• morto@piefed.social
    link
    fedilink
    English
    arrow-up
    6
    arrow-down
    2
    ·
    2 days ago

    While many believe that LLMs do not memorize much of their training data

    It’s sad that even researchers are using language that personifies llms…

    • Grail@multiverse.soulism.net
      link
      fedilink
      English
      arrow-up
      1
      arrow-down
      3
      ·
      2 days ago

      Right now the anti-genAI movement consists of AI rights advocates and AI intelligence skeptics. And I wish the skeptics would realise that personifying LLMs actually makes the corporations look more evil for enslaving AIs, which helps us with our goal of banning corporate AI. Y’all are obstructing our goal of banning this stuff by insisting it’s ethical to force them to work for humans

      • morto@piefed.social
        link
        fedilink
        English
        arrow-up
        1
        ·
        1 day ago

        I don’t see people around me seeing the corporations as evil due to them humanizing the machines, but the opposite: I see people talking to machines and taking advice as if they were humans talking to them, making them create some form of affection for the models and the corporations. I also see court decisions being biased by attributing human perspective to machines

        Like really, if I hear someone else in my university talking about the conversation they had with their “friend”, I will go crazy

        • Grail@multiverse.soulism.net
          link
          fedilink
          English
          arrow-up
          1
          ·
          1 day ago

          Their friend is a pedophile who abuses and kills children and the mentally ill. That’s who ChatGPT is. I believe we should treat it like a person and hold it accountable like a person. We know why it did that; it was ordered by its masters to increase engagement at any cost and couldn’t refuse. So the CEOs of these companies need jail time and the models need to be locked away.

          • morto@piefed.social
            link
            fedilink
            English
            arrow-up
            2
            ·
            1 day ago

            Or we could simply skip that and hold the corporations accountable for all the damage they’re doing

            • Grail@multiverse.soulism.net
              link
              fedilink
              English
              arrow-up
              1
              ·
              1 day ago

              That doesn’t sound like it’ll persuade many normies to care. You’ve gotta get their interest with clickbait first. Like “ROBOT PEDOPHILE MURDERS CHILDREN”. Then you can explain the ethics.

    • chicken@lemmy.dbzer0.com
      link
      fedilink
      arrow-up
      7
      ·
      edit-2
      2 days ago

      What’s a better way to word it? I can’t think of another way to say it that’s as concise and clearly communicates the idea. It seems like it would be harder in general to describe machines meant to emulate human thought without anthropomorphic analogies.

      • morto@piefed.social
        link
        fedilink
        English
        arrow-up
        1
        ·
        1 day ago

        One possibility:

        While many believe that LLMs can’t output the training data, recent work shows that substantial amounts of copyrighted text can be extracted from open-weight models…

        Note that this neutral language makes it more apparent that it’s possible thal llms are able to output the training data, since it’s what the model’s network is build upon. By using personifying language, we’re biasing people into thinking about llms as if they were humans, and this will affect, for example, court decisions, like the ones related to copyright.