• ji59@kbin.social
    link
    fedilink
    arrow-up
    0
    ·
    9 months ago

    The problem is that these answers are hugely incorrect and if some child learning about history of England would see this, they would create bias that England was always diverse.
    The same is true for some recent post, where people knowing nothing about Scotland history could learn from images that half of Scotland population in 18th century was black.
    So from my perspective these images are just completely wrong and it should be fixed.
    Also if you want diversity, what about handicapped people?

    • groet@feddit.de
      link
      fedilink
      arrow-up
      0
      ·
      9 months ago

      Repeat after me:

      “Current AI is not a knowledge tool. It MUST NOT be used to get information about any topic!”

      If your child is learning Scottish history from AI, you failed as a teacher/parent. This isn’t even about bias, just about what an AI model is. It’s not even supposed to be correct, that’s not what it is for. It is for appearing as correct as the things it has been trained on. And as long as there are two opinions in the training data, the AI will gladly make up a third.

      • GregorGizeh@lemmy.zip
        link
        fedilink
        arrow-up
        0
        ·
        9 months ago

        That doesn’t matter though. People will definitely use it to acquire knowledge, they are already doing it now. Which is why it’s so dangerous to let these “moderate” inaccuracies fly.

        You even perfectly summed up why that is: LLMs are made to give a possibly correct answer in the most convincing way.

      • Skull giver@popplesburger.hilciferous.nl
        link
        fedilink
        arrow-up
        0
        ·
        9 months ago

        Fancy autocomplete may be nothing more than fancy autocomplete at the moment, but that hasn’t stopped lawyers, college students, and kids everywhere from taking anything it shits out as factual.

        The “most of what this says is complete nonsense” disclaimers are small (except on OpenAI, but there they take the form of popups I think most users will subconsciously dismiss) and the AI query box is right up there with the actual web search box.

        Most of the world has no idea how AI works, what a training set looks like, and what the implications of prompt temperature are. People do believe everything AI says, and it’s only getting worse from what I can tell.

    • stoneparchment@possumpat.io
      link
      fedilink
      arrow-up
      0
      ·
      edit-2
      9 months ago
      • it’s true that this would mislead children, but the model could hallucinate about literally anything. Especially at this stage, no one-- children or adults-- should be uncritically accepting what the model states as fact. That said, I agree LLMs need to improve their factual accuracy

      • Although it is highly debated, some scholars suggest Queen Charlotte might have had African ancestry, or that she would be considered a POC by today’s standards. Of course, she reigned in the 17-1800s, but it isn’t entirely outlandish to have a “Queen of Color”, if we aren’t requesting a specific queen or a specific race

      • People of color did live in England in the middle ages? Like not diverse in the way we conceive now, but here are a few papers discussing the racial diversity at the time. It was surely less intermingled than today, but it’s not like these images are impossible

      • Other things are anachronistic or fantastical about these images, such as clothing. Are we worried about children getting the wrong impression of history in that sense?

      • Of course increasing visibility and representation of all kinds of marginalized people is important. I, myself, am disabled, so I care about that representation too-- thanks for pointing out how we could improve the model further. I do kinda feel like people would be groaning if the model had produced a Queen with a visible disability, though… I would be delighted to be wrong on this front :)

      • KeenFlame@feddit.nu
        link
        fedilink
        arrow-up
        0
        ·
        9 months ago

        These are not hallucinations. The image generator system prompt has been intensely altered to mix all races and genders. The model is probably not inaccurate up until it being misused. The misuse could be at any level of interaction, so it’s very misleading to base it on such an example

      • ji59@kbin.social
        link
        fedilink
        arrow-up
        0
        ·
        9 months ago

        I know that POC lived in England and it was possible to meet someone like that. But I would prefer if the model gave most possible, most general answers. If I ask for an image of a car I would like to give me four-wheeled red or gray or green car, not three-wheeled pink car just because there exist some car like that.

        • stoneparchment@possumpat.io
          link
          fedilink
          arrow-up
          0
          ·
          edit-2
          9 months ago

          That’s valid! I agree. I think in this case it would be reasonable for the model to give multiple (or like, at least one, jeez) images with white queens. I don’t disagree with anyone in that sense. I just also don’t think it’s worth pitching a fit when the dumbass model that has been trained to show more racial diversity produces (frankly comical) hallucinations.

          The ethos of the trainers is a good one. Attempting to counter the (demonstrated, measurable) bias of many models toward whiteness is a good choice. I prefer that the trainers choose to address the bias even if it (sometimes, in early versions) makes the model make silly mistakes like this. That’s all.