Presumably it was looking at the math and realizing the best chance the humans had to catch up was the daily doubles so was actively looking for them based on statistical records. Which means taking a lot of 1200/1600 dollar clues as they're the most likely to have daily doubles.
enlightenedbum on
Self-righteousness is incompatible with coalition building.
I wanted him to sound like Darrell Hammon doing Sean Connery on SNL's Celebrity Jeopardy instead. "That'sh not what your mother shaid lasht night, Trebek!"
At the end of yesterday's show it appeared that the contestants had watsons number, you could see that they were ringing in immediately the light went on and then thinking while Alex said their name. I bet they were told not to do that today, since Watson obliterated them.
I wouldn't be surprised if they were told to back off. Watching Jennings and Rutter answer is totally not the point of this whole thing. We already know they can answer Jeopardy questions by deciphering plain human language. We want to see if Watson can too, and obviously the answer is yes.
What's disappointing is how simple all the clues have been. Not just in terms of content (Jeopardy's gotten progressively easier over the last decade), but in terms of format. There's been little to none of the clues based around puns, homonyms, before & after (eg, Tom Cruise Missile), etc., - the exact types of "natural language" answers that were supposed to really test Watson.
Not that it really matters for tech demo purposes. No one who actually needs Watson's help is going to be asking for it with puns.
It is certainly true, none of the questions have been of the sort where I don't even know what sort of thing they are asking until I hear the answer which is very common on Jeopardy on a normal day. I'd like to see them throw out something really hard though. Tom Cruise Missile would be a perfect example, something like a 'Star Weaponry' category. Where the answer itself and the question make no sense considered in isolation.
That would be Before & After, which Ken says he's seen Watson sweep.
I still think that 'irrational' answers have to be his weak point, since they don't make sense without a complete understanding of clue and answer and category. The questions we've seen so far have all been solvable with a vaguely logic understanding of language.
To everyone saying Jeopardy/IBM told contestants to "ease off"
FUNNY BUSINESS?
Q. Given that it is in IBM's (marketing) interest for Watson to win, what sort of oversight (if any) exists to make sure that Watson isn't getting the questions provided to him ahead of time or any funny business isn't taking place?
A. KEN JENNINGS :
Not rigging a quiz show in America isn't just a good idea--IT'S THE LAW. Since the scandals of the 50s, it's actually been made a felony to screw around with the outcome of a game show. There was VERY careful oversight (an independent auditor) on every aspect of gameplay here. Plus, from IBM's point of view, this is a research experiment. (Well, and a PR bonanza.) They don't want their data tainted either.
Yeah, I think it's mostly Ken's point about computers being able to work faster than us on the whole buzzer thing. I mean, these aren't even tournament of champions level questions for the most part (though last night's categories were not favorable for me) so with Ken and Brad it's just a matter of buzzing.
Which makes this an interesting tech demo, but not great TV.
enlightenedbum on
Self-righteousness is incompatible with coalition building.
That would be Before & After, which Ken says he's seen Watson sweep.
He also says IBM did a lot of work specifically targeted at wordplay categories. Which makes it all the more disappointing that we're not seeing Watson actually get to try them.
Even if they tweaked the hell out of this thing, I doubt it could do as well on a You Don't Know Jack-styled show as it does on Jeopardy. The questions in Jeopardy are all phrased in a very specific way, without pop-culture obfuscation added to confuse the query.
(1) The term "contest'' means any contest broadcast by a radio station in connection with which any money or any other thing of value is offered as a prize or prizes to be paid or presented by the program sponsor or by any other person or persons, as announced in the course of the broadcast.
So I know it's a very small part of this, but Watson can hear and understand anybody talking to him? Like, when they were showing the test Jeopardy sparrings, and the not-Trebek host told him it was wrong, does Watson input that or is it fed to him digitally?
E: Also, I want to see Watson guest on other game shows. Specifically, Wheel of Fortune, Hollywood Squares, and the Dating Game.
MetroidZoid on
Steam
3DS FC: 4699-5714-8940 Playing Pokemon, add me! Ho, SATAN!
Yeah, I think it's mostly Ken's point about computers being able to work faster than us on the whole buzzer thing. I mean, these aren't even tournament of champions level questions for the most part (though last night's categories were not favorable for me) so with Ken and Brad it's just a matter of buzzing.
Which makes this an interesting tech demo, but not great TV.
Yeah, as interesting as it is from a Com Sci perspective, the game itself is pretty boring to watch.
It was good the first day by the end, when it looked like it would tighten up, but by the second day it was just "Watson always buzzes in first and gets the right answer 99% of the time".
(1) The term "contest'' means any contest broadcast by a radio station in connection with which any money or any other thing of value is offered as a prize or prizes to be paid or presented by the program sponsor or by any other person or persons, as announced in the course of the broadcast.
Even if they tweaked the hell out of this thing, I doubt it could do as well on a You Don't Know Jack-styled show as it does on Jeopardy. The questions in Jeopardy are all phrased in a very specific way, without pop-culture obfuscation added to confuse the query.
That could make for an interesting Game With Fame when YDKJ comes out.
Yeah, I think it's mostly Ken's point about computers being able to work faster than us on the whole buzzer thing. I mean, these aren't even tournament of champions level questions for the most part (though last night's categories were not favorable for me) so with Ken and Brad it's just a matter of buzzing.
Which makes this an interesting tech demo, but not great TV.
