Nadella described the solution as “an architectural imperative to separate the supply of intelligence” from the authority granted to it.

  • schnoopy
    link
    fedilink
    English
    arrow-up
    4
    ·
    49 minutes ago

    Sometimes I read something that makes so little sense I wonder if I am illiterate.

    Meaningfully, in concrete terms, what the fuck are they talking about? It sounds like “Hey copilot please bold the text between this point and that point, make no mistakes” “Oh move the image right” “No undo that, less right” “No I didn’t mean make the text less accurate”.

    They already have tool chains to create docx format documents from decorated emissions (or maybe they emit the XML directly which would explain some of the utterly bizarre documents I have seen). Copilot is (grossly simplified) a chat window, bunch of scripts that are called conditionally, a monitoring loop of emitted text, and an LLM. Copilot cannot run word inside itself. What does that even mean? There’s a chat window interface directly hooked into a running word instance?

    ARRRRRGH

    • supersquirrel@lemmy.caOP
      link
      fedilink
      English
      arrow-up
      2
      ·
      31 minutes ago

      Meaningfully, in concrete terms, what the fuck are they talking about?

      Here I will explain, it is very simple.

      Would you find corporate-slop rationalizing micromanaging workers as if they were intelligent but highly untrustable monkeys intent on chaos, sloth and mischief unless overseen by a panopticon-like structure of corporate surveillance (both digital and human/cultural) to be meaningfully comprehensible in argument in any way?

      I think you would react precisely the same to that kind of business dogma as you would to this AI dogma, both are the same pattern of denying fundamentally addressing the substance of things by partitioning it up into tinier and tinier boxes further and further down the chain of abstraction from management to the point that the humanity of the elements far below becomes more of a thought experiment than a reality.

      In the end the demand on reasonable application of intelligence is placed on the human workers at the bottom of the rung that are tasked with fixing the monumental pile of bullshit that AIs are constantly creating for human workers higher up the chain who attempt to use it to do their work but then get a broken result and demand angrily that it is fixed by a mechanical turk.