Artificial intelligence mavens have often pointed out that it’s easier for a computer to play master-level chess than to pick up the pieces and put them on their intended squares.
But now even this final redoubt of human exceptionalism is falling to the machine. Yesterday in Toronto, at the AAAI Robotics Fair, Kenneth Regan, a computer scientist and a noted chess master, lost two games to a robot that moved the pieces all by itself.
Regan, a professor at the University of Buffalo, writes in an email that an optical system “identified my move and fed it to the Houdini chess engine, and then the Lego-build arm executed the engine's reply. Apparently as close to ‘The Turk’ as anyone has come.”
“The Turk” was a mechanical contraption topped by a turbaned and mustachioed wooden head that physically moved pieces on top of its base, a cabinet that opened to show that nobody was inside. Of course, someone was inside—a chess-playing dwarf, of master-level strength it appears, given that The Turk defeated Benjamin Franklin (a very good player) as well as Napoleon Bonaparte (a very bad one).
No less than Edgar Allen Poe observed the contraption in the 1830s and called it out as a fraud. He was right, but for the wrong reason: He thought that any machine able to play chess at all would necessarily play it perfectly.
Not so. The hard part is to put the pieces where you want them to go.
According to Regan, his robot overlord did have a bit of trouble holding onto the knights, the only truly asymmetrical piece. At least the machine didn't call it a horse.