• urmums401k [she/her, they/them]@hexbear.net
    link
    fedilink
    English
    arrow-up
    11
    ·
    edit-2
    1 month ago

    They did the math determining that it was a dead end. They were right. Yes it requires a certain bigness of data set to function, but after that the returns diminish logarithmically. It basically peaks at the software I’m using to type this on my phone. These massive data centers may as well be using that energy and water to breed monkeys and make typewriters.

    There are other niche uses, but this is mostly it.

    • Hexboare [they/them]@hexbear.net
      link
      fedilink
      English
      arrow-up
      9
      ·
      edit-2
      1 month ago

      These massive data centers may as well be using that energy and water to breed monkeys and make typewriters

      What if we breed monkeys to be our telepathic typewriters my-hero

    • urmums401k [she/her, they/them]@hexbear.net
      link
      fedilink
      English
      arrow-up
      1
      ·
      edit-2
      1 month ago

      I should clarify; the utility of large language models isn’t in the utility of the technology, but the utility of the propaganda. It’s selling a dream, so we stay asleep, it’s selling the liberal holy Grail of the endless perfect excuse and abdication, and the fascist one of a perfect inhuman system, simultaneously an impossible future and a bullshit homogenized soup of half remembered past and a composite of us all in which we are a consciousness wothout a body (but without all thar pesky interiority)that will grind us all into the mud

      Its not about being good at stuff. If you see an “AI” engineer, offer them a cocktail made with your favorite whiskey and drain cleaner.