• FierySpectre@lemmy.world
    link
    fedilink
    arrow-up
    11
    ·
    7 months ago

    They actually tested that, trained a model using only the outputs of the previous generation of model. It takes less iterations of that to completely lose quality than you’d think.

    • jacksilver@lemmy.world
      link
      fedilink
      arrow-up
      4
      ·
      7 months ago

      Do you have any links on that, it was something I had wanted to explore, but never had the time or money.

    • WarmSoda@lemm.ee
      link
      fedilink
      arrow-up
      3
      ·
      7 months ago

      They go insane pretty quickly don’t they? As in it all just become a jumble.