Robot Botches Geography on ‘Jeopardy!’; Why the Machine will never crush the Human

The thing I find most interesting about the whole Watson vs Humans contest on jeopardy is not the competition. I have no doubt that a computer can process things faster, and if taught right, interpret human questions well. There a basic laws of our language and also processors are becoming faster and faster as well as memory cheaper and cheaper.

What's really interesting is when the computer is wrong, it is way wrong, by like a human-like mile. In the case below, the question was asking about a US city, and Watson named Toronto. Watson didn't get Wichita, Topeka, Dallas or Pittsburgh; but instead Toronto. The human brain, even when wrong, knows not to be wrong by too much. Think if your GPS... It is right most of the time. But sometimes when it is wrong, it takes you to the wrong place, or on a route that is way off target and you double the time. I see the same thing with Watson here.

If machines do indeed rise against the human (ok, being a little cheeky here), this will always be it's downfall. We just need to wait for the opportunity when it strikes against itself, or does something blatantly stupid.

From The New York Times:

ARTSBEAT BLOG: Robot Botches Geography on ‘Jeopardy!’

In its second outing on "Jeopardy!", IBM's Watson computer answered the first 24 questions correctly but got the last one wrong when it confused Toronto for a city in the United States.

http://nyti.ms/hb42R3

Posted via email from Suneil Mandava's Posterous