• koper@feddit.nl
    link
    fedilink
    English
    arrow-up
    0
    ·
    1 year ago

    Congratulations, you are technically correct. But does this have any relevance for the point of this article? They clearly show that LLMs will provide false and misleading information when that brings them closer to their goal.

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

      Anyone who understands that it’s a statistical language algorithm will understand that it’s not an honesty machine, nor intelligent. So yes, it’s relevant.

      • koper@feddit.nl
        link
        fedilink
        English
        arrow-up
        0
        ·
        1 year ago

        It’s rather difficult to get people who are willing to lie and commit fraud for you. And even if you do, it will leave evidence.

        As this article shows, AIs are the ideal mob henchmen because they will do the most heinous stuff while creating plausible deniability for their tech bro boss. So no, AI is not “just like most people”.