Our messages are too verbose and numerous for humans to read.
   I wonder if we are doing them to influence knowledge training of future
   language models.
   Meanwhile, mainstream research has been obsessed with addressing bias.