Watson @ Jeopardy Feb 14-16

**Julian** · 2011-02-15 12:31

Originally Posted by Stiker

I highly doubt it was an oversight, but more of a "this isn't worth it" sort of thing. Happens in programming all the time. Odds of that exact situation coming up were rare, so it got bypassed. The assumption was that 99% of the time, the answer that Watson had as the most correct wouldn't be both wrong and said by another contestant.

Watson was being fed questions via text files, so you would have to add yet another delivery system to not only receive/parse the answers, but then also create additional logic to handle the other answers, and then calculate if Watson should click his buzzer based on the new information. And adding that on could also affect the speed at which Watson could respond with a second answer.

It's a calculated risk, and honestly, it made Watson seem a lot more human/genuine by "screwing up". Any thoughts of "well, Watson is being fed answers" died right there.

This, seriously. How often is Watson going to pick a wrong answer, and then have someone else ring in first, and say the exact same answer that Watson was thinking?

I imagine it wouldn't be very high, and I assume the people working on it thought along the same lines. Adding something that would fix this "small oversight" would add a lot of extra complexity onto the system, that isn't really needed at all. Also, for this scenario to happen, Watson has to pick a wrong answer. Working on making it better to further and further reduce this number would be a better solution to the problem.

**Surealistic** · 2011-02-15 12:43

look, ibm said "we want to take our computer on jeopardy and show how well he can do"

i dont care how hard it is to program something, or if its a "rare situation", if you are putting yourself out there with something of this complexity, you make sure its right, you dont overlook anything

you can sit there and defend ibm all day long about how hard it is, but guess what, they have an unfinished product on stage (and yes it is doing incredibly well), but after watching jeopardy a few times in my life, i know that people will ring in and give wrong answers and you can see that someone else was gonna say the same thing, and having to have someone enter all the questions in by text, really? they couldnt get voice recognition software? there is some pretty good stuff out there that im sure would do the job just fine, and actually put watson on more of a level playing ground of actually having to hear and recognize whats being said and how, making it better understand puns and whatnot

wait another 6 months, finish watson, then do this, right now, its close, but its not done

**tyven** · 2011-02-15 12:53

YO KEN GONNA FUCKING RAPE TONIGHT

**Darus Grey** · 2011-02-15 12:53

ITT: People pretending they understand computational linguistics.

My best professional guess at why they don't have watson analyzing(you analyze voice, you parse text) voices is because they know that the best method of doing that is to phonetacize the corpus. In Comp.Ling. you always get the best results comparing sound to sound, and text to text, much in the same way we humans understand natural language. Phonetacizing the corpus is probably more expensive then the entire watson project to-date however, so it's out(for the moment).

**Saga** · 2011-02-15 12:53

Originally Posted by Surealistic

they couldnt get voice recognition software? there is some pretty good stuff out there that im sure would do the job just fine

stop, stop! you're being too scientific! "stuff"?? "fine"???? our brains can't process all these expert-level terms you're arbitrarily throwing around because you think you have an idea of what you're talking about! please explain!!

**Knightmare** · 2011-02-15 12:55

Originally Posted by Surealistic

look, ibm said "we want to take our computer on jeopardy and show how well he can do"

i dont care how hard it is to program something, or if its a "rare situation", if you are putting yourself out there with something of this complexity, you make sure its right, you dont overlook anything

you can sit there and defend ibm all day long about how hard it is, but guess what, they have an unfinished product on stage (and yes it is doing incredibly well), but after watching jeopardy a few times in my life, i know that people will ring in and give wrong answers and you can see that someone else was gonna say the same thing, and having to have someone enter all the questions in by text, really? they couldnt get voice recognition software? there is some pretty good stuff out there that im sure would do the job just fine, and actually put watson on more of a level playing ground of actually having to hear and recognize whats being said and how, making it better understand puns and whatnot

wait another 6 months, finish watson, then do this, right now, its close, but its not done

exactly, how will IBM takeover the world if they mess up on jeopardy

**Skidz** · 2011-02-15 12:58

I wasn't saying I'm not impressed by Watson, just that I'm amazed that situation did not come up in the test sessions (or maybe Stiker's right, and they just said "not worth it"). I just assume that since it was "optimized" for Jeopardy!, they would take all possible scenarios into account.

Also, regarding the question-feed time, it would have to be delayed a bit to keep the contest fair, no? Or maybe I overestimate the computing power of Watson.

**Karshe** · 2011-02-15 13:04

One thing I thought was very strange. Generally, when someone has control of the board, contestants will go vertically down a column and then move onto the next subject. Watson was going horizontally across all questions as much as it could... it would do all the $200s, and then all the $400s, etc.

But here's what I thought was strange... the very first time it got control of the board, it randomly picked an $800 question in an untouched category and got the Daily Double. Wtf? If it had been picking random dollar amounts for the whole round it wouldn't have seemed as weird.

