IBM Supercomputer Beats Ken Jennings At Jeopardy
This is interesting:
IBM Corp.’s “Watson” computer defeated the top two Jeopardy! champions of all time during a practice round Thursday, showcasing the future of artificial intelligence.
IBM Corp.’s “Watson” computer defeated the top two “Jeopardy!” champions of all time during a practice round Thursday, showcasing the future of artificial intelligence and setting the stage for filming Friday of two official matches of man vs. machine on the game show.
Undeterred by the lights, cameras and dozens of spectators Watson coolly beat Ken Jennings and Brad Rutter to the buzzer right out of the gate, answering the first four questions correctly.
Adding insult to injury, the category was “Chicks Dig Me.” Granted, the former “Jeopardy!” champions were rusty, pointing out during a question-and-answer session afterward that their hands hadn’t gripped a buzzer in five years.
Still, Watson’s victory offered reason for IBM researchers, who have dedicated themselves to the project for the past four years, to breathe an initial sigh of relief.
At least for now.
The final score was $4,400 for Watson, $3,400 for Jennings and $1,200 for Rutter, close enough to make the official showdown anyone’s — or any computer’s game.
I agree with Ann Althouse that the computer would seem to have a bit of an unfair advantage when it comes to buzzer-response, which is often the key to winning on Jeopardy since you get the first shot at coming up with the correct question.
Anyway, here’s the video:
Former Jeopardy contestants say it’s all about buzzing in at the right time. How did the computer buzz in? Machine will have an advantage if programmed properly. I wonder how many the computer got wrong versus how many the humans got wrong?
This is the problem with Jeopardy! in general: it’s based in large part on figuring out the timing of when they’ll flip the switch that lets you buzz in. While that can be “fair” when it’s all humans playing against each other (though note that whomever won the last game will have a huge experience advantage against the others, esp. considering they film 5 games in a row right after each other), with a computer it’s just an enormous advantage that knowledge can’t overcome (unless the computer was programmed to have to guess like the humans, which it doesn’t appear to be).
I’d be far more interested to see what a Watson-esque computer could do playing the more academic game of quizbowl, which allows players to ring in during the question and groups clues from more difficult to less difficult. Thus, the computer’s calculations would have to be even faster and would have to take into account a kind of Bayesian updating about the likelihood of a given answer as new information was read.