• skooma_king@lemm.ee
    link
    fedilink
    arrow-up
    8
    ·
    5 months ago

    I asked GPT 3.5, Claude 3, Llama 3, and Mixtrial and they all said “eleven eleven.” No fun.

    • lugal@sopuli.xyz
      link
      fedilink
      arrow-up
      3
      ·
      5 months ago

      It’s not that LLM will do that exact thing, they have a big dataset so they know how “eleven” works. If you trained a NN with this small dataset, than this will happen or rather it’s a comment on how AIs work in general. With more complex things something like this might happen and then you have this meme to relate to