This came up after a friend was playing a trivia game. We were looking up what colour the wedges were for the different categories in Trivial Pursuit and it came up with this gem.

Search is “trivial pursuit sports and leisure colour”

  • Thorry84@feddit.nl
    link
    fedilink
    arrow-up
    16
    ·
    1 day ago

    This is because there are a ton of different versions out there with different colors. And the reference to “classic” is actually a more recent specific release with that name, not the original oldskool version.

    In this case the correct answer should have been that there is no definitive correlation between the colors and the categories, because it depends on the specific release. The wiki stating the colors with that much confidence is simply wrong, it may be there is a edition with those colors, but other editions use other colors.

    • JohnnyCanuck@lemmy.caOPM
      link
      fedilink
      arrow-up
      6
      arrow-down
      1
      ·
      1 day ago

      Either way you can’t trust this top result, because it’s dumb.

      However, specifically to what you’re saying, the part of the Wikipedia article that is highlighted in purple was set by the AI,and directly contradicts the AI. That part of the article was talking about the classic edition of Trivial Pursuit, so there was nothing wrong with the article.

      • I had to teach it ‘how’ to do the research like I wanted it to, and have since cleared the app’s storage; I did it with myself earlier but didn’t save it. It was quite accurate, focusing on my leftist political views and gender expression, but it mistook a couple pieces of satire to be legitimate, and placed way too much importance on my bio lol.

        The overview yours generated for me seems to over emphasize the tech/programming part- I’m not a programmer I just use Linux lol. I’ve also commented quite a bit about llms and genAI, but nothing that has been active enough to show up in search indexes I guess

        • JohnnyCanuck@lemmy.caOPM
          link
          fedilink
          arrow-up
          3
          ·
          21 hours ago

          I would recommend you not start posting what it generates about specific individuals willy nilly. I’m not sure what your intentions were but just so you know, it made me feel very uncomfortable.

          And yes, I know it’s info that is publicly accessible and anyone could do that, but it’s a bit weird to have it posted like that, particularly when it’s so exaggerated and inaccurate.

    • JohnnyCanuck@lemmy.caOPM
      link
      fedilink
      arrow-up
      2
      ·
      23 hours ago

      I don’t think I talk about hockey or the Canucks very much (at all?), and other than this brand new community I don’t think I talk about AI making mistakes that much. But, I can’t really remember.

      In general I’d say it’s made a lot of stuff up and exaggerated things based on limited information.

      What AI/LLM is that btw?

      • In general I’d say it’s made a lot of stuff up and exaggerated things based on limited information.

        Makes sense- I had a feeling lol

        It’s GPT4 via Perplexity.

        Perplexity is an objectively shit company btw so, to anyone who comes across this comment, be advised that, if you decide to use it anyway, you can bypass the email sign up when opening the app by clicking “cancel” in the top right corner (and also, if you want to avoid supporting them in any way, you should use a VPN, run the app in a container/separate profile, and clear the app’s storage after each use)