Live data from Hacker News

Day 1: IBM Watson Ties for first on Jeopardy

venturebeat.com

21–30 of 99 posts

Re: Day 1: IBM Watson Ties for first on Jeopardy

#21
post #16
post #8

"Watson can’t adjust its answers to what the other players say and so it simply answers with whatever comes up as its top answer." This is false, in fact they specifically mentioned Watson learning from the answers of the other players. In the game, Watson answers "The 1920s" after Jennings answers incorrectly with "the 20s". Basically, it didn't write off its own answer because it thought Jennings' was different eno…

I believe that it learns from the correct answers. i.e. once the question is complete, it can use the known correct answer in its analysis. But it can't "hear" anything.

That's correct. Watson only learns from correct answers after a question has been completed. This is mentioned specifically int he Nova documentary.

Re: Day 1: IBM Watson Ties for first on Jeopardy

#22
post #10

I couldn't have wished for a more exciting opening to this game. Watson just shot right off into the lead like there was no tomorrow, jetting ahead by thousands! I was blown away. Then he slowly started to crumble a bit and his answers were faulty more and more. But his strategy is amazing! He hops from category to category, while most humans pick a category and stick with it. I wonder if IBM has him do that purely b…

I've seen elsewhere it was suggested that whereas humans will want to get used to the question style of each category and learn from each one as they progress to harder answers, Watson has no such issues with jumping straight for high-value questions in all categories.

Re: Day 1: IBM Watson Ties for first on Jeopardy

#23
post #10

I couldn't have wished for a more exciting opening to this game. Watson just shot right off into the lead like there was no tomorrow, jetting ahead by thousands! I was blown away. Then he slowly started to crumble a bit and his answers were faulty more and more. But his strategy is amazing! He hops from category to category, while most humans pick a category and stick with it. I wonder if IBM has him do that purely b…

He hops from category to category, while most humans pick a category and stick with it. I noticed the same thing, but I really hoped it was unintentional. It doesn't feel like it's a fair fight otherwise, since it means he's exploiting a weakness intrinsic to his competitors. The whole point of this competition is to show that a computer can compete and win at a human level. If it weren't, then why force him to physi…

I wonder if the Watson developers did this because they were unable to simulate the benefit human players get by graduating through progressively difficult questions in a category.

Re: Day 1: IBM Watson Ties for first on Jeopardy

#24
post #21
post #16

Earlier quoted context omitted.

I believe that it learns from the correct answers. i.e. once the question is complete, it can use the known correct answer in its analysis. But it can't "hear" anything.

That's correct. Watson only learns from correct answers after a question has been completed. This is mentioned specifically int he Nova documentary.

Looks like this isn't too clear. An IBM spokesperson said he can't hear what other players are saying at all and only takes input from the question board. (http://latimesblogs.latimes.com/technology/2011/02/ibms-wats...)

Regardless, Watson only selectively takes input, if it does at all.

Re: Day 1: IBM Watson Ties for first on Jeopardy

#26

Going the Sci-Fi route, I would love this technology to expand and become an interactive Wikipedia. How cool would it be to go to your local library, (you remember those?) and ask it a series of questions and get answers. But thinking about how quickly archaic the library is turning into these days, I see it going this route: You call this service from your computer or phone and ask it anything. But by then Goggle wi…

I too am inspired by all the possibilities this implies for human-computer interaction. I picture a portable device that records events from your surroundings and provides personalized reasoning similar to this. It could remind you of people's names, it would know all the details of your personal finances and could provide budget and stock tips, remind you of events if you have a pattern of forgetting to plan ahead, etc. Aside from the fact that Watson fills up a whole room and its knowledge is tailored to Jeopardy, it really doesn't seem that far off combined with a few other technologies like machine vision and translation that are already out there.

Re: Day 1: IBM Watson Ties for first on Jeopardy

#27
post #11

Earlier quoted context omitted.

I suspect the hopping categories was to try and buy time to calculate what sort of questions would be asked in each. But it's first move of getting into the lead (and the daily double) I would say is a combination of good strategy and PSYOP.

I thought Watson was searching for the daily double spot to jump ahead $1000 right away.

And also to steal the humans their chance for a late comeback in the game.

Re: Day 1: IBM Watson Ties for first on Jeopardy

#28
post #10

I couldn't have wished for a more exciting opening to this game. Watson just shot right off into the lead like there was no tomorrow, jetting ahead by thousands! I was blown away. Then he slowly started to crumble a bit and his answers were faulty more and more. But his strategy is amazing! He hops from category to category, while most humans pick a category and stick with it. I wonder if IBM has him do that purely b…

> He hops from category to category, while most humans pick a category and stick with it.

Ken Jennings himself did that for the express reason of throwing off his opponents. As the category selector he would have an extra couple seconds to start thinking in that direction. Totally fair game for Watson to do too.

I agree that Watson started with a higher value clue in order to look for the Daily Double. Statistically the fourth row has the highest concentration of them. DDs are valuable not just because of the opportunity to double up or bet more than the clue's actual value, but also because you get exclusive access to the question without buzzer competition.

As for Watson's early lead, that happened because the clues chosen early happened to be the types of questions it could answer quickly and reliably. Computers are great at stuff like 4-letter words and searching Beatles song lyrics. Computers are less great at clues where the hint drives the answer, like the one about "don't name Voldemort" or the "Final" category where the hint is in the category name. It just so happened that those harder clues for Watson came after the easier clues.

Re: Day 1: IBM Watson Ties for first on Jeopardy

#29
post #13

Does anyone want to create a community site for the contest? I am looking for another Django developer who can participate in an all-night code sprint. The site would: 1. Show an online score board. 2. Track Watson's performance (like shkb is doing on Google Docs) 3. Aggregate pictures, videos, and written analysis of the event

I think you should've been doing this on Sunday...

Re: Day 1: IBM Watson Ties for first on Jeopardy

#30

Going the Sci-Fi route, I would love this technology to expand and become an interactive Wikipedia. How cool would it be to go to your local library, (you remember those?) and ask it a series of questions and get answers. But thinking about how quickly archaic the library is turning into these days, I see it going this route: You call this service from your computer or phone and ask it anything. But by then Goggle wi…

Have you seen Qwiki? http://www.qwiki.com/
Post reply on HN