Large language artificial intelligence models, such as ChatGPT, often misjudge what people outside the West might value as a moral priority, according to our new research published in the Proceedings of the National Academy of Sciences.

  • scutiger
    link
    fedilink
    English
    arrow-up
    9
    ·
    6 days ago

    This is dumb. LLMs do not reason and don’t prioritize viewpoints, they determine the order of a string of words statistically based on context given to them directly, generally in the form of a query.

    If you anthropomorphize the LLM, you make it seem like it’s making decisions when it’s not.

    • Wren@lemmy.todayOPM
      link
      fedilink
      English
      arrow-up
      2
      arrow-down
      1
      ·
      6 days ago

      It’s not anthropomorphism, it’s explaining a bias in their output based on training data.

      • calcopiritus
        link
        fedilink
        English
        arrow-up
        2
        ·
        6 days ago

        That’s like the definition of training data.

        All output of all kinds is biased by the training data.

        • Wren@lemmy.todayOPM
          link
          fedilink
          English
          arrow-up
          1
          ·
          6 days ago

          I’m not debating what is or isn’t training data, I’m saying it’s not anthropomorphism to say an LLM has a bias.

          • thenextguy@sh.itjust.works
            link
            fedilink
            English
            arrow-up
            1
            ·
            6 days ago

            Large language artificial intelligence models, such as ChatGPT, often misjudge what people outside the West might value as a moral priority

            How is that anything but anthropomorphism?

            • Wren@lemmy.todayOPM
              link
              fedilink
              English
              arrow-up
              1
              ·
              6 days ago

              It helps to read the rest of the article which explains the methodology behind that thesis.

              Besides, how is it anthropomorphism? Is there any other way to convey that point?

                • Wren@lemmy.todayOPM
                  link
                  fedilink
                  English
                  arrow-up
                  1
                  ·
                  6 days ago

                  I don’t agree that those terms obfuscate what AI is when taken in context of the whole article.

                  I’m honestly asking how to get that point across in other terms, do you have an example of what would work better?

                  • thenextguy@sh.itjust.works
                    link
                    fedilink
                    English
                    arrow-up
                    1
                    ·
                    6 days ago

                    “LLMs output may be biased by their input data and users should be careful to account for this when using them.”