• tourist@lemmy.world
    link
    fedilink
    English
    arrow-up
    17
    ·
    3 hours ago

    oh my fucking god

    just dial the doomsday clock to 1am, because we’ll probably miss it anyway if it’s up to a few morons and glorified predictive text, both seemingly in “fucking kill everything” mode

  • DrunkenPirate@feddit.org
    link
    fedilink
    English
    arrow-up
    7
    ·
    3 hours ago

    I wonder when does the AI also hallicunate the sources such as webpages, pdfs, reports to prove its hallucinated point.

    • Zwuzelmaus@feddit.org
      link
      fedilink
      English
      arrow-up
      6
      ·
      3 hours ago

      Not far. Current LLMs have kinda stubbornness. They cling to their own conclusions even after you told them they are wrong. So if you ask them for proof, then most times they come up with something that’s convincing.

      • brsrklf@jlai.lu
        link
        fedilink
        English
        arrow-up
        2
        ·
        2 hours ago

        You know the thing where AI companies describe their alignment direction as “Honest, Harmless, Helpful”?

        The only one that’s marketable to the idiots buying AI services is Helpful. So that’s always top priority even when it goes against the other two.

        Average decider wouldn’t stand an “assistant” that sometimes answers “I don’t know”. So instead agent providers train their already hallucination-prone model to be 100% confident in whatever it says. Who cares if it’s true if it looks like it is?

  • Tollana1234567@lemmy.today
    link
    fedilink
    English
    arrow-up
    6
    arrow-down
    1
    ·
    4 hours ago

    sooner or later they are going to have friendly fire, and hallucinate one of thier own US navy vessels as “foreign with nuclear capabilities”

    • NewOldGuard@lemmy.ml
      link
      fedilink
      English
      arrow-up
      20
      ·
      6 hours ago

      Right it’s not that it’s a killer sentient machine, it’s that it’s a faulty statistical model that outputs incorrect stuff all the time and credulous morons believe it uncritically (or even automate a response based on that inherently inaccurate output).

    • Telorand@reddthat.com
      link
      fedilink
      English
      arrow-up
      58
      ·
      8 hours ago

      It doesn’t help that the people currently in charge are fucking stupid, trigger happy, and emotionally insecure.

    • AeronMelon@lemmy.world
      link
      fedilink
      English
      arrow-up
      10
      ·
      8 hours ago

      The only solution is to give the firing codes to the AI. That way we don’t acccidentally use them.

    • nieceandtows@programming.dev
      link
      fedilink
      English
      arrow-up
      3
      ·
      6 hours ago

      Yeah the more these AI companies are making AI seem the scary boogeyman, the more probable it is some disaster is going to happen because people put too much stock into it. Like the school bombing.