Poll
5 votes (25%) | |||
3 votes (15%) | |||
1 vote (5%) | |||
1 vote (5%) | |||
3 votes (15%) | |||
4 votes (20%) | |||
No votes (0%) | |||
1 vote (5%) | |||
2 votes (10%) |
20 members have voted
Ken Jennings won 74 games of Jeopardy, earning over $2,000,000. He placed 2nd in the Ultimate Jeopardy! Tournament, losing to Brad Rutter. He also won $500,000 on "Are You Smarter Than A Fifth Grader?", and won the championship title on Grand Slam in 2007.
Brad Rutter has never lost a match on Jeopardy!, and as a 5-time undefeated champion, has gained 3 Jeopardy! titles: 2001 Tournament of Champions winner, Million Dollar Masters Tournament winner, and Ultimate Tournament of Champions winner.
Watson is a supercomputer built by IBM. It will not have Internet connectivity during the match; Everything it knows is stored in its memory or storage. It will take the natural language clues Alex Trebek reads, and formulate a ranked list of possible responses. A mechanical piston will "buzz in" when Watson is ready to answer. It has won and lost against human opponents in live practice games (record unknown).
Survey question above: Who will win? Man or machine? Will it be a blowout, or a close match?
4 years of development by IBM for this. I think however that it has many real world applications.
Quote: boymimboIt really depends on how Watson handles the questions and figuring out how to interpret the questions.
The answers. Contestants ask the questions. That's jeopardy's gimmick, born, according to legend, out fo the 50's game show scandals.
Quote:4 years of development by IBM for this. I think however that it has many real world applications.
It should. The answers given on jeporady are often missleading and/or not clearly stated. A computer that can interpret those correctly might serve to take verbal orders from users. Ok, just now I've ahd about 4 hours of sleep and one long drive, minus a bad breakfast, so I can't think of any really useful aplication.
FWIW I'll be proving my ignorance by taking the online Jeopardy test tonight. 50 questions (yes, questions) and 15 seconds each. 35 correct is passing. I figure I'll get about 20 :)
By the way, as I told the Wizard at his coffee, I was on Jeopardy, so if anyone has any questions about the show, feel free to ask me.
I love those SNL Jeopardy skits with Will Ferrell, et al
He does like to get all the pronunciations right and projects a bit of smugness sometimes, but he does his job really well.
As of now, there are 10 votes, and it is split 50/50, human victory vs. computer victory.
Quote: teddysMy friend was on the Nova show last night for about five seconds. He was also on Jeopardy and won over $100,000.
By the way, as I told the Wizard at his coffee, I was on Jeopardy, so if anyone has any questions about the show, feel free to ask me.
When the first to buzz in answers incorrectly, how is the next to buzz in decided? Is it similar to the first buzz in, where the players must wait for Trebek to read the entire question before the judges make the buzzers, "live"?
After the first commercial break during the questions I was happy as I was the only one on here that voted for Watson to crush the humans. As the questions got tougher you could see Watson was having trouble with deciphering some of the reasoning. Overall I'm still impressed with his intelligence and think the next two days will be interesting.
Quote: OneAngryDwarfI thought so at first, but then I realized that what happened made sense. (SPOILER ALERT) Watson did very well with the easier questions early on, but started to falter a bit when they got to the higher-value ones. Since the $800 and $1000 clues are usually trickier and involve more wordplay and reasoning skills, it makes sense that a computer would have more trouble with them.
There were more things than just that which made me think that every second of it was planned and/or scripted:
1. Near the beginning it sounded like the whole thing was an elaborate advertisement for IBM's Power7 family of servers, as opposed to an interesting project by some creative and ambitious people who happen to work for IBM.
2. On Watson's first pick he conveniently chose the square that just happened to be the Daily Double.
3. They made it seem like he was going to wipe the floor with the humans, then in the last few questions he ends up tied with one of the humans... again, so convenient that it seemed like an intentional ploy to gain viewers and ratings.
I don't know exactly what I was expecting, but I guess it makes sense that it's all just an advertisement for IBM. As a programmer who understands the logical difficulties behind the code that runs Watson, I am probably looking at the situation differently than non-programmers.
Quote: JBNear the beginning it sounded like the whole thing was an elaborate advertisement for IBM's Power7 family of servers, as opposed to an interesting project by some creative and ambitious people who happen to work for IBM.
I don't know that that is surprising though. Getting a plug in for IBM.
