• ZILtoid1991@kbin.social
    link
    fedilink
    arrow-up
    7
    arrow-down
    3
    ·
    2 years ago

    Bad comparison. The scale is way greater. I can’t go through multiple terabytes of reference images.

    As an artist, that’s my issue with current models. It shouldn’t be opted-in by default with questionable opt-out options.

    • Knighthawk 0811@lemmy.one
      link
      fedilink
      arrow-up
      1
      arrow-down
      5
      ·
      2 years ago

      but that’s exactly how the models are being trained. by manually going through all the images and describing them.

      and you can’t really complain that computers go through data at a greater scale than humans… that’s the whole point. that argument doesn’t hold

      • zalack@kbin.social
        link
        fedilink
        arrow-up
        9
        ·
        edit-2
        2 years ago

        It’s worth entertaining the observation that these learning models piggyback off underpaid human effort to funnel wealth towards the 1%.

        It’s just that the solution isn’t to stop the tech. That never works. It’s out in the world now.

        The real solution is much harder. We need to overhaul our economy and put in mechanisms to recirculate wealth downwards.