Tuesday, June 22, 2010

Who is Watson? A Tool.

If Watson were to be connected to the Internet and coupled with the massive databases compiled by Google, its power would be amazing...
I had the same reaction that this reader did of the recent New York Times Magazine article about IBM's attempt to create an artificial intelligence capable of beating humans at Jeopardy, a game more subtle and nuanced than the purely logical chess.  Nicknamed "Watson" after the father/son duo who led IBM for more than 50 years, the system uses thousands of concurrent algorithms to analyze millions of digitized texts and ranks potential "questions" for the Jeopardy "answers" by their probability of being correct.  

I was going to try and incorporate Google Books and the upcoming Google Editions into this post, but at 500+ words, it was already getting lengthy.  Being inclined to think about how to maximize the return given a certain set of parameters, I naturally wanted to pair a system capable of beating the best Jeopardy contestants with the soon-to-be most comprehensive collection of human information.

While the reader chose to worry about the possibly darker route--a la Kubrick, Cameron, and the Wachowski brothers...
(and closer to the "Skynet" concept than we would be comfortable with). 
... I am more prone to believe that such a system would be closer to C-3PO, especially at ~20 seconds in this clip from The Empire Strikes Back.

And here's why: Watson merely measures the connections
that were already made by humans.  He is built for an unbelievably minute portion of the human experience, and I'm sure that sales of Trivial Pursuit will continue at their current rate. Although I admittedly played my fair share of quiz bowls--alright, I was the nerdy captain of a state finalist team--I always knew that such activities had almost zero correlation with life success.
Being able to find an answer doesn't make you intelligent or creative or an intelligent problem solver; it makes you a tool.  And that's all Watson is, a tool designed to quickly answer trivia.  And even as a tool, he's not as good a one as the best human at his specific task.  If and when he plays Ken Jennings, my--and Watson's designers'--bids are all or nothing on Ken.
That said, I'm still incredibly excited about the implications for Watson on several fields I'm hoping to impact after Harvard.  The article lists potential applications for a verbal input Watson-esque system, including the ability to quickly look up legal cases and findings that could possibly apply to a lawyer's current client.  As a time-saving device, such power might rival "Google" as the verb for "search," but it would still take a human brain to come up with the best way to present the evidence.

As a teacher, I was breathlessly waiting for someone to realize that the people who ask the hardest questions and have the most curiosity--children--are often the ones with the least resources to find the answers.  I can't wait to push companies like I.B.M., Google, and Apple to team up for a learning tool that combines their respective expertise.

I mean, imagine if you had this as a kid, the full breadth of human knowledge and experience, and the user interface was simply, "Watson, why is the sky blue?"