Wednesday, June 16, 2010

Meet Deep Blue's offspring, Watson

I just read a fascinating story in the New York Times about Watson, a supercomputer that plays Jeopardy. I'm sure I enjoyed the article in part because I was a computer science major in college. I am intrigued by tasks that are easy for computers and hard for humans versus tasks that are easy for humans and hard for computers (the most famous probably being the Turing Test). I gained a real appreciation for this difference when I had a project in college to write a computer program to count the number of people in a lecture hall. Wow. Hard. It was around this time that I thought about this exact problem--getting a computer to play Jeopardy.

I didn't find myself reading the article through the lens of a person interested in computer science, though. I was thinking as a teacher. Assuming this technology advances, and assuming it becomes readily available (it's current incarnation will cost companies a million bucks just for the hardware), what are the ramifications for education?

Will we be having the good ol' calculator debate in history class? Will educators fight over whether or not students should be allowed to use their Watson on their science test? In some ways, we're already close to being there with google, although I think there is a fundamental difference between the current technology that necessitates some analysis and the Utopian potential of Watson. On the math end, could/would/should a tool like this change instruction? Would it maybe help resolve the long going debate on what tools are appropriate when? (On somewhat of a side note, I read an article today by R. James Milgram (pdf) who argued that we should continue to teach the standard algorithm for long division to elementary school kids even though calculators (and phone with calculators) are ubiquitous because it introduces them to the important task of estimating and revising)

I'm also intrigued by the different algorithms and checks that this computer does to test the plausibility of its answers. The article only mentions a few, but I wonder if humans do these same (or similar) checks or if these checks could/should be taught.

It made me think...which is almost always a good thing.

1 comment:

  1. I haven't read the whole Watson article but I read the beginning. It reminds me a little of the 20-questions game-playing device/program. The first time I saw it, it kind of freaked me out, but then after a bit the more interesting questions were:
    1) What can I think of to predictably trick it?
    2) How does it work?
    3) How could it get better?

    Actually I think it would be fun to give that to a class and ask them to reverse-engineer it... how would you make something like this? What is the computer doing?