I have been thinking about why people are unimpressed with Watson.  I'm also here to tell you that everyone should be in awe of what Watson is doing.

The chess match between Deep Blue and Garry Kasparov

 In the end, though, Chess is just like Othello, Checkers, or Tic-Tac-Toe for that matter.  All of these games have specific rules that govern what you are allowed to do on each turn.  The way computers play these games is to search through all of the possible moves.  For a game like Tic-Tac-Toe, it is easy to figure out every possible combination of moves and to assure that you do no worse than a draw in every match.

 More sophisticated games have more possible moves, and that makes them a bit more difficult for computers.  Still, it is theoretically possible to 'solve' each of these games the way that Tic-Tac-Toe has been solved.  In 2007, for example, the game Checkers was solved, as computers were able to look through every possible move in the game.

 For chess, there are more possible moves in the game than there are elementary particles in the universe, so it hasn't been solved yet.  But computers still search through vast sets of possible moves and then use expert ratings built into the system to help the system evaluate possible moves.  In addition, the computer has a huge library of openings and endgames to help the computer win.

 All this is to say that Deep Blue was an impressive technical feat, but it wasn't nearly as smart as it looked.

 Another reason why people are less than impressed with Watson is that computers have also gotten good at making recommendations for us.  If you purchase things from Amazon.com, then eventually you start getting recommendations of other products from the system there.  Many of these recommendations are pretty impressive.  I have discovered new music and authors by following these recommendations.

 In this case, the computers are doing something interesting.  People make many purchases from Amazon.  You can track their purchases and find patterns of things people tend to buy.  Amazon generates a huge volume of sales, and so they have lots of data to work with.

 What makes these recommendations so successful is not the intelligence of the computers that analyze the data, but the similarities across different people.  The computer is just detecting basic statistical properties of the patterns of purchases people make.  The reason those statistics are useful to you is that liking a particular kind of music, or art, or literature really does predict other things you will enjoy as well.

 The third thing that dampens your enthusiasm for Watson is the internet.  When I was in college, there was an all-night trivia contest every semester that was broadcast on the college radio station.  The team that won the previous semester's contest would host the next one.  Teams would stay awake and try to answer all sorts of obscure questions like the names of the 64 colors in a box of Crayola crayons.  As I recall, that question was asked at about 2am, and we ended up calling a toy store in Australia to get the answer.

 Now, of course, these kinds of trivia contests are no fun.  People just Google the answer to everything.  At dinner, conversations about who starred in a particular movie used to energize discussion.  Now, somebody whips out a smart phone, hits IMDB and gets the answer. 

 So, it makes sense that Watson would know lots of stuff if it has access to the internet.  Anyone should be able to win at Jeopardy if they can Google the answers quickly.

 And that's where you're wrong.  What Watson is doing to answer these questions is truly impressive.  I'll talk about why in the next blog entry in this series.

 Follow me on Twitter.