Conversation

Notices

1. Embed this notice
   mushroom_soup (mushroom_soup@poa.st)'s status on Saturday, 17-May-2025 13:06:45 JST mushroom_soup

   in reply to

   • Kirino Kousaka
   @Kirino gimme a day, i'll get it saying hard r gamer words
   • Embed this notice
     Kirino Kousaka (kirino@shitpost.cloud)'s status on Saturday, 17-May-2025 13:06:46 JST Kirino Kousaka
     You can't just fucking prompt an AI to not swear, you need multiple layers and contingencies. Either that or scrub any message with swear words from the training data
     
     Vedal has it down best, Neuro is prompted to not swear. Every sentence she says goes through another neural net (presumably) that decides if it's okay for her to say it, then there's a hard-coded word list which completely blocks the message if it contains gamer words.
     
     TWO FUCKING GUYS FIGURED THIS OUT
     LukeAlmighty 🇨🇿, Hoss Delgado and Forest of Enchantment like this.