The big
thing for all of geek culture this week, heck even most of normal
culture, was the big test of the IBM Watson computer system on the
game show Jeopardy. As I sat waiting for the day that it would happen
given that it was announced almost a month in advance I sat in the
same crazed ideas that most of us if we would admit it would be
geeking out over. The usual geek outs went something like this. "This
is it! The battle we have all been waiting for! Humans VS Computers,
the smack down of a life time!" Of course we were all missing
the point and some of us still never got the point. This was not a
match to see if a computer could finally beat a human or vice verca
it was more of a test to see how well a computer could understand
natural language, or to put it in better terms, how the human language
works. I cannot think of a better way of pulling this off then by a
game of Jeopardy and belive me I thought long and hard. Jeopardy has
always been up there in my list of favorite game shows as it plays to
my way of thinking. Clues that dont have a straight out way of saying
things that you have to decrypt and think quick to get the answer to
before anyone else can. In the case of this test it meant that the
processors on Watson would have to work double time to beat a humans
speed which I know sounds like childs play to someone that believes that
someday computers will take over but is actually a feat to someone
like me who doesn't have paranoia about that subject at least.Watson
was designed to look for key terms in text and then run it through
its database seeing all things that would fit with the context before
spitting out an answer that would be suitable. In the case of
Jeopardy if it wasn't sure enough of it's answer it wouldn't even try
to buzz in. To a person just watching with the awe of a smackdown in
mind they would see either a tense battle or ,in some cases like my
aunt, would see the most boring thing they could have ever watched in
their life. To those watching to see what Watson was capable of there
was another layer to the story. As the game went on we were able to
see how sure Watson was of the top three answers he came up with and
if he even had the right ones in mind. Categories involving things
like characters from The Beatles songs, of which I only got one wrong
because I didn't know Johns mothers name, or specific context clues
Watson would excel at while categories like decades for instance he
struggled a bit with coming up with the proper response. For the
researchers in the audience this made a wealth of research data that can
be used to refine the technology even further to make it that much
more useful in practical applications like medicine although it had
better not try to shoot me up with anything or I may kill it. The
point of this whole match was to continue to make computers better at
helping with everyday tasks but still not perfect enough to run us
over. In the end Watson won having a few Daily Doubles that were even difficult for him to figure out but with luck did come up with the answers and even a failed final Jeopardy question that swore Toronto
was in the United States! The point however is not that Watson
won..it's that now he can improve even more in his use case. In the
end Watson was made by humans after all and I think that's something we
can all be just a bit proud of.
The Homepage for all of the content of Samuel Lewis and the other members of the Samcast Crew. Along with some other content we like as well.
Sunday, February 20, 2011
Watson on Jeopardy and what the point was- Commentary
Labels:
commentary
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment