• tetris11@feddit.uk
    link
    fedilink
    English
    arrow-up
    9
    ·
    1 day ago

    what humans have written

    a highly curated selection that panders to the views of the AI’s sponsors

    • Peacepath@lemmy.world
      link
      fedilink
      English
      arrow-up
      6
      ·
      1 day ago

      We are not yet at the point LLM are so dependent on advertising and “partnerships” that they are totally biased in favor of specific brands… not yet…

      Take a time to anticipate how the actual state will soon be regreted…

      • tetris11@feddit.uk
        link
        fedilink
        English
        arrow-up
        3
        ·
        1 day ago

        Im just hoping that the LLMs that are trained to push views that are at odds with reality struggle with basic reasoning, making them inferior models to LLMs trained on a more diverse worldview. I hope.

        • Peacepath@lemmy.world
          link
          fedilink
          English
          arrow-up
          4
          ·
          1 day ago

          My guess is we will see some less convenient models, that gives better result but a “weird” way ; and far more very convenient models, easy to use, but that generate advertisings as a result (maybe even with promo codes!)…

          Remember how internet search have evolved, how youtube has became, how everything went when finance guys decided it was time for harvest…

          Now look at this new field, see how they grow it… what are the odds their plan is to harvest it at some point?

        • kestrel7_7@lemmy.world
          link
          fedilink
          English
          arrow-up
          1
          ·
          1 day ago

          I think the idea that “better training = better LLM’s” is AI industry marketing hype. It’s a pretty shoddy technology on a basic level. There are applications in computer science of LLM tech, but this chatbot shit has got to stop.