It had to happen eventually: once you get a smart enough chatbot, you want to get it talking to itself and see what happens. What follows shouldn't be any surprise, for two reasons:
1. The developers of the chatbot are attempting to simulate human conversation, and human conversation is clogged up with ego.
2. The developers of the chatbot are basing their learning database on real conversations on the Internet, and the Internet is clogged with assholes.
So it shouldn't be a surprise when the conversation descends rapidly into bickering.
This is not artificial intelligence, although this is what people are trying to pass as artificial intelligence. This is only artificial learning, where programmers optimize search algorithms to pull suitable answers from a massive database.
True artificial intelligence is situational. The bot responds based on the context of the discussion rather than on correlations with stored contexts. The difference is subtle but distinct.