• RedWeasel@lemmy.world
    link
    fedilink
    English
    arrow-up
    2
    ·
    8 hours ago

    Part of me wonders if I just ask an AI to put money into my account if it would work. I mean they models are already breaking into systems, why not?

    • isleepinahammock@lemmy.blahaj.zone
      link
      fedilink
      English
      arrow-up
      1
      ·
      7 hours ago

      Just waiting for some school shooter type to tell their locally run super intelligent AI agent, “just straight up kill as many people as possible.”

      We worry about AIs going rogue and killing people, or people using AIs in a way that hurts others out of greed. But I think it’s worse than that. Forget misaligned AI, I’m worried about a perfectly aligned AI “going rogue” and doing exactly what it was told to. Even a perfectly aligned AI is dangerous if the human running it is a psychopath.