Thanks so much for posting this, @Chris. Ever since the first announcement from IBM I've been looking forward to seeing Project Debater in action. I was struck by several things: the language and speech synthesis were close to flawless--very impressive. The AI managed to marshal many relevant studies from its database in support of its position, which lent it an air of authority that its human opponent lacked (IMO). OTOH, I thought that the AI's rebuttal was weak--it was really more of a restatement of its own opening than a rebuttal of the human's arguments, which were open to several rather obvious lines of attack. It seemed pretty clear to me that although the AI was composing coherent and consistent arguments, it did not really understand the human's claims.
This stuff is developing faster than I had thought possible. I read recently that the OpenAI team has developed a new language model that is so powerful that they are withholding its release because of ethical concerns:
Fascinating stuff, I think.