Friday, February 18, 2011

What Is: A Strangely Routine Game of Jeopardy?

On February 14th, in a grotesque mockery of a day devoted to human emotion, two of Jeopardy!'s most renowned contestants faced off against Watson, an IBM computer programmed specifically to play the game. The game show featured two rounds of play over three days, with Watson easily defeating it's carbon-based opponents Ken Jennings and Brad Rutter. 

I followed the Watson competition with special interest, because I myself am a former Jeopardy! contestant. (My first appearance in the Teen Tournament aired seven years, to the day, before Watson's). Naturally, I was inclined to root for the human beings, not only in my desire to prevent a full-fledged robot-run dystopia, but also because I felt that Jennings and Rutter were on "my team."

Computerized opponents have been around for a number of years - Deep Blue is the famous chess machine that managed to beat Gary Kasparov. But where chess is a mathematical game set in a closed system, Jeopardy! relies on the contestant to know a myriad of obscure facts, and be able to make quick connections. The show is famous for its use of wordplay, and many of the "answers" are only knowable through deciphering groanworthy puns. Where Deep Blue had been programmed with the rules of chess, Watson had to be programmed to understand the intricacies and ambiguities of the human language. Could this be done? 

It turns out that the answer is "yes," but with a few caveats. Watson won a decisive victory over his two opponents of flesh-and-blood, but he also made several embarrassing mistakes that not even the most frail Jeopardy! amateur would commit. 

The first two episodes of the three-episode series were sort of a disappointment. The Jeopardy! team knows a good gimmick when they see one, and keeping in mind that February is a sweeps month, they perhaps extended the "Watson" tournament for longer than they should have. On Monday night, in between the constant explanations of Watson's computational abilities, there was only time for one round - thirty questions in all. 

In fact, Monday and Tuesday started to feel like an infomercial for IBM that just happened to take place on the Jeopardy! stage. The viewer was subjected to interviews with the scientists involved, Alex Trebek's visit to Watson's "brain," and all sorts of propaganda about how great computers are (and IBM computers in particular). I would have preferred to see all these extraneous explanations confined to a half-hour special before the game, but because of the specific demands of syndicated television programs, the Jeopardy! team was instead forced to split one game over two nights, interspersed with this IBM cheerleading. 

That's not to say that Watson isn't interesting, but the never-ending IBM talk destroyed the pace of an already unexciting game. Watson made a few blatant errors - at one point, he rang in with an incorrect answer Jennings had already tried, and in Final Jeopardy, he answered "Toronto" in the category "U.S. Cities." But these errors, though laughable, weren't enough to stop the machine from gaining a decisive victory on the first day. 

Wednesday made for better viewing - this episode featured the entirety of the second round. Not only did this liven up the pace, but Ken Jennings actually managed to give Watson a run for his money, and the final result was still up in the air when it came time for Final Jeopardy. This time, however, Watson was able to interpret the convoluted clue: "William Wilkenson’s An Account of the Principalities of Wallachia and Moldavia inspired this author’s most famous novel." This involves several intuitive leaps - from Wallachia and Moldavia to Transylvania, Transylvania to Dracula, and Dracula to Bram Stoker. All three contestants got the right answer, but Watson finished the day with the most money. 

Are computers smarter than humans? I have to take issue with two aspects of this exhibition. First, according to Jeopardy! rules, contestants are not allowed to ring in until Alex finishes reading the clue and a light turns on above the board. Any contestant who rings in too soon will have their buzzer deactivated for one whole second. A lot of Jeopardy! skill requires finding the right timing to ring in at the correct moment - not too soon, but still before the other contestants. 

Naturally, Watson was able to ring in first nearly every single time. The computer will never err and ring too soon, never doubt himself for that crucial split second, never slip with a sweaty hand and not push the button down the whole way. I'm not sure if Watson knew more than Rutter or Jennings, but he certainly was first to the buzzer enough times to make up for what he didn't know. 

Secondly, the two matches had a dearth of two of Jeopardy!'s most famous categories - "Potent Potables" and "Before and After." Give me a computer who can provide pun-laden answers like "Tom Cruise Control," and then I'll be impressed.

But the strangest thing about the Watson competition was not how impressive the comptuter was, or how scary its intellectual capabilities were. Instead, it was amazing how uninteresting the match was. After all the IBM infomercials, the episode turned into just another round of Jeopardy!. The idea of a computer competing on a game show seems ridiculous, like some lame joke from The Jetsons. But the actual setup was almost boring in its routine - we had the same short quips from Trebek, the same applause when Watson hit a Daily Double, and so on. Watson fit in so well that it almost became dull. It struck me how we really are living in the future in 2011 when we can get a computer to do this and it doesn't even seem like a huge deal. A computer besting humans on Jeopardy!?  Ho-hum, change the channel. Perhaps chess computers have already inoculated us to the idea of machines with the ability to trounce human opponents, but I think it's a sign of how ubiquitous powerful computers have become that a robot game show contestant doesn't seem out of the ordinary. 

Still, I retain hope for the human race. Underneath his Final Jeopardy answer, Ken Jennings showcased his nerdiness by scrawling "I, for one, welcome our new computer overlords." This garnered a huge round of laughter from the audience, and solidified one thing - no matter how good computers are at Jeopardy, human contestants are still more entertaining to watch.