I'd like to see a computer that could design a human good enough to beat it at chess.
So I remarked in a comment. Mark Liberman kindly responded: "Excellent premise for a story. There could be a prize involved: the Mayhew Prize."
In other words, the whole premise of "computers are getting better than humans" at something is flawed. Of course you can make a computer better than a human at almost any given task.