Watson kicked some serious carbon-based brain on Jeopardy.
The three shows that aired were interesting, outlining the background of the IBM project and detailing some of the challenges the engineers faced as they fine tuned the system to compete at a championship level - Watson's AI against Jeopardy's two most successful champions, Ken Jennings and Brad Rutter. I also watched the Nova episode which covered the development of Watson in much greater depth.
The most daunting challenge for the computer scientists and engineers who designed Watson was dealing with the not-so-straightforward stuff that comes up on Jeopardy: puns, double meanings, jokes, homonyms, and Jeopardy's always enjoyable "before and after" questions (extremely terrible example: Pays your dinner bill while organizing your contacts. What is a Credit card file, Alex?). In other words, how do you design a computer that can parse language and tease out nuances?
While watching the games, I felt that Watson had an advantage. There were lots of fact-based trivia questions that required little of the sophisticated language parsing skills that the Watson team worked so hard to develop. Answering these questions with the massive databases that Watson had at its disposal seemed trivial. I assume that the categories and questions were selected randomly, but it would have been more interesting if there were trickier questions. Additionally, and maybe more importantly, Watson seemed to be able to buzz in faster than the other contestants. Consistently, and to the chagrin of Jennings and Rutter.
In general, I think it would have been more fun (and maybe a fairer fight) if Watson was forced to wrestle with more of those tricky language questions, but I came away very impressed. I'm not in the camp of "oh shit, next stop, Skynet!!" I think this is more along the lines of Star Trek TNG's Majel Barrett-voiced ship's computer - a system that can assist humans with sorting intelligently through massive amounts of potentially relevant data in order to answer a question, learn something new, etc.
The thing that really struck me, however, was not the idea of a new, hyper-intelligent computer, but rather was the human ingenuity that went into designing and refining this system. The Nova episode manages to give the viewer a peek into the process, and it's both amazing and inspiring.
0 comments:
Post a Comment