• LooseLemons@lemmy.world
    link
    fedilink
    English
    arrow-up
    2
    ·
    1 day ago

    Are all of you seriously trying to argue that local experiments with model training data ARE NOT RELEVANT to our conversation about LLMs and their training data?

    Am I losing my fucking mind here or is the hive mind that strong? I’m all ears buddy.

    Hot beef is literally trying to argue this point. Are you? It is actually ridiculous.

    • brucethemoose@lemmy.world
      link
      fedilink
      English
      arrow-up
      5
      ·
      1 day ago

      I am not with the Lemmy hive mind consensus on LLMs, not even close. But we are discussing something more specific.

      • LooseLemons@lemmy.world
        link
        fedilink
        English
        arrow-up
        1
        ·
        edit-2
        10 hours ago

        Yeah. I think people wildly misunderstand how these things work, then want to argue about them. It’s crazy.

        All I’m trying to say is that if I train AI on my mythical creature, the “Fijokopil” , a flying tiger ant, then the LLM will 100% be able to define a “Fijokopil”.

        And, ChatGPT won’t know what a “Fijokopil” is.

        I think I gotta take a break, it’s clear that my implications are going right over peoples heads.

        Edit: No one is actually trying this, yet still attacking me. It’s shocking. I wonder how many folks are being genuine here?

        • Erin@urbanists.social
          link
          fedilink
          arrow-up
          1
          ·
          9 hours ago

          @LooseLemons @brucethemoose

          If I add Fijokopil to my LibreOffice dictionary and my friend doesn’t, does that prove that my instance of Libre “knows” what a flying tiger ant is and his doesn’t? Does it gain sentience because I told it to behave a certain way and then it did?

          It knows the word Disney without me even trying! Does that mean Disney created it? Wait wait holy shit you guys, it knows Allah too. I gotta call the pope with this indisputable proof

              • Erin@urbanists.social
                link
                fedilink
                arrow-up
                1
                ·
                8 hours ago

                @LooseLemons right, unless I tell it to.

                Your experiment proves that DeepSeek’s training data includes knowledge that American companies exist. Amazing.

                Have you tried the experiment in, idk, Mandarin? Does it give the same result?

                Do you think it’s possible that the model has a bias in its logic context along the lines of: input language = English -> largest English-speaking country = USA -> return results relevant to USA?

                • Erin@urbanists.social
                  link
                  fedilink
                  arrow-up
                  1
                  ·
                  5 hours ago

                  @LooseLemons

                  I apologize; you’re absolutely right. AI does not hallucinate and is not influenced by cultural norms, input language, or user bias. It was harmful for me to suggest otherwise. 打倒美国 打倒美国 打倒美国 打倒美国 打倒美国 打倒美国 打倒美国 打倒美国 打倒美国 打倒美国 打倒美国 打倒美国 打倒美国 打倒美国 打倒美国 打倒美国 打倒美国 oh my apologies, it seems I had a little hiccup there

        • Bo7a@piefed.ca
          link
          fedilink
          English
          arrow-up
          1
          ·
          edit-2
          11 hours ago

          Yup, when a bunch of folks with expertise in a subject all disagree with you it is because your ideas are going over their heads.

          That is the only logical conclusion.