• SoftestSapphic@lemmy.world
    link
    fedilink
    English
    arrow-up
    13
    arrow-down
    3
    ·
    edit-2
    16 hours ago

    You are demonstrating in this comment that you don’t really understand the tech.

    The “efficient” models already spent the water and energy to train, these models are inferior to the ones that need data centers because you are stuck with a bot trained in 2020-2022 forever.

    They are less wasteful, but will become just as wasteful the second we want it to catch up again.

    • Sibyls@lemmy.ml
      link
      fedilink
      English
      arrow-up
      2
      arrow-down
      6
      ·
      15 hours ago

      You are misunderstanding the tech. That’s not how this works, models are trained often, did you think this was done only a few years ago? The fact that you called them bots says everything.

      You’re just hating to hate on something, without understanding the technology. The efficiency I’m referring to is the MoE architecture that only got popular within the last year. There are still new architectures being developed, not that you care about this topic but would prefer to blindly hate on what’s spewed from outdated and biased news sources.

      • SoftestSapphic@lemmy.world
        link
        fedilink
        English
        arrow-up
        5
        arrow-down
        2
        ·
        15 hours ago

        Yeah nah

        Same shit people said in 2022

        In 3 more years you’ll be making the same excuses for the same shortcomings, because for you this isn’t about the tech, it’s about your ideology.

        • Sibyls@lemmy.ml
          link
          fedilink
          English
          arrow-up
          2
          arrow-down
          4
          ·
          14 hours ago

          You make weird assumptions seemingly based on outdated ideas. I’ll let you be, perhaps you need some rest.