Poll: Who do you think will win Jeopardy with IBM's Watson playing?

20 points by solipsist ↗ HN
IBM's Watson (http://www-03.ibm.com/innovation/us/watson/) - DeepQA designed to answer natural language questions

Brad Rutter (http://en.wikipedia.org/wiki/Brad_Rutter) - biggest all time money winner

Ken Jennings (http://en.wikipedia.org/wiki/Ken_Jennings) - longest winning streak (74 consecutive games)

In case it affects your decision, the current results after Day 1 are posted here: http://twitter.com/robotwisdom/status/37200070742904832

20 comments

[ 1.8 ms ] story [ 116 ms ] thread
After the first commercial break, I was too busy packing ammo for the upcoming War Against The Machines to pay much attention.
Watson only wins 60 percent of the time but the odds are 1:∞:15 (15 votes for Watson, 0 points for Brad Rutter, 1 point for Ken Jennings). =)
I'd say it's going to be between Rutter and Watson in the end, and between them it's probably a tossup. (Full Disclosure: I voted for Rutter) I just don't see Jennings winning, especially after today. I could be wrong but considering that Jennings has never beaten Rutter and that most of his money today came from buzzing in and only then trying to come up with an answer I don't see him being able to come out ahead. However, I do think that Jennings did the best job of adapting to the machine opponent's weaknesses, as opposed to a human's, by taking advantage of his ability to buzz in before he has a final answer and by starting with higher value question in the categories so who knows what will happen.

Bottom Line: I think Watson and Rutter is an excellent match-up and that the winner will be decided between them by luck as much as anything else. Jennings, with his experience and apparent strategy, should be able to win but since he has never beaten Rutter anytime they faced I don't see him being able to beat both Rutter and the Computer at once. I guess we'll see.

I think it's too early to tell based on day1 scores. In the preview round few weeks back, it was the opposite, Ken and Watson was tied before the last $1k question to put Watson in the lead and Brad was _way_ behind with only a couple hundred bucks.

"Jennings has never beaten Rutter" Hmm, they only matched up like once or twice, right?

Yes, AFAIK Jennings and Rutter have only ever faced off in the 'Tournament of Champions that Jeopardy has every so often, so probably 3 times at most (I'm almost positive it was more than once).

Also, I actually came up with my prediction over the weekend so the scores obviously didn't factor into it; however I do feel like I was heavily biased toward a human winning this challenge, or at least winning one of the games. The human players have enough game-play advantages (anticipation, buzzing before they have an answer, etc.) that either Brad or Ken will be able to stay very close to Watson until Final Jeopardy, at which its anybody's game.

Then again, as an optimistic Cub's fan, my track record for predictions is, well, less than stellar. :)

Note Rutter has never actually lost at Jeopardy (he has been behind on a daily total, but always pulled a win off in the end).
I was really impressed by Watson. His first choice was the daily double though, so I think he hacked it, haha. It seemed like for the answers it was very confident on it could react faster than Brad and Ken. It always was the first to buzz in, and jumped to a huge lead. I don't know if that is true, but its pure reaction time may be its winning factor.
> its pure reaction time may be its winning factor.

Indeed. Jennings himself has said that his buzzer instincts and reflexes created his winning streak more so than actual knowledge. You can easily see that when he clicks and then comes up with the answer a few seconds later. About half of all Jeopardy questions are known by all three players so it's a tossup as to who buzzes the fastest.

I'm not entirely clear how Watson buzzes. The general Jeopardy buzzer system is that an offstage operator presses a button to enable the buzzers as soon as Alex is done speaking. That button also turns on a light near the board to alert the contestants. If you press too early, there's a 300ms lockout until you can press again. We've been told that Watson uses a physical arm to press the Jeopardy buzzer, but we haven't been told exactly how Watson times it: does it listen for Alex to stop speaking, or use an optical sensor on the light, or some other method. An optical sensor would seem pretty hard for humans to beat on timing.

Watson will pull it out in Final Jeopardy: name the first billion digits of pi.
There's no way they would ask that question.
You're right. They would state the answer, and the contestant would ask the question. My bad.
Interesting factoid that might affect your voting... anyone know more about how Watson's confidence level factors into its betting?

Watson has an “Achilles heel,” because its betting is restricted to its confidence level, according to “Final Jeopardy,” a book about Watson by technology journalist Stephen Baker. Despite previous reports, Watson actually lost the practice match in January to Jennings in Final Jeopardy, because Jennings had bet aggressively on the Daily Doubles and Final Jeopardy, Baker said in an excerpt to final chapter of his book.

source: http://www.eweek.com/c/a/IT-Infrastructure/IBMs-Watson-Ties-...

Can someone give me the TL;DR; on how Watson actually answers questions?
tl;dr; with a voice synthesizer of some sort.
haha no I meant how does it come up with answers!
They store lots of data into Watson's memory, including wikipedia, imdb, the bible. They run hundreds of algorithms attacking a question from different angles, and produce a confidence level for each answer.
I think after todays performance it was interesting that watson had a hard time with the much harder questions (specifically the $800 dollar questions) but was able to answer the much easier questions. I have a suspicion that in the next round watson may have more problems answering questions but I still think that watson overall will win.