This week, Jeopardy had a unique series of shows air Monday through Wednesday (just in time for sweeps, coincidence?) Two former Jeopardy champions, Ken Jennings and Brad Rutter, played two games against Watson, an IBM supercomputer.
If you don't know what happened and don't want spoilers, stop reading now (of course I think most of you are well enough informed that you probably ran across the coverage by now!)
Everyone gone who hasn't seen it? Good.
Watson kicked this asses. You might think that a computer would easily win at Jeopardy, but it's a much more complicated problem that you think. Jeopardy clues are frequently based on puns and wordplay. The answers often give clues to the question. Sometimes I can think of more than one right question, but a combination of the answer and the category guides you to the right one. Computers are very good at looking up information in a database. Understanding natural language, puns and wordplay, that's a much more difficult computer program to write!
One of the interesting things they did was have Watson produce a graphic showing his top three choices for answers and his confidence level in each. Watson had a threshold to buzz in, he had to be confident of his answers or he wouldn't buzz in. Watson was very quick on the buzzer and you could tell Jennings and Rutter were frustrated. They knew many of the answers and were just beaten to the buzzer. Watson missed final Jeopardy the first night but was very far ahead and made a small wager (I took consolation in the fact that I knew that final Jeopardy!)
Watson made mistakes and didn't always understand the question. Once Watson said the same incorrect answer that was just given by Jennings.
Okay, so computers are now superior right? Well, one thing I noticed is that the second game was much closer than the first, Jennings and Rutter seemed to learn how to play against Watson. What would have happened in a third game? With practice, could they beat it? Does Watson have the ability to learn (without the input of external programmers) to stay ahead of the human challengers? It was an interesting exhibition, but I think these are a few questions that would be fun to explore as well.
And the quote in the title of my blog? Ken Jennings added that to his final Jeopardy answer on the second day. It's a reference to the Simpsons episode when Homer flies on the space shuttle. Some ants in a science experiment escape and one of them floats in front of a camera during a live broadcast leading Kent Brockmant to conclude that the shuttle was taken over by giant ants and say, "I for one welcome our new insect overlords". I don't think Watson understood the joke.
UPDATE: Here is the clip from the Simpsons
Subscribe to:
Post Comments (Atom)
Posts
3 comments:
I knew the computer had won, I didn't know the details. I can't imagine the programming it took.
I remember reading about Artificial Intelligence in the 60's. They said it would be here soon and that if we were lucky, the machines would keep us as pets.
There was a magnificent show on PBS about the programming of Watson and what had to take place before he/she/it was ready to compete. If you didn't see the program, you should look for it.
When they thought they had it, they hired an emcee and ran trials against humans. The initial trials failed. They had to teach it to "listen" and not repeat wrong answers already given. I noticed that it did that once anyway. They also had to program in that percentage-of-correctness thing because it wanted to ring in every time. (and I thought that was a man thing) ;>
I also gotta say what a brilliant publicity stunt it was, both for Jeopardy which got some of its highest ratings ever and IBM which got loads of positive publicity.
I will see if I can find that show...was it Nova Science Now or something?
Post a Comment