5 ms·

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 lear

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 content.

ramraj07 3y ago
Will it? If the LLM in the body is allowed to read nytimes on a tablet I'm sure they wouldn't care.
- logicchains 3y ago
  >If the LLM in the body is allowed to read nytimes on a tablet I'm sure they wouldn't care. Why should the law treat a LLM in a body reading NYT on a tablet differently than a LLM browsing the content from a website online and reading that?
type_Ben_struct 3y ago
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.
- logicchains 3y ago
  >or produced replicas from memory and sold them (the most comparable example), I’d be sued. This is not the most comparable example, because it's not what ChatGPT is doing. The most comparable example is if you were hired as a contractor and the employer asked you to write verbatim some copyright content you'd memorised. If the employer then published it, they'd be the one liable, not you. >Humans produce derivative work all the time, and it’s fine for LLM’s to do that, but you can’t do it verbatim. Nobody's suggesting preventing humans from consuming any copyrighted content just because in future they might recite some of it verbatim, but that's what NYT want for LLMs.
  - tsimionescu 3y ago
    > The most comparable example is if you were hired as a contractor and the employer asked you to write verbatim some copyright content you'd memorised. If the employer then published it, they'd be the one liable, not you. No, you'd both be liable. You are not allowed to create copies of a copyrighted work, even from memory, for any commercial purpose. Making it public or not is irrelevant. This is more obvious with spftware: if I copy a version of AutoCAD that my previous employer bought and sell it to another company, or even just use it for my current employer without showing it to anyone else, I am violating the copyright on that software, and I am liable. Even though obviously no "publishing" happened. Similarly, if you hire a decorator to paint Mickey Mouse on the inside walls of your private kindergarten, the decorator is violating Disney's copyright just as much as you are, even if neither of you has made that public.
    - YuccaGloriosa 3y ago
      Your previous employer never bought AutoCAD, they licenced its use, paying a subscription. When you start working for them that licence was no longer available to you. So you would be unable to subsequently use it.
      - tsimionescu 3y ago
        Unable legally, but I may find illegal ways. And the reason it is illegal to copy is copyright at the end. The license is only (legally) required because of copyright.
- kromem 3y ago
  Then we should be focused on policing the usage of the model, not the training of it. That's the point at which infringement occurs in your example. It's not the memorizing that's the infringement, it's the reproduction from your memory. We shouldn't be regulating your hippocampus encoding the book, but your reproducing the book from that encoding. Similarly, we shouldn't be regulating the encoding of material into the NN, but the NN spitting back out the material.
CJefferson 3y ago
Are those LLMs independant citizens we are going to give rights to? Then I'm fine with that. Are they all owned by one mega-corporation, which is going to do as capitalism does, and use them to squeeze money out of all of us? Then I'm happy to ban them.
- kromem 3y ago
  "Let's ban something capable of diagnosing medical conditions and letting coma patients to communicate with an EEG because it learned the relationships between words from a giant data set of scraped data and is owned by a company" is a pretty callous take IMO. The opportunity cost of holding this technology back is going to literally be millions of people's lives given current trends in its emerging applications. Police usage, not training.
makeitdouble 3y ago
Memorising isn't the issue. It's providing it back verbatim and/or cutting access to the source. You'd get the same problem with someone with a photographic memory who a group of people would turn to recite them the news instead of buying the newspaper. As of now public performance of copyrighted material is infringement.
- kromem 3y ago
  That's not the case, as they aren't trying to get a ruling on the forced reproduction by prompt as infringement, but rather to get a ruling that training is infringement. I fully agree with the perspective that infringement in usage needs to be limited even if I strongly disagree that training is infringement.

ramraj07 3y ago

Will it? If the LLM in the body is allowed to read nytimes on a tablet I'm sure they wouldn't care.

logicchains 3y ago
>If the LLM in the body is allowed to read nytimes on a tablet I'm sure they wouldn't care. Why should the law treat a LLM in a body reading NYT on a tablet differently than a LLM browsing the content from a website online and reading that?

type_Ben_struct 3y ago

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.

logicchains 3y ago
>or produced replicas from memory and sold them (the most comparable example), I’d be sued. This is not the most comparable example, because it's not what ChatGPT is doing. The most comparable example is if you were hired as a contractor and the employer asked you to write verbatim some copyright content you'd memorised. If the employer then published it, they'd be the one liable, not you. >Humans produce derivative work all the time, and it’s fine for LLM’s to do that, but you can’t do it verbatim. Nobody's suggesting preventing humans from consuming any copyrighted content just because in future they might recite some of it verbatim, but that's what NYT want for LLMs.
- tsimionescu 3y ago
  > The most comparable example is if you were hired as a contractor and the employer asked you to write verbatim some copyright content you'd memorised. If the employer then published it, they'd be the one liable, not you. No, you'd both be liable. You are not allowed to create copies of a copyrighted work, even from memory, for any commercial purpose. Making it public or not is irrelevant. This is more obvious with spftware: if I copy a version of AutoCAD that my previous employer bought and sell it to another company, or even just use it for my current employer without showing it to anyone else, I am violating the copyright on that software, and I am liable. Even though obviously no "publishing" happened. Similarly, if you hire a decorator to paint Mickey Mouse on the inside walls of your private kindergarten, the decorator is violating Disney's copyright just as much as you are, even if neither of you has made that public.
  - YuccaGloriosa 3y ago
    Your previous employer never bought AutoCAD, they licenced its use, paying a subscription. When you start working for them that licence was no longer available to you. So you would be unable to subsequently use it.
    - tsimionescu 3y ago
      Unable legally, but I may find illegal ways. And the reason it is illegal to copy is copyright at the end. The license is only (legally) required because of copyright.
kromem 3y ago
Then we should be focused on policing the usage of the model, not the training of it. That's the point at which infringement occurs in your example. It's not the memorizing that's the infringement, it's the reproduction from your memory. We shouldn't be regulating your hippocampus encoding the book, but your reproducing the book from that encoding. Similarly, we shouldn't be regulating the encoding of material into the NN, but the NN spitting back out the material.

CJefferson 3y ago

Are those LLMs independant citizens we are going to give rights to? Then I'm fine with that. Are they all owned by one mega-corporation, which is going to do as capitalism does, and use them to squeeze money out of all of us? Then I'm happy to ban them.

kromem 3y ago
"Let's ban something capable of diagnosing medical conditions and letting coma patients to communicate with an EEG because it learned the relationships between words from a giant data set of scraped data and is owned by a company" is a pretty callous take IMO. The opportunity cost of holding this technology back is going to literally be millions of people's lives given current trends in its emerging applications. Police usage, not training.

makeitdouble 3y ago

Memorising isn't the issue. It's providing it back verbatim and/or cutting access to the source. You'd get the same problem with someone with a photographic memory who a group of people would turn to recite them the news instead of buying the newspaper. As of now public performance of copyrighted material is infringement.

kromem 3y ago
That's not the case, as they aren't trying to get a ruling on the forced reproduction by prompt as infringement, but rather to get a ruling that training is infringement. I fully agree with the perspective that infringement in usage needs to be limited even if I strongly disagree that training is infringement.