I recorded Jeopardy last night because asking my wife to watch 2880 processors play Jeopardy against two humans rather than go out for a Valentine's Day dinner was probably not a good idea. I just finished watching the first day this morning. What follows are some thoughts and observations as the latest man vs machine contest unfolds.
First and foremost, I want to give some credit to IBM. Aside from a fascinating machine, they are doing an excellent job of explaining (as best as can be done in layman's terms) Watson and how "he" works. In particular, I think they are helping to bring some reality in to the viewing public. There are far to many movies and television shows that use computer equipment to create "computers like things" that do not exist. From the infinite zoom capability to the huge multi-physics simulations (run on what computer?), computers are portrayed as convenient plot enhancers with no concept of what is or is not possible. Although multi-physics is on the horizon, infinite zoom, or creating information, is just plain insulting. I find it depressing that there are a lot of very interesting real things possible with computers (and science), probably more interesting than what is fabricated in movies and television, but these concepts and capabilities never reach the masses. Instead, we get nonsense. Sigh. Thanks IBM, you are helping here.
How did Watson do? At the end of the show, which was about half Jeopardy play and half background, Rutter and Watson are tied at $5,000 each. Jennings has $2,000. In the early play, Watson seems to dominate. It did seem that as the questions got harder the other players had a chance to answer. If Watson knows the answer, he seems to respond very quickly. He did get a few wrong and even gave the same wrong answer as Jennings. I had thought, according to the Nova episode, that the same wrong answer issue was solved, but hey, it shows that Watson is embarrassingly real, though not embarrassed. The one category with which he seemed to have the most trouble was identifying time "periods" (i.e. 1920's, 1950's, not exact years) when different events happened.
While I am amazed at how well he can assimilate knowledge, I am also amazed at how much knowledge he holds. There is no Internet connection. The range of topics was quite vast from Beatles songs, to cough, cough, a reference to Grendel and Beowulf (the literary version). The match will continue tonight and tomorrow (Feb 15,16). I'll have more commentary as well.