I’m sure a LLM trained on Reddit conversations will be a well-balanced, polite, and reasonable conversationalist and not a contradictory mess trying to figure out which brand of spite is most acceptable. Remember in 2016 when Microsoft plugged the chatbot “Tay” into the internet and it went full psycho Nazi in 24 hours?