• psycotica0@lemmy.ca
    link
    fedilink
    arrow-up
    4
    ·
    17 days ago

    True! But I’m pretty sure most, if not all, LLMs remove all whitespace as, like, step zero of their process as they tokenize stuff. So the LLM is literally blind to the data this contains.

    • Meron35@lemmy.world
      link
      fedilink
      arrow-up
      8
      ·
      17 days ago

      So you’re saying we need to return to the 2000s and start communicating with ASCII art to stop AI scrapers? In.