John Henry, the mythical (or maybe not so) steel-driving man who blew out his heart defeating a steam-powered hammer, will always be the icon for any man vs. machine grudge match. For weeks, the game show Jeopardy! has been hyping the ultimate showdown between man and computer, which it is airing on three nights this week.
The epic contest pitting the two best-known and successful Jeopardy! champions — Brad Rutter and Ken Jennings — against a super computer known as Watson, has instead been just a neat parlor trick. To begin with, this is not a fair contest. The computer is provided with the answer (remember, in Jeopardy! the clues are answers and you have to come up with the question) instantaneously as a digital file when the host, Alex Trebek, begins to read it. By the time he is done, Watson already has a response and with a mechanical advantage rings in ahead of Brad and Ken almost every time.
Apparently, the computer is programmed to parse sentences to extract and interpret the meaning, what they call understanding natural language. It is impressive how Watson so frequently comes up with the correct response, but it isn’t all rosie. A little info box at the bottom of the screen shows the top three responses Watson has calculated, with a percentage indicating the confidence Watson has in each response. (Watson won’t ring in if its confidence doesn’t reach a reasonable threshold.) While Watson has had the correct response a very high percent of the time, it is interesting to monitor the second item in Watson’s response list, which is often an answer no human would have even considered. I haven’t written these down, so can’t provide an exact example, but one time, for instance, the second response on Watson’s list was Anna Karenina, when the correct response was a place, I believe.
Additionally, in the Final Jeopardy round of the first game, the clue made it clear the correct response would be a U.S. city, and Watson came up with Toronto — although it did indicate a high level of uncertainty in that response. Sometimes Watson isn’t even smarter than a fifth grader.
It is clear Brad and Ken know these answers, but it is next to impossible for them to beat Watson to the buzzer, so this has never been a legitimate contest.
But the worst aspect, in my view, is that during the first two nights Jeopardy! has become an infomercial for IBM, turning the two greatest Jeopardy! champs into props and Alex Trebek into Suzanne Somers.* We’ll see how tonight’s final show goes — better, I expect, since they will have to fill the whole show with typical game action, meaning most of the IBM preening will be pared out.
*Okay, just to be clear, I don’t mean Alex had a sex change operation.