• A_norny_mousse@piefed.zip
    link
    fedilink
    English
    arrow-up
    9
    ·
    2 天前

    The researchers proposed an “amplification spiral” in which chatbots mirror a user’s language, generate increasingly personalized responses, and validate unsupported beliefs instead of challenging them.

    I find it weirdly fascinating that people do not seem to realize the simple fact that so-called AI is based on human input, both in the moment and through the data it was fed.

    But I bet some will start demanding external “senses” for AI and possibly even agency (partly already happening with agentic stuff).

    • Natanael@infosec.pub
      link
      fedilink
      English
      arrow-up
      2
      ·
      1 天前

      “my multimodal model has embodiment in my… don’t ask what kind of robot body I put it in”