Comment on Ars Technica Pulls Article With AI Fabricated Quotes About AI Generated Article

limer@lemmy.ml ⁨3⁩ ⁨weeks⁩ ago

• terminatortwo@piefed.social ⁨3⁩ ⁨weeks⁩ ago

  I’m pretty sure it lies

  source

  • atrielienz@lemmy.world ⁨3⁩ ⁨weeks⁩ ago

    Saying Generative AI lies is attributing the ability to reason to it. That’s not what it’s doing. It can’t think. It doesn’t “understand”.

    So at best it can fabricate information by choosing the statistically best word that comes next based on its training set. That’s why there is a distinction between Generative AI hallucinations and actual lying. Humans lie. They tell untruths because they have a motive to. The Generative AI can’t have a motive.

    source
  • limer@lemmy.ml ⁨3⁩ ⁨weeks⁩ ago

    People made AI to lie. When companies make something that does not work and promote it as reliable, that’s on the people doing that.

    When faulty products are used by people, that’s on them.

    I can no more blame AI than I could a car used during a robbery . Both are tools

    source

    • terminatortwo@piefed.social ⁨3⁩ ⁨weeks⁩ ago

      but what if the car lied though.

      source

      • kamenlady@lemmy.world ⁨3⁩ ⁨weeks⁩ ago

        The car’s ai lied.

        source

        • -> View More Comments
      • limer@lemmy.ml ⁨3⁩ ⁨weeks⁩ ago

        It’s exactly like the “guns kill people” arguments. I would like all this AI stuff to go away, the tech is not ready to be used.

        source
• Lost_My_Mind@lemmy.world ⁨3⁩ ⁨weeks⁩ ago

  AI does not lie.

  Last year AI claimed “bleach” is a popular pizza topping. Nobody claimed this as their own work. It’s just what a chatbot said.

  Are you saying AI didn’t lie? Is bleach a popular pizza topping?

  source

  • XLE@piefed.social ⁨3⁩ ⁨weeks⁩ ago

    What it did was assemble words based on a statistical probability model. It’s not lying because it doesn’t want to deceive, because it has no wants and no concept of truth or deception.

    Of course, it sure looks like it’s telling the truth. Google engineered it that way, putting it in front of actual search results. IMO the head liar is Sundar Pichai, the man who decided to show it to people.

    source
  • ulu_mulu@lemmy.zip ⁨3⁩ ⁨weeks⁩ ago

    To be able to lie, you need to know what truth is. AI doesn’t know that, these tools don’t have the concept of right vs wrong nor truth vs lie.

    What they do is assemble words based on statistical patterns of languages.

    “bleach is a popular pizza topping”, from the “perspective” of AI, is just a sequence of words that works in the English language, it has no meaning to them.

    Being designed to create language patterns in a statistical way is the reason why they hallucinate, but you can’t call those “lies” because AI tools have no such concept.

    source
• Sunshine@piefed.ca ⁨3⁩ ⁨weeks⁩ ago

  AI has a high rate of hallucinations…

  source