• Jhex@lemmy.world
    link
    fedilink
    English
    arrow-up
    1
    arrow-down
    1
    ·
    2 days ago

    but to dismiss their utility completely is just idiotic.

    not what I said at all. I simply stated AI answers cannot be trusted without verifying them which makes them a lot less useful

    • 3abas@lemm.ee
      link
      fedilink
      English
      arrow-up
      1
      ·
      8 hours ago

      You’re moving the goalposts. You said you need nuance in how to measure a shirt size, you’re arguing just to argue.

      If a model ever starts answering these curiosities inaccurately, it would be an insufficient model for that task and wouldn’t be used for it. You would immediately notice this is a bad model when it tells you to measure your neck to get a sleeve length.

      Am I making sense? If the model starts giving people bad answers, people will notice when reality hits them in the face.

      So I’m making the assertion that many models today are already sufficient for accurately answering daily curiosities about modern life.

      • Jhex@lemmy.world
        link
        fedilink
        English
        arrow-up
        1
        ·
        4 hours ago

        You’re moving the goalposts. You said you need nuance in how to measure a shirt size, you’re arguing just to argue.

        I said I needed context to verify AI was not giving me slop. If you want to trust AI blindly, go ahead, I’m not sure why you need me to validate your point

        If a model ever starts answering these curiosities inaccurately, it would be an insufficient model for that task and wouldn’t be used for it.

        And how would you notice unless: you either already know the correct answer (at least a ballpark) or verify what AI is telling you?

        You would immediately notice this is a bad model when it tells you to measure your neck to get a sleeve length

        What if it gives you and answer that does not sound so obviously wrong? like measuring the neck width instead of circumference? or measure shoulder to wrists?

        So I’m making the assertion that many models today are already sufficient for accurately answering daily curiosities about modern life.

        And once again I tell you that you can trust it blindly while I would not and I will add that I do not need another catalyst for the destruction of our planet so I can get some trivia questions answered. Given the environmental cost of AI, I would expect a significant return, not just a trivia machine that may wrong 25% of the time