• Vupware@lemmy.zip
    link
    fedilink
    English
    arrow-up
    14
    arrow-down
    2
    ·
    1 day ago

    Despite the problems with these LLMs, I have found them tremendously valuable when it comes to finding sources regarding historical events.

    It’s much easier to find historical evidence that goes against the status quo with LLMs than it is with a conventional search engine. I suspect that this is an unintentional side effect of the technology; the fact that these LLMS are black boxes might have something to do with it.

    • ivanafterall ☑️@lemmy.world
      link
      fedilink
      English
      arrow-up
      4
      ·
      3 hours ago

      I understand what you mean and don’t entirely disagree. If you use it like a calculator with information YOU’VE ALREADY ASSEMBLED. If you rely on it to have all information and/or give an accurate rendering of whatever it’s trained on, you’re probably gonna have a bad time. I’m working on a book that includes, in part, information about the Mormons’ history with slavery. If you ask AI, it will sometimes insist they never had slaves, at all. In fact, it will argue with you until you get legitimately pissed. Ask me how I know.

      • Vupware@lemmy.zip
        link
        fedilink
        English
        arrow-up
        3
        ·
        3 hours ago

        Certainly! that’s why you use them to find sources. I don’t immediately trust anything an LLM spits out.

    • smayonak@lemmy.world
      link
      fedilink
      English
      arrow-up
      8
      ·
      12 hours ago

      They are both great and also biased. Because LLM training data includes vast libraries of history books, including such works as Dark Alliance, it is capable of running down paths that few historians would walk. I recently used an LLM to find an unimpeachable source for the CIA’s connection to cocaine trafficking aircraft. This was not available through a simple Google search because the site was likely deindexed.