Quote: JBThere were more things than just that which made me think that every second of it was planned and/or scripted:
2. On Watson's first pick he conveniently chose the square that just happened to be the Daily Double.
Very suspicious. After he chose the Daily Double, he started picking $200 questions. My wife and I both agreed that it seemed like something was up.
Quote: JB
2. On Watson's first pick he conveniently chose the square that just happened to be the Daily Double.
My wife and I did turn to each other and called 'bullshit' at the same time on that one. But it could just be chance, as Watson only had $400 at the time and it wouldn't make sense to choose the 'golden ticket' right away.
Quote: FinsRuleVery suspicious. After he chose the Daily Double, he started picking $200 questions. My wife and I both agreed that it seemed like something was up.
Watson was given past played games to analyze. I suppose a computer can arrive at some unusual strategies given any particular circumstances.
Quote: cellardoorMy wife and I did turn to each other and called 'bullshit' at the same time on that one. But it could just be chance, as Watson only had $400 at the time and it wouldn't make sense to choose the 'golden ticket' right away.
under full disclosure, I work for IBM, but I have nothing to do with Watson.
I don't think there is anything deceptive going on. I believe Watson would be programmed to try and take the daily double out of play as soon as possible, since the daily double could radically change the game at a later point. Given that, everyone knows the daily double is never found under the 200 answer, and is rarely found under the 400 answer. Also, if you are playing to minimize the damage your opponent can do, and that would be the strategy to take with these two champions, you would pick the 800 and 1000 questions first, so there would be a greater chance that the category link would not quite be understood yet.
It seems old hat now, but remember back when IBM's Deep Blue took on Kasparov? Everyone said a computer could not possibly take on a master grand champion, or whatever the top players are called. When Deep Blue went toe to toe with Kasparov the rumor was rampant that Kasparov was throwing the matches. It wasn't true back then, and I don't believe it is true that this contest is fixed either. As with most corporations, we have ethics fed to us all the time, and fixing this contest for good TV, or to promote IBM, would be unethical to the utmost. I think it is a fair contest.
Quote: RaleighCrapsAs with most corporations, we have ethics fed to us all the time, and fixing this contest for good TV, or to promote IBM, would be unethical to the utmost. I think it is a fair contest.
I meant to put it on my post but I forgot. I do think the game is being played fair but I thought it was funny when Watson picked the Daily Double first.
What better venue than Jeopardy to showcase this new technology? Make it fun!
Quote: zippyboy. I suppose that brings up the question of "How close are we to the Terminator movies?" Ten years? Twenty?
What better venue than Jeopardy to showcase this new technology? Make it fun!
I thought you were going to say: What better venue than Jeopardy to showcase Armegeddon, or the end of us all or something.
I guess in the scenario of all things odd, THE END OF MANKIND might as well start with a gameshow. While we're all watching space for the killer comet, it's Alex Trebeck who's ushering in our doomsday machine.
Quote: rxwineWhile we're all watching space for the killer comet, it's Alex Trebeck who's ushering in our doomsday machine.
He lived in an icy lair (Canada), had a simple yet effective disguise (mustache) and has a twin (Will Ferrell, SNL). Gadzooks, rxwine, you may be on to something!
*you guys are cracking me up with the Trebeck/Terminator stuff. I am in fact going to try to see if I can study Trebeck further, looking for android-like mannerisms.
*Having played computers for years at various things, as have many of us, one ugly thing always raises its head: the computer cheating. This is justifiable, as they are acknowledged to "cheat" as a matter of design in many cases; and then there was the claim Deep Blue was cheating, which I bought; I think IBM was at least not transparent about the programming if not "unethical". I have a blog post on all that. Maybe because of this known and alleged cheating, it was guaranteed to come up instantly here it seems. As it did in our "dice wars" phase.
PS: I finally decided the dice wars bots were not cheating, but we were reacting to being "fooled by randomness" as that guy put it in the title of his book.
Daily Doubles tend to appear in the $800 or $1000 spots, with $600 coming in third. There have been very rare times they've shown up in the $400 spot, but never to my knowledge at $200. He went to the first category on the board, perhaps by chance, and started at $800. Dumb luck brought up the Daily Double. Seems like a smart move to try to find it before anyone else.
Watson also jumped from category to category. A few human players have used this strategy to throw off their opponents, to make them have to change gears in their thought process more often. There is no rule that says you have to go $200 to $1000 straight down a category, but it makes clues go down smooth, from easy to hard all in the same topic.
I believe in the NOVA special that they said Watson will be getting the correct response fed to him AFTER it is given, so if he is confused about what kind of response to give in a category, he can look at what the other correct questions have been to determine what they are looking for. I find it on the unfair side that Watson is penalized for giving the same WRONG response someone else gave, because he can't hear them. Then again, I think I've seen human players do that inadvertently.
