could have mentioned this in your reply). Most of the large AI companies relevant to our discussion are based in the US.
        throw5959 2y ago
        Will that remain true if they have copyright issues in the US?
        ben_w 2y ago
        (I'm not the other commenter) 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.
- CaffeineLD50 2y ago
  You could argue intent. The model has no intent to infringe. No mens rea. AI models will get broad federal immunity is my prediction for 2025. I'll bet DOGE coins on it.
  - angoragoats 2y ago
    AI models are not entities that can be sued, accused of a crime, or given legal immunity, to my knowledge. But I get the gist of what you’re saying.

realusername 2y ago

I tried and I could not make it work. And even if you could, that has to be the most inefficient way to pirate books on earth.

menzoic 2y ago

> reproduce books word by word, page by page 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.

rhubarbtree 2y ago
The NYT has extracted long articles from ChatGPT and submitted the evidence in court.
- ben_w 2y ago
  Crucial size difference between an article and a book. 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.
- Lerc 2y ago
  Given that it has been submitted in court, does that mean you can say what the longest verbatim extract was? It seems like that would be a fact that couldn't be argued with.
angoragoats 2y ago
> At most LLMs can produce partial excerpts. Glad you agree that LLMs infringe copyrights.