Yeah, as interesting as it is from a Com Sci perspective, the game itself is pretty boring to watch.
It was good the first day by the end, when it looked like it would tighten up, but by the second day it was just "Watson always buzzes in first and gets the right answer 99% of the time".
Indeed, I think what we've shown here is that all three contestants are equally quick to think of the answer, probably equally likely to be right and that Watson is faster at pushing a buzzer.
Now, the fact that Watson is so great at jeopardy is amazing, but it's not very interesting.
1: Even though he won in the end, Watson really fumbled today. A metric ton of questions he didn't know, and wasn't even close on in his calculations. Still, bloody amazing either way, and hey, nobody knows everything. :P
2. Did anybody else get a glimpse of Ken and Brad when Watson got the second DD? You could just see the "Aw fuck move they did.
3. Ken Jennings is a God.
The Wolfman on
"The sausage of Green Earth explodes with flavor like the cannon of culinary delight."
1: Even though he won in the end, Watson really fumbled today. A metric ton of questions he didn't know, and wasn't even close on in his calculations. Still, bloody amazing either way, and hey, nobody knows everything. :P
2. Did anybody else get a glimpse of Ken and Brad when Watson got the second DD? You could just see the "Aw fuck move they did.
It was super fucking impressive, but contrary to IBM's suggestion that one of the key areas where this sort of open question technology would be useful is in medicine, I'm still pretty sure I don't want my healthcare managed by a set of algorithms that came up with "Toronto" as an answer in the category "U.S. Cities."
Although within another few years of actually refining the algorithms and also tailoring a system to a specific application like medicine, I can possibly see myself changing my mind.
It's not like they were saying it's ready for that use right this instant. Like you said, give it a couple/few more years for development.
Right.
Although based on the Legal E's category, I would totally expect it to ace the MBE portion of the Virginia State Bar Exam if it took it tomorrow. Eminent Domain? Alright, that one's easy. Escalator clauses? Holy shit.
Edit: Oh God. Think of the tens of thousands of unemployed paralegals who wouldn't be able to find work in Manhattan. Times Square would be like Tahrir Square, only instead of shoes they'll be flinging soy lattes.
Posts
It is certainly true, none of the questions have been of the sort where I don't even know what sort of thing they are asking until I hear the answer which is very common on Jeopardy on a normal day. I'd like to see them throw out something really hard though. Tom Cruise Missile would be a perfect example, something like a 'Star Weaponry' category. Where the answer itself and the question make no sense considered in isolation.
I still think that 'irrational' answers have to be his weak point, since they don't make sense without a complete understanding of clue and answer and category. The questions we've seen so far have all been solvable with a vaguely logic understanding of language.
On the other hand, I love that he's predisposed to bet like a dick and win by a buck.
Which makes this an interesting tech demo, but not great TV.
He also says IBM did a lot of work specifically targeted at wordplay categories. Which makes it all the more disappointing that we're not seeing Watson actually get to try them.
Does this even qualify as a real quiz show? Is Watson going to get the money/prizes if he wins?
http://law.justia.com/us/codes/title47/47usc509.html
I'd say so.
E: Also, I want to see Watson guest on other game shows. Specifically, Wheel of Fortune, Hollywood Squares, and the Dating Game.
3DS FC: 4699-5714-8940 Playing Pokemon, add me! Ho, SATAN!
Yeah, as interesting as it is from a Com Sci perspective, the game itself is pretty boring to watch.
It was good the first day by the end, when it looked like it would tighten up, but by the second day it was just "Watson always buzzes in first and gets the right answer 99% of the time".
Half of the humans' winnings go to charity while all of Watson's does
That could make for an interesting Game With Fame when YDKJ comes out.
In the NOVA special, the lead IBM guy actually said that that's his ultimate goal.
Indeed, I think what we've shown here is that all three contestants are equally quick to think of the answer, probably equally likely to be right and that Watson is faster at pushing a buzzer.
Now, the fact that Watson is so great at jeopardy is amazing, but it's not very interesting.
thats the kind of false pattern humans will fall for but not computers
I love Ken Jennings
You calling me fallible?
Those are fighting words.
Fisticuffs at noon.
Ok maybe with gloves.
Ok never mind.
For the record, they will do this for any blind/deaf contestant.
Cables at dawn, good sir
3DS FC: 4699-5714-8940 Playing Pokemon, add me! Ho, SATAN!
2. Did anybody else get a glimpse of Ken and Brad when Watson got the second DD? You could just see the "Aw fuck move they did.
3. Ken Jennings is a God.
Ken was really smashing that button today.
I never asked for this!
Although within another few years of actually refining the algorithms and also tailoring a system to a specific application like medicine, I can possibly see myself changing my mind.
Right.
Although based on the Legal E's category, I would totally expect it to ace the MBE portion of the Virginia State Bar Exam if it took it tomorrow. Eminent Domain? Alright, that one's easy. Escalator clauses? Holy shit.
Edit: Oh God. Think of the tens of thousands of unemployed paralegals who wouldn't be able to find work in Manhattan. Times Square would be like Tahrir Square, only instead of shoes they'll be flinging soy lattes.
http://www.youtube.com/watch?v=o6oS64Bpx0g
Saw him scribbling in Final Jeopardy.
Never guessed what he was pulling.
Although given how the other guy was sucking up with his charity, I figure Ken knew he had to hedge his bets.
Why I fear the ocean.