• RedstoneValley
    link
    fedilink
    English
    arrow-up
    40
    arrow-down
    1
    ·
    edit-2
    1 year ago

    “public” does not mean you’re allowed to steal it and republish it as a work of your own. There are things like copyright and stuff

    • cooljacob204@kbin.social
      link
      fedilink
      arrow-up
      15
      arrow-down
      1
      ·
      1 year ago

      “public” does not mean you’re allowed to steal it and republish it as a work of your own

      That is not what they or LLMs do. And while there is questionable morals around it acting like they are straight up stealing and republishing work hurts having serious discussions about it.

      • Kichae@kbin.social
        link
        fedilink
        arrow-up
        19
        ·
        1 year ago

        LLMs create statistical distributions of words and phrases based on ingested data, and then sample those distributions given conditional probabilities.

        Why should for-profit companies have the right to create these statistical distributions based on our written works without consent? They’re not publishing these distributions, and the purpose of ingesting these texts is not to report on the distributions.

        They’re just bottom-trawling the internet and acting as if they have every right to use other peoples’ written works. While people are having “serious discussions” around it, they’re moving forward, ignoring the discussions entirely, and trying to force the conclusion of those discussions to be “well, it’s too late now, anyway”.

        • Even_Adder@lemmy.dbzer0.com
          link
          fedilink
          English
          arrow-up
          6
          ·
          edit-2
          1 year ago

          Original analysis of public data is not stealing. If it were stealing to do so, it would gut fair use, and hand corporations a monopoly of a public technology. They already have their own datasets, and the money to buy licenses for more. Regular consumers, who could have had access to a corporate-independent tool for creativity and social mobility, would instead be left worse off with fewer rights than where they started.

      • pjhenry1216@kbin.social
        link
        fedilink
        arrow-up
        9
        ·
        1 year ago

        You should have the discussions first. Not after you already profited off someone else’s work. If the argument should be about whether they can use the data or not, then harvesting it first is absolutely harmful to the discussion you claim is important. You can’t just argue one side is in bad faith when the other side is already objectively acting in bad faith if we are to assume the discussion is real.

        • cooljacob204@kbin.social
          link
          fedilink
          arrow-up
          8
          ·
          edit-2
          1 year ago

          You should have the discussions first.

          But we are way past that. And legally while they are walking a thin line it seems that LLMs are going to win the legal challenges.

          I don’t think stopping or slowing LLM development is going to work, because then more questionable countries who really don’t give a fuck about IP will pull ahead.

          If you want my honest opinion I don’t think these LLMs companies are stealing and I do think artists are getting the shit end of the stick at the same time. We are heading towards and AI dystopia and I think the way to address is is through more solid social welfare programs instead of fight about IP. While artists are the focus, this AI revolution is coming for all labor. Artists are unfortunately the first ones being impacted by it.

          I think people should stop fighting about the minor things and instead prep for the inevitable unemployment this will bring. LLMs are really just the tip of the iceberg.

      • RedstoneValley
        link
        fedilink
        English
        arrow-up
        3
        ·
        1 year ago

        Yeah, you’re right that it is different from simply stealing content. However the LLMs still use protected material as input and it seems that at least parts of those works can be uniquely identified in the output. That can be considered problematic, even if the data is deconstructed into embeddings inbetween input and output.

    • little_water_bear@discuss.tchncs.de
      link
      fedilink
      English
      arrow-up
      3
      ·
      1 year ago

      Thank you. I haven’t thought about copyright just now. This is indeed something that needs to be addressed.

      Although I personally still don’t have much of a problem with that. I think copyright laws are highly debatable.