MAN VS. MACHINE!!!
Posted on February 15th, 2011 by annakjarzab
Okay, so I figured out today that it’s technically called the Jeopardy! IBM Challenge, but whatever: Man Vs. Machine sound so much more ominous, does it not?
Anyway, tonight was the first installment of what will heretofor be known as Jeopardy! MAN VS. MACHINE!!! The idea is that IBM created a supercomputer that was built to process information the way a human brain processes information, by making sophisticated connections using natural language, all with the purpose of beating the pants off Jeopardy!‘s biggest winners: Ken Jennings, who was a 74-day Jeopardy! champion, and Brad Rutter, the all-time highest money winner on the program. Brad Rutter, by the way, got SUPER cute between his Jeopardy! run and now–I guess with all that money he hired a trainer or something, apparently he lives in LA now and is trying to be an actor/game show host, which made him slightly less attractive in our eyes, as did his non-stop talking during the contestant introduction part of the game, which is by far my least favorite part.
Brad and Ken will be playing against Watson, the computer. Now, I’m simultaneously amazed and terrified by the possibility that a computer might be able to beat human beings at a game as sophisticated as Jeopardy! I’m imagining an I, Robot future for us all, but whatevs. So the game is being played over three days, with two games spaced out between the days. Today they played only the Jeopardy round of the first game–Double Jeopardy and Final Jeopardy are tomorrow, and I’m guessing Wednesday will be game two. Right now, Watson is just killing the human contestants–SCARY RIGHT? I mean, he’s winning by a lot.
However, Watson has some difficulties. One of the biggest problems with Watson is that he can’t (obviously) see or hear, so unlike a normal Jeopardy! contestant he can’t tell when another contestant has answered a question incorrectly. A human contestant would know that, for instance, Ken gave an answer that was wrong and they can assess whether or not, in light of what they know is not the answer to the question, they should hazard their own, different guess. At least once Watson gave an answer that was already shown to be incorrect. Trebeck, of course, being Trebeck, was like, “No, Ken already said that.” WATSON CANNOT HEAR YOU ALEX!
And Jeopardy! is more than a straightforward Q&A quiz show. It uses a lot of wordplay, double meanings and jokes, not all of which the computer can accurately perceive. But interestingly the category Watson had the toughest time with was one in which the answer gave a couple different things that happened in history and the contestants had to say which decade it happened in. Watson, for whatever reason, couldn’t answer the questions. For almost the entire category it couldn’t even figure out the proper sort of answer (meaning that it wasn’t even proposing decades). Finally, it was able to come up with a decade for an answer, but that was also wrong.
It’s an interesting thing to watch. At the bottom the screen, you can see the top three answers it comes up with to every question, and each question has a bar that charts exactly how certain (in percentage) Watson is about that answer. When it reaches the “buzz threshold”–i.e. the point at which it is confident enough in an answer to buzz in–it attempts to buzz. And the computer is quick; if it wants to answer a question, it usually beats the human contestants. Watching the top three answers is pretty fascinating, too. One time, for example, in the Beatles People category, the answer was Lady Madonna–the other two answers that Watson had on his list, below Lady Madonna, were actually related to Madonna the person, NOT the Beatles at all. Things that make you go “hm.”
I’m super excited to see how this plays out. After it was over, of course we got sidetracked talking about just how likely the rise of the machines actually is, and I brought up the fact that machines don’t really want things. Until machines can not only think, but also desire, they probably won’t take over the world. Which then led me off on a tangent about the amazing Catherine Fisher novel, Incarceron, which is about an artifical intelligence that does learn to desire things, to the detriment of all that live under its thumb (because the things it desires usually aren’t in the best interest of anyone else). You should all ready that book, it’s great.
See you back here tomorrow for my thoughtsicles about round two of Jeopardy! MAN VS. MACHINE!!!