Quote:
Originally Posted by Chris is me
My post was worded pretty badly - the language recognition is a historical accomplishment, but the passing of a "Jeopardy test" really isn't...
|
Watching
this video (~20 minutes), I would certainly say it is. You can see that early on, Watson was a terrible, terrible Jeopardy contestant. If Deep Blue's victory over Kasparov was noteworthy, this certainly is too. Kasparov was at least a challenge to Deep Blue, and the Watson we saw play Jeopardy steamrolled Ken & Brad. I have no doubt they could win in a match structured to confuse Watson, but I think Watson would win the vast majority of normal matches.
I'm no computer science guy, but I think it should be pretty clear to all engineers that this one heck of a complicated system. We like to reply "It's more complicated than that" when people make suggestions about our systems they don't fully understand. Pretty sure this one is way over my head, and I'm willing to give IBM the benefit of the doubt. They do a pretty good job of explaining why Jeopardy makes such a good test case for interpreting language in their promo video too.