3 ms·

> I feel as though, from an information-theoretic standpoint, it can't be possible that an LLM (which is almost certainly <1 TB big) can contain any substantial

by gspr 7mo ago

> I feel as though, from an information-theoretic standpoint, it can't be possible that an LLM (which is almost certainly <1 TB big) can contain any substantial verbatim portion of its training corpus, which includes audio, images, and videos. It doesn't need to for my argument to make sense. It's a problem if it reproduces a single copyrighted work (near)-verbatim. Which we have plenty of examples of.

NewsaHackO 7mo ago
Do we? Even when people attempt to jail break most models with 1000s of prompts they are only able to get a paragraph or two of well known copyrighted works and some blocks of paraphrased text, and that's with giving it a substantially leading question.
- gspr 7mo ago
  It surely doesn't matter how leading or contorted the prompt has to be if it shows that the model is encoding the copyrighted work verbatimly or nearly so.
  - NewsaHackO 7mo ago
    It definitely does, which is why I put substantial amount of verbatim material. If someone can recite the first paragraph of Harry Potter and the sorcerers stone from memory, it surely doesn't mean they have memorized the entire book.
    - gspr 7mo ago
      Of course not. But if the passage they can recite is long enough that it is copyrightable, then surely distributing a thing that (contortedly or not) can do said recitation is a form of redistribution of the work itself?
      - NewsaHackO 7mo ago
        No. It is against their TOS to attempt to jailbreak their models. While I don't agree that the models can recite longer periods of verbatim copyrighted material, even if it could, the person who is at fault is the person subverting the system, not the creator of the system. If I steal a library book and make copies of it to distribute illegally, it wouldn't make sense to hold the library at fault for infringing on the book publisher's copyright.
        gspr 7mo ago
        This is an interesting take that I hadn't considered. Your analogy with a library break-in is good. I'll need to digest this. Thanks.

by gspr 7mo ago

NewsaHackO 7mo ago

Do we? Even when people attempt to jail break most models with 1000s of prompts they are only able to get a paragraph or two of well known copyrighted works and some blocks of paraphrased text, and that's with giving it a substantially leading question.

gspr 7mo ago
It surely doesn't matter how leading or contorted the prompt has to be if it shows that the model is encoding the copyrighted work verbatimly or nearly so.
- NewsaHackO 7mo ago
  It definitely does, which is why I put substantial amount of verbatim material. If someone can recite the first paragraph of Harry Potter and the sorcerers stone from memory, it surely doesn't mean they have memorized the entire book.
  - gspr 7mo ago
    Of course not. But if the passage they can recite is long enough that it is copyrightable, then surely distributing a thing that (contortedly or not) can do said recitation is a form of redistribution of the work itself?
    - NewsaHackO 7mo ago
      No. It is against their TOS to attempt to jailbreak their models. While I don't agree that the models can recite longer periods of verbatim copyrighted material, even if it could, the person who is at fault is the person subverting the system, not the creator of the system. If I steal a library book and make copies of it to distribute illegally, it wouldn't make sense to hold the library at fault for infringing on the book publisher's copyright.
      - gspr 7mo ago
        This is an interesting take that I hadn't considered. Your analogy with a library break-in is good. I'll need to digest this. Thanks.