Today’s paper is “AI Makes You Smarter, But None the Wiser: The Disconnect between Performance and Metacognition”. AI users wildly overestimate how brilliant they actually are: [Elsevier, paywalled…
  • CrayonDevourer@lemmy.worldBannedBanned from communityEnglish
    313·
    9 個月前

    You can set them up so that they don’t do that. It’s just the companies running the chat bots set them up that way because that’s what people expect now.

    • selfAEnglish
      9·
      9 個月前

      Pay no attention to ChatGPT, we purposely trained him wrong… as a joke.

      yeah ok

    • BlueMonday1984English
      8·
      9 個月前

      “you can set the sycophancy engines so they aren’t sycophancy engines”

      I’ll take “Shit that’s Impossible” for 500, Alex

    • zbyte64English
      7·
      9 個月前

      It’s still telling you what you want to hear but with a different aesthetic. This will always be the case for systems that can’t even perceive it’s own lies.

    • shalafi@lemmy.worldBannedBanned from communityEnglish
      18·
      9 個月前

      Have you run one? Your claim sounds reasonable to me, but I haven’t tried. Not sure I have the hardware unless I run a shitload in parallel. (If that’s even a thing.)

      • selfAEnglish
        8·
        9 個月前

        Your claim sounds reasonable to me,

        quiz show “wrong” buzzer, poster disappears down a trapdoor