Americans must be either very excited about the artificial intelligence that IBM has built into a new computer called “Watson” or very scared. This week, when Watson competed on ABC’s Jeopardy against two of the best players in the quiz show’s history, the network got its highest ratings in six years. Crammed full of data from reference books and trained to understand questions in regular human speech, Watson wiped its human rivals out, correctly answering questions on everything from who wrote the Études-Tableaux for piano (Sergei Rachmaninoff) to who designed the Emmanuel College chapel at Cambridge (Christopher Wren).
The feat has been compared to the 1997 victory of IBM’s chess-playing computer, Deep Blue, over Garry Kasparov, the world’s champion at the time. But for a computer to master language is a far more unsettling encroachment on the sanctum of uniquely human behaviours than superiority in a game played on an 8-by-8 grid. Outside the walls of IBM headquarters, Watson has provoked mostly anxiety – over the practical question of what jobs it will destroy, and the metaphysical question of whether talking machines will erode our sense of what it means to be human.
To some extent, this is a misunderstanding. Watson is not a smart machine that has shown its intelligence by winning at Jeopardy. It is a Jeopardy-playing machine which, after years of tinkering by dozens of IBM’s top scientists, now works reasonably well. As big as a room, it combines a supercharged version of the grammar check on your word-processing software with a supercharged version of Google’s “I’m feeling lucky” button.