IBM's Watson On Jeopardy!

Published on Feb 16, 2011 by Jamie Munro

I've been a fan of Jeopardy! for quite some time now.  Even more so over the past year.  Now that I am the Father of twins, going out at night just isn't as easy as it was!  When I first heard about Watson being on Jeopardy, probably about 3 or 4 months ago, I was extremely excited.  Actually watching it over the past two nights has been just as thrilling.

Learning about the technology behind it, being a geek and trying to think and understand some of the algorithms used, watching it think and tell you it's best guess and a little bit about the process to find it has been nothing but entertaining.



In case you haven't watched it yet, spoiler alert coming, Watson has absolutely dominated the contestants.  With 30 Jeopardy and 30 Double Jeopardy questions being asked over the past two nights, I would estimate that Watson rang in first for probably 50 of them and getting only a handful of them wrong!  Not only that he got all 3 Daily Doubles!  Granted he had some pretty weird betting numbers (clearly a random number generator).

Even with Watson's, what seems like an insurmountable lead, there are clearly a few issues that IBM needs to work out.  Don't get me wrong, I was impressed, but getting Final Jeopardy wrong, not just wrong, but embarrasing wrong isn't a good thing.  During the show they talked about using this for medical research and diagnosing patients.  I wouldn't want Watson making the same kind of mistake.  In case you missed it, the Final Jeopardy question was asking for a US City with two airports named for different events involving World War 2.  Somehow Watson thought the answer was Toronto!  Granted it had a lot of question marks because it was very unsure.  But a Canadian City being your number 1 filtered down result, yikes!

One of the other ideosynchroncies that I noticed was on part 1 during the Jeopardy round.  The category was "Decades" and the clue was telling you the Empire State Building was constructed and a few other hints.  Watson was only 50% sure that the answer was the 1930s.  Any human looking at the possible filtered answers would immediately bump that up to 100% accurate simply because the category was "Decades".  None the less, Watson's domination was nothing but impressive!

Finally, what I thought was most interesting was my side research.  My wife, like most people, know how to use a computer and surf the Internet; however, she was having a really hard time comprehending how Watson knew all of this information and could come up with one answer.  My simplest explaination to her was, it's just like asking Google a question.  That got me thinking, Google is by far the leader in search technology, let's put it to the test!  During last night's program I directly typed a few questions into Google.  Since Google isn't built to give back one word answers, I had to manually look at the results and find it, but I tried about 3 questions and somewhere in the top 10 results was the answer to the question!

More on this tomorrow, don't forget to catch the final part in the three part series tonight!

Tags: jeopardy | watson | ibm | Theory

Related Posts

blog comments powered by Disqus