Sebastian Benthall, PhD
Everybody has gotten excited about new chatbots based on LLMs. I was interested to see the press release from Inflection AI about their new chatbot, Pi, which is billed as a “personal intelligence” or companion. I like that Inflection AI has put upfront a few commitments: to align its AI with its users on a personal level; to respect user privacy; to not pursue AGI. These are all good things.
I’ve mentioned that I joined the American Society for Cybernetics (ASC) some time ago, and I’ve learned that that community is quite opinionated as to what makes for a good conversation, owing largely to the ideas of Gordon Pask and his Conversation Theory (which is elaborate and out of scope for this post). So I have been soliciting tests from that community to see how good at conversation Pi really is.
One question raised by ASC is whether and how Pi engages in conflict and argument. So I have engaged Pi in a debate about this. The transcript is below.
Sebastian Benthall, PhD