You are correct to recognize Donald Michie's (full disclosure: I am his son) work on artificial intelligence developing new insights rather than brute force, as well as the importance of AI communicating these insights to humans (The Guardian view on bridging human and machine learning: it's all in the game, 30 March). This groundbreaking study is significant for the reasons you state, but it also speaks to discussions about whether the rise of robots would lead to their enslavement of humans.
My father argued that future robots and artificial intelligence (AI) must be forced (programmed) to explain what they were doing and why in human-understandable terms. We'll be in serious trouble if that doesn't happen — from a routine standpoint.(What caused the self-driving car to crash?) to the existential.
Jonathan Michie is the head of Kellogg College and a professor of innovation and knowledge exchange at the University of Oxford.
• Your essay was intriguing, but the AI system featured in it, NooK, did not play bridge. First, it skipped the bidding, which is considered the most challenging portion of bridge. There are more conceivable auction sequences than there are possible card hands, and remembering the auction has an impact on how the cards are played. Second, even in card games, NooK assumed the less difficult role of declarer. A robot bridge program took over the defender roles.
NooK's eight champions all played the identical hands against the same robots, and the outcomes were compared. ItIt'd be interesting to learn what incentives these human players had, and whether their performance on the last hands dipped. It takes a lot of effort for us humans to maintain high concentration levels over 100 hands, but not for NooK.
You must be logged in to post a comment.