Given a snippet of dialogue between a chatbot and a human, the system predicts how convincingly human you or I would rate the chatbot’s response. This could be useful for building better virtual assistants.
Turing test
Would it be possible to cut humans out of this process altogether and just automate the Turing test instead? To find out, Lowe designed an AI system that automatically rates how human-like a piece of chatbot-generated dialogue sounds.
He chose 1000 short Twitter conversations, and got human volunteers to add a response. Then he got several chatbots of varying abilities to add their own response. Then he asked a different group of human volunteers to rate all the responses according to how human-like they were.
See the full story here: https://www.newscientist.com/article/2140809-ai-coach-will-train-hopeless-chatbots-to-pass-the-turing-test/