• Hereforpron2@lemmynsfw.com
    link
    fedilink
    English
    arrow-up
    3
    ·
    1 year ago

    It could be that the people's typing (and therefore the training data) includes so many different unique movements that an "average" of all of them doesn't actually look like any real person's individual typing pattern. Sorta how some one-size-fits-all designs don't fit any one person perfectly.

    • effward@lemmy.world
      link
      fedilink
      English
      arrow-up
      3
      ·
      1 year ago

      This seems plausible to me.

      They should make it smarter and group users based on their swipe technique and have a separate model for each group.