**Dantrag** · 2011-02-15 13:17

Lol when I saw it pick the daily double After what? One question I kinda went wtf.

**Norellicus** · 2011-02-15 13:24

Originally Posted by Surealistic

look, ibm said "we want to take our computer on jeopardy and show how well he can do"

i dont care how hard it is to program something, or if its a "rare situation", if you are putting yourself out there with something of this complexity, you make sure its right, you dont overlook anything

you can sit there and defend ibm all day long about how hard it is, but guess what, they have an unfinished product on stage (and yes it is doing incredibly well), but after watching jeopardy a few times in my life, i know that people will ring in and give wrong answers and you can see that someone else was gonna say the same thing, and having to have someone enter all the questions in by text, really? they couldnt get voice recognition software? there is some pretty good stuff out there that im sure would do the job just fine, and actually put watson on more of a level playing ground of actually having to hear and recognize whats being said and how, making it better understand puns and whatnot

wait another 6 months, finish watson, then do this, right now, its close, but its not done

If you think they put Watson on this show because they thought it was "done", you're a moron.

Everything it does, even on national TV, is research. This whole stunt is proof of concept. You can GUARANTEE they took something useful away from the show to apply to the project going forward.

**Qalbert** · 2011-02-15 13:54

Originally Posted by Dantrag

Lol when I saw it pick the daily double After what? One question I kinda went wtf.

I'm wondering if it analyzed past games for which squares had high probabilities for being the double or something

**Conrad** · 2011-02-15 15:04

It's always at least the third item of any given category, I believe. Can't recall it ever being in the first two.

**Stiker** · 2011-02-15 15:13

Originally Posted by Surealistic

look, ibm said "we want to take our computer on jeopardy and show how well he can do"

i dont care how hard it is to program something, or if its a "rare situation", if you are putting yourself out there with something of this complexity, you make sure its right, you dont overlook anything

you can sit there and defend ibm all day long about how hard it is, but guess what, they have an unfinished product on stage (and yes it is doing incredibly well), but after watching jeopardy a few times in my life, i know that people will ring in and give wrong answers and you can see that someone else was gonna say the same thing, and having to have someone enter all the questions in by text, really? they couldnt get voice recognition software? there is some pretty good stuff out there that im sure would do the job just fine, and actually put watson on more of a level playing ground of actually having to hear and recognize whats being said and how, making it better understand puns and whatnot

wait another 6 months, finish watson, then do this, right now, its close, but its not done

The bold is the reason Watson is currently on stage. They believe it's at the point that it can win a game of Jeopardy. This is a test to see if they are right. And because it severely "botched" one question, it's a failure? Ok, sure.

PS: I've watched humans do the same shit Watson did before.

**Qalbert** · 2011-02-15 15:29

No one called it a failure, dumbass.

**Stiker** · 2011-02-15 16:03

Right, calling it an unfinished product and saying it needs another six months of development is calling it an unbelievable success. Hell, the post I quoted explicitly calls out for not "mak[ing] sure its right" as well as saying it should have included voice recognition over text input. Furthermore, said post says that Watson isn't even on a level playing field because of lack of said voice recognition.

Sure looks like someone considers it a failure.

Dumbass.

**Headspace** · 2011-02-15 16:04

sure is some nitpicking in this bitch

**Qalbert** · 2011-02-15 16:07

Originally Posted by Stiker

Right, calling it an unfinished product and saying it needs another six months of development is calling it an unbelievable success. Hell, the post I quoted explicitly calls out for not "mak[ing] sure its right" as well as saying it should have included voice recognition over text input. Furthermore, said post says that Watson isn't even on a level playing field because of lack of said voice recognition.

Sure looks like someone considers it a failure.

Dumbass.

Pointing out a flaw != calling it a failure. Stop whiteknighting a fucking computer.

**Surealistic** · 2011-02-15 16:13

if i was calling it a "failure" i would have said so, unfinished product =/= failure

reading comprehension ftw

**Stiker** · 2011-02-15 16:17

Originally Posted by Qalbert

Pointing out a flaw != calling it a failure. Stop whiteknighting a fucking computer.

Re-read the post I quoted, it attacked some of IBM's direct goals of Watson, specifically, the ability of Watson to comprehend complex linguistics such as puns. That's a little bit more than pointing out a flaw of repeating an answer.

Surealistic wasn't just pointing out a flaw, but saying that what he believed caused it (lack of audio input), was pointing to a larger failure of the system.

But hey, keep speaking for him. Arguing via proxy is amusing.

**Stiker** · 2011-02-15 16:19

Originally Posted by Surealistic

if i was calling it a "failure" i would have said so, unfinished product =/= failure

reading comprehension ftw

As I said before, you attacked on of Waston's direct goals, is it really that shocking that I drew that conclusion?

Thread: Watson @ Jeopardy Feb 14-16

Thread Tools

Latest Threads

Up & Coming Threads

Hottest Threads