This AI can tell true hate speech from harmless banter

The algorithm can figure out when the word “moron” is used in good fun, and when a seemingly innocuous word like "animal" is being used to demonise people