• Pyr_Pressure@lemmy.ca
    link
    fedilink
    English
    arrow-up
    8
    ·
    1 year ago

    Perhaps it was some sort of “ethics” avoidance thinking you were trying to use it for betting purposes?

    • Teknikal@lemmy.world
      link
      fedilink
      English
      arrow-up
      3
      ·
      1 year ago

      Yeah it was weird still though because it was answering similar questions beforehand and it actually did a search before acknowledging it was incorrect.

      I think it probably can admit it was wrong but is still limited to it’s first decision.

      • Teppic@kbin.social
        link
        fedilink
        arrow-up
        1
        ·
        1 year ago

        I think if you tell it it is wrong it will always agree with you (regardless how right it might have been, or otherwise). Presumably it is designed that way so it is always non-confrontational.

    • jrs100000@lemmy.world
      link
      fedilink
      English
      arrow-up
      3
      ·
      1 year ago

      Since the AI does not have opinions of its own and lacks the ability to tell fantasy from fact, a human can usually convince an AI that just about anything is true, if they are allowed to argue long enough. The easiest way to make sure this does not happen is to prevent the argument from taking place, either by locking the AI into a safety response or by shutting down.