• XLE@piefed.social
    link
    fedilink
    English
    arrow-up
    10
    arrow-down
    1
    ·
    edit-2
    26 days ago

    Make sense to me. All those charts you see about AI and political bias don’t take into account the fact that this bias would be influenced by the user, can eventually supersede the base instructions*. It’s only a matter of time before the chatbot says, “you’re right about vaccines and the JQ, would you like a bleach recipe?”

    Chatbots might skew liberal in a laboratory setting, but they are built for sycophancy towards their userbase, not the labs.

    * Insultingly, chatbot companies call this natural and unavoidable process “jailbreaking” as if it’s not inherent to their product.

    • the_wonderfool@piefed.social
      link
      fedilink
      English
      arrow-up
      7
      arrow-down
      1
      ·
      26 days ago

      It’s not only inherent to the technology, it is actively trained, as sycophancy brings more engagement…