Clobbering the two best Jeopardy! contestants out there is no small feat. But last week, during one final Jeopardy! round, Watson — the artificial intelligence IBM computer that won a three-day tournament against Ken Jennings and Brad Rutter — answered incorrectly. The machine answered "Toronto" for a clue under the heading "U.S. Cities." So was this some sort of snide remark? Was Watson intentionally trying to knock our nationally sensitive identity? Or is the machine’s gaffe a sign that with the right practice — and a good dose of Canadian trivia — this superhuman, high-tech machine can be beat?
Who better to give us a real competitive edge than Stephen Baker, a former BusinessWeek senior writer and author of Final Jeopardy: Man vs. Machine and the Quest to Know Everything. Baker has been following the development of Watson since its early days as the apple of IBM’s eye.
How did you get the idea to write a book about Watson?
I had been looking for my next book project and was eager to find a real story to tell. Ideally, this gripping story would introduce readers to some technology poised to change the world (if that wasn’t asking too much). So one day in late 2009 I was having lunch at IBM, and people started talking about Watson, the Jeopardy! machine. This was my book! It even featured a do-or-die championship match. My only fear was that someone else had already claimed it. Luckily, no one had.
What is the machine’s most endearing quality?
Its happy confidence as it botches clues in ways that no person could dream up. It is then, when Watson errs, that you realize just how foreign our knowledge and language remain to this machine, and how surprising it is that gaffes don’t happen even more often.
What is the best way to trump it?
That’s easy. Just step out of the Jeopardy! format and ask it a question most people would consider simple, perhaps, "What’s today?" or "Where are we?" Watson will launch an exhaustive search through millions of documents with references to time and place, never comprehending that the question requires something it utterly lacks: awareness. Tripping up Watson in Jeopardy! is much more difficult. But it does have troubles with clues involving multiple levels of analysis. This type of arithmetic clue, for example, ties it into knots: ____ Wonders of the World x ____ commandments =? (What is 70?)
Watson mistook Toronto for an American city. Is this a matter of semantics, or should we be worried that America is trying to assimilate us?
Poor IBM. Before the match, the team building Watson was weighing the potential damage to IBM’s brand from a silly gaffe on the part of the machine. If Watson had low confidence in a response, mightn’t it be better simply to raise the white flag and say, "I don’t know?" Humans, you see, are blessed with common sense and can make reasonable guesses. But when Watson starts guessing, it comes up with the most outrageous responses. In the end, instead of quieting Watson in these cases, they had it add question marks, as if to announce that it was only a guess. In any case, Watson was stumped by a clue about U.S. cities. Yet compared to some of the dumb answers I’ve seen (It once confused Oliver Twist with the Pet Shop Boys), Toronto wasn’t so far off. It is a city, after all, a North American city, even if it’s on the other side of a border.
Who do you think would win at a round of pure Canadiana trivia? Margaret Atwood or Watson?
That’s a tough one. I would suspect that in depth of Canadian knowledge, Margaret Atwood would clobber the computer (a dilettante if there ever was one). But I’m not so sure about her range. Is she up on the ski lingo at Whistler, seasonal festivals in Nunatsiavut, hurricanes that have battered the Maritimes? If so, she’d probably fare well against Watson — assuming she’s fast on the buzzer.