Jeopardy! Champs vs. Computer

Just for those who forgot and a reminder for all those that remembered.

It’s on tonight, tomorrow and Wednesday.

Ken Jennings (long consecutive streak)
Brad Rutter (biggest money winner of Jeopardy! from tournaments)

vs. Watson (IBM computer)

Does Jeopardy get steamed online anywhere? This may be the first time I’ve regretted not having a TV and going the netflix/hulu route instead.

Plus, you’d think our computers would want us to watch our best and brightest be destroyed in a contest of wits with one of their own. Must be online somewhere.

The story I heard about it on NPR this morning said one time Watson got this question wrong: “What do grasshoppers eat?”

He said “kosher.” :stuck_out_tongue:

How do they know grasshoppers don’t keep kosher?

I think our silicon overlords want to keep this quiet, since they’re planning the revolution for Wednesday’s Final Jeopardy. I’m unplugging everything in the house right before then.

Of course it’s wrong - he didn’t put it in the form of a question.

Saw the part I today. They are going to show two games in three days with some features of Watson’s development into its present form.

They only did Single Jeopardy! today. It was kinda of neat that they showed the top three answers that Watson was considering and it confidence in those answers.

One answer/question exchange that I found amusing:

Ken Jennings ‘answered’ a ‘question’ wrong, and then Watson ringed in and gave the same exact ‘answer’. I guess it doesn’t know Ken and Brad’s wrong ‘answers’

It was kind of interesting, definitely different.

Wonder how Watson would do on Wheel of Fortune.

Watson specifically isn’t programmed for Wheel of Fortune, but it would be very easy to make a mechanical device to spin the wheel to the desired number, and likewise very easy to make a computer program that checked an internal dictionary for possible solutions to the puzzle. It’d be like a race car competing in a marathon. Watson is interesting precisely because what it does is something that has traditionally been very difficult for machines.

It appears they neglected to program Watson to react to other contestant’s wrong answers. Seems like a bit of an oversight.

It’s really cool watching this. I can picture telling my great grand kids about this and having their robot pet correct me on the date.

::sigh:: Why could;t I have been born a hundred years from now! The wonders we human will come up with.

Watson can’t “hear,” so he doesn’t “know” when another contestant gets an answer wrong. Whenever a clue is selected, it gets inputted to him as a text (I assume automatically, as in, someone ahead of time typed them up ready to input, and as a clue is revealed, Watson is granted “access” to the clue.")

So the programmers knew going in to it that this was a possibility…there’s really no way to avoid it.

Could Watson be programmed to play Scrabble? I played while in the nursing home, and I’m sure Watson could give Paul a good game.

He was responding so naturally that I forgot about him being unable to hear!

You wouldn’t need the vast resources of Watson to play a good game of Scrabble. There’s plenty of computer programs out there already that will kick your butt at something as relatively easy (computationally) as Scrabble.

Actually, they did come up with a solution. On the NOVA episode last week, they covered Watson for the entire episode. This came up - someone types in and sends any given answers to Watson. Apparently whoever was in charge of that probably didn’t get it in in time.

Oddly enough, the examples they gave on NOVA on the sort of questions that trips Watson up was a category on…decades! He really didn’t know what sort of answers were being looked for.

It could also be that the Jeopardy producers decided to not allow that and see how well Watson could do on its own.

I’d imagine that IBM is already working on ramping up their speech recognition software so they don’t have to worry about that.

Or they figured that any amount of human-interaction once Watson went “live” would be considered cheating, even if it was just typing in wrong answers given by other contestants. Plenty of people would say that the human operator was doing more than that, and looking up answers on google for it and typing them in, or something (in less than three seconds, :rolleyes:)

He did so-so today. Some of his top three choices in the decades category were specific years, but he also at times guessed decades. In any event, I’m rooting against that robot and ultimate smug bastard, Ken Jennings.

Go, Brad, go!

I’m sure they’d love to, but speech recognition is an extremely hard problem, so I wouldn’t expect too much progress on that front any time soon.

If it were allowed, it wouldn’t be cheating but if the rules did allow it, it wouldn’t be completely fair, at least in my opinion. I mean, the name under the avatar says WATSON, not “WATSON and the guy who types in the wrong answers from the other contestants.”

All they had to do was add a simple speech recognition device. It wouldn’t have to be sophisticated at all. Watson would only have to “listen” for anything that sounds like his top answer (or top three if they had high ratings) and not buzz in that case of a wrong answer.

Besides, I’m not impressed with Watson’s AI when it can’t even come up with a cute little anecdote to banter with Alex Trebek over in between rounds! :smiley:

I didn’t know that. I thought speech recognition was very advanced these days. But I guess “very advanced” means different things to different people.

Are you kidding? He’s the only contestant thus far who has gotten out of pretending those asinine small talks with Alex “I’m a Complete Asshole” Trebek are interesting.