Originally Posted by
Deepsepia
Very cool, much more impressive than playing chess well, which is just a calculation. Playing jeopardy involves knowledge, rhymes, jokes, allusions-- things that are "special" about human thinking.
And remember, what one computer can do can be implemented in other machines, fairly quickly. The world demand for computers that play great chess was never significant-- the demand for natural language understanding is massive.
The thing I'd like to know is to see inside it's algorithm-- how did it pick "Toronto" for the final jeopardy answer, for example? Thinking through that one, its that it could have been that Lester Pearson (for whom the airport is named) both served in WW II, and won the Nobel Peace Prize (although not for WWII service). So its correctly found a city with an airport that honors a hero who served in WWII. I'm not sure what Toronto's second biggest airport would be . . . and unlike the computer, a human could very quickly narrow down the field to cities with two major airports-- whose names you could be expected to know. New York would be first guess (Kennedy . . . but LaGuardia wouldn't work). Los Angeles? Has a bunch of airports, I think LAX is named after Tom Bradley. Atlanta? There's Hartsfield, and second airport is what? And then Chicago-- not many people would know who O'Hare is without looking him up, but "Midway" is famous.