Its partners? Ken Jennings, who is renowned for having won 74 consecutive Jeopardy games in a row, and Brad Rutter, the highest earning contestant (3.2+ million).
I’ve been following the progress of IBM’s Watson for a while now. When I first saw a video about it, I was sure it was a gag… something from The Onion perhaps, and was therefore surprised to learn that it was completely real.
Check out the short clip below. It’s a practice match between Watson, Ken, and Brad:
(Note: You may need to visit the post directly at TylerCruz.com if you’re reading this via e-mail or RSS in order to see it.)
Amazing, huh? That is real! Watson is also a complete standalone program – it is not connected to the Internet, although it basically has an archive of a good portion of it already downloaded.
If you don’t think this is an amazing advance in computer AI, then you might not really realize what you are witnessing here.
First off, simply answering extremely general trivia questions which could be anything under the sun is already very difficult. But what makes it all the more challenging is the way Jeopardy asks questions (or “answers”). They are usually fairly convoluted and are often presented in riddle-form.
Deep Blue vs Kasparov
Most of you will probably remember when IBM’s Deep Blue challenged Garry Kasparov (arguably the best chess player of all time) to a chess match back in 1996.
Garry ended up winning the first match 4-2, but IBM made a few tweaks and came back the next year to win the second match 3.5-2.5.
That was 14 years ago – centuries in computer age, but chess is child’s play compared to Jeopardy. Chess is a simple game with very well-defined rules. A computer just needs a fast processor so that it can basically brute-force its way through every possible combination of moves until it finds a forced checkmate (“mate in 42”).
But with Jeopardy, Watson needs to be prepared to answer basically any trivia question.
What’s cool with Watson, and what separates him from Deep Blue, is that he actually has some basic forms of actual computer learning.
I watched the Nova special on Watson a couple of days ago, and Watson does adapt based on the situation. One good example from the documentary is when Watson was playing a practice game against 2 humans.
The category was a tricky one in which the answers (or “questions”) were all months of the year, but there was no real clue or reference to that. Watson didn’t understand the questions.
But after a bunch of them were correctly answered by the other contestants, Watson picked up a pattern that all the answers were months, and then finally understood what was going on. Confident in its realization, Watson picked the last question of that category and correctly answered "May”.
Now that’s smart.
Remember, he is a standalone program, so it’s not like there’s somebody behind the computer telling him what to do. He was stuck, but adapted to the situation.
I would highly recommend watching the 1-hour Nova documentary on Watson here. Unfortunately, you have to be in the US to watch it
I love how IBM and Jeopardy decided to put the best of the best up against Watson. Just like with Kasparov and Deep Blue, IBM likes to challenge the best of what the human race has to offer, in order to confidently say who is better (at Chess or Jeopardy anyway )
Who Do You Think Will Win?
Cast your vote in the poll below. I’ve set the poll to close once the 3rd day ends.
Personally, I think it will be very close. My money is on Watson or Ken, just because Ken proved his incredible consistency with his sick 74-game winning streak (I mean, come on!)
Check out my photoshopped image below. Not bad for 90 seconds, eh?
I wonder why they spanned it across 3 days… will there be more questions than usual?
The grand prize for this competition will be $1 million with second place earning $300,000 and third place $200,000. Rutter and Jennings will donate 50 percent of their winnings to charity and IBM will donate 100 percent of its winnings to charity.
Tune in tonight!