should i train ai on search queries (moral?) to determine what it’s asking, or should i determine it algorithmically