• This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn more.

Watson kicking ass in Jeopardy.

Edrondol

Well-known member
#1
The way they have it set up is actually kinda biased against the Humans. Watson is being fed the answers at the same time as the Humans, but Alex has to read the clue. THEN they get to click in. So Watson can extrapolate when to click faster than the Humans.
 

Quillz

Well-known member
#2
I don't think the timing really matters. Pretty sure the supercomputer will win regardless.

It's like when they have supercomputers take on the greatest chess players in the world and the computers win almost every time.
 

Chimpie

Well-known member
#3
The contestants can read the question as soon as it appears. I read once that some contestants 'block out' Alex and read the questions themselves so they can start thinking of the answer quicker.
 

SneakyDave

Well-known member
#4
Is Watson really trying to figure out the answers, or is he just doing some fancy internal googling to get them? I think I got disenchanted with the whole Deep Blue / AI saga when I heard that part of its "intelligence" was just recording or memorizing all the moves Kasparov had ever done given a scenerio. I don't think that's real intelligence, that's just an encyclopedia.
 

Trombones13

Well-known member
#7
It's only been through the first round (Double Jeopardy is tonight), but Watson and Brad are tied at $5k while Ken has $2k, if I remember correctly.
 

Edrondol

Well-known member
#8
When I first wrote the post Watson was at about $3K while the other two were both hardly on the board.

What I think is interesting is that Watson's first choice was the first column $800 and was the Daily Double. From there it went straight across the rows -> $200, $200, $200, $400...So how did it go right to that area? It'll be interesting to see how it chooses things today.

And I thought it was going to be 2 full games not one game stretched over 2 days.
 

SneakyDave

Well-known member
#9
I heard that Watson had a gaffe in which he stated a "question" for an "answer" that a competitor already had given wrong. So they didn't build Watson to "hear" answers from competitors, and find out if those answers were wrong or not?
 

Edrondol

Well-known member
#10
I heard that Watson had a gaffe in which he stated a "question" for an "answer" that a competitor already had given wrong. So they didn't build Watson to "hear" answers from competitors, and find out if those answers were wrong or not?
Nope. Watson receives the text of the question at the time that it's revealed to the other contestants.
 

SneakyDave

Well-known member
#11
No, I meant that I read that a contestant answered a question incorrectly, then Watson "buzzed in", and gave the same incorrect answer.

Here's the story I was reading which I found interesting. I guess they didn't think to include some kind of voice recognition of contestant answers to make sure Watson wouldn't give the same one?
http://www.pcmag.com/article2/0,2817,2380351,00.asp

Watson's most obvious gaffe was its inability to recognize when an answer had already been given. One of two of the all-time Jeopardy! champions pitted against the super-machine, Ken Jennings, answered a question incorrectly with, "What is 1920s." Watson repeated the same answer. Steve Camepa, IBM's general manager of global media and the entertainment industry explained that "Watson only takes his input from the question board so the fact that somebody else gave the same answer doesn't factor into what Watson says. He can't hear what the other players are saying, but maybe that's a feature we can add in the future."
 

Trombones13

Well-known member
#12
And I thought it was going to be 2 full games not one game stretched over 2 days.
Oh, yeah; I think you're right (except that it's over three days, not two).

No, I meant that I read that a contestant answered a question incorrectly, then Watson "buzzed in", and gave the same incorrect answer.

Here's the story I was reading which I found interesting. I guess they didn't think to include some kind of voice recognition of contestant answers to make sure Watson wouldn't give the same one?
http://www.pcmag.com/article2/0,2817,2380351,00.asp
Yep; I think Ken said "the 20's" and then Watson said "1920s." Alex said, "No...Ken just said that." haha
 

Fred Sherman

Well-known member
#13
Interesting observation. I wonder why the program team didn't write Watson to hear incorrect answers and apply that to the decision making heuristics as an exclusion data set?
 

Salty

Active member
#15
There were a lot more problems than just hearing. Some of the more ambiguous questions, at the end, Watson's first and second possible answers were incorrect.

Point being human contestants were able to parse the words in the question and understand the perceived answer while Watson was incapable of understanding the question at hand. As shown by numerous answers in which the first two were totally incorrect. Which happened often.

There was a Harry Potter question in which Watson had absolutely no clue what the answer was.
 

Edrondol

Well-known member
#19
How would we know for certain it doesn't? If it can profess love then who are we to say? As our brains are basically organic computers (in the most simplistic of explanations) how can we say for certain that a machine can't, won't or HASN'T gained some measure of sentience? Hell, most PEOPLE I know don't know if they are in love or what they are feeling at a specific time. Feelings are non-quantifiable and can therefore be claimed by a machine as well.

In the end, does it matter if a machine is not sentient but acts that way due t programming? If it reacts to its environment and acts accordingly then we'd never know for certain as to the voracity of its claims of acknowledgment of itself as an aware entity.