I also thought it was good strategy for Watson to continually change categories to try and keep the human opponents from getting into a groove in any given category format.
Judging by most blog sites, that should be fairly easy today.
On the contrary, the Turing test remains a very formidible challenge.Quote: pacomartinMost of the old guys here remember the "Turing test" proposed by Alan Turing 60 years ago. His test was simply could a computer fool the "subject person' in a blind test where you conduct a conversation. The subject has to guess if the responder is human or a computer.
Judging by most blog sites, that should be fairly easy today.
I think that given that IBM had a large team of paid people to build the computer, it's fair for them to get some free advertising out of the deal. Many people of course who didn't see the Nova program last week (which would be the great majority of viewers) would want to know about how Watson was programmed, that it's not connected to the internet, and how much data was stored in the system. After all, TV is not free.
Finally, certainly, the chinks in Watson's armor showed up, especially in the decades category when on several occasions, its top answer was a year and not a decade. I also loved the fact that it showed its top three questions.
I think Trebek was told to slow down as Watson is sent a text when the question is read and that it should have a fair amount of time to "compute" the answer.
Great drama.
On the point of the questions being read more slowly than usual. This could be. I hadn't noticed myself, but if it were true, this would be a bit of an advantage to Watson, although I don't think as much as you may think. Most of the great players ring in as soon as they are permitted, many times before they even know the answer for certain. They are relying on their knowledge of the category, and trust they can pull out the correct answer as soon as they finish reading the clue.
For those that don't know, I believe Wed night will be another complete Jeopardy game, played as normal, but the total from Game 1 will be added to the total of Game 2 to determine the overall winner.
SPOILER ALERT - Do not read the last sentence as it pertains to Final Jeopardy.............
I did find the Final Jeopardy for Game 1 to be a bit absurd. First is the wager that Watson makes, and then second is the answer given. Watson's answer wasn't even a US city! There is no way that answer should have even been in the possible list of correct answers. That part does seem to be a bit contrived.
Quote: RaleighCraps
I did find the Final Jeopardy for Game 1 to be a bit absurd. First is the wager that Watson makes, and then second is the answer given. Watson's answer wasn't even a US city! There is no way that answer should have even been in the possible list of correct answers. That part does seem to be a bit contrived.
I'm wondering if Watson is able to incorporate the 'rules' of the category properly in order to narrow down answers. I am not knowledgeable at all of the programming that does into it though. The mention of a US city was not mentioned in the answer, only in the category. In some other categories today, like "Church" and "State", Watson's preferred question had either "church" or "state" in them but the other two questions did not.
His Daily Double wager amounts were rather odd as well, I couldn't get the logic behind that.
Overall it looks like my original guess of Watson easily winning is looking good.
Watson's already clearly faster on the button than the humans. The next step would be next year when Watson 2012 can actually hear the answers by Trebek, and hear the wrong questions of other contestants. The producers need to come up with more answers that are more punny, innuendo, double-speak, mixing languages, etc.
The audience for this Jeopardy segment seems to be full of IBM programmers proud of their creation. I guess technology needs to move forward, but this seems like the beginning of the end to me.
And this is definitely not "thinking".
We were able after sometime able to recognize answers that Watson wouldn't get due to the complexity of the wordplay. Still, its performance last night was impressive.
Jennings looked so frustrated last night... I felt bad for him.
But is Watson revolutionary because it can understand a text message and come up with the answer and measure how certain it is of the answer. I'm not sure.
Why on earth would they allow humans to get into a contest on how fast you can click in? Of course the machine will be faster! that is really bothering me.
I am not finding these answers easily online.
Quote: boymimboWhat really surprised me last night was Watson's inability to get the city name last night. I knew right away that O'Hare was a WW2 hero and that the WWII battle was Midway. It couldn't connect the two and came up with a non-US city.
In the early days of web browsing, searches for very specific terms, say 6 liter V8 engine, returned some results that were far, far, far off the mark. Altavista (remember it?) was notorious in this respect.
Today results more nearly reflect what is being looked for, but sometimes you enter "3:2 BJ in Las Vegas" and get a result about prostitution in Nevada.
Quote: odiousgambitIs someone entering the question as text for watson?
I think so.
Quote:Why on earth would they allow humans to get into a contest on how fast you can click in? Of course the machine will be faster! that is really bothering me.
It's not how fast you can click, it's how fast you can figure out the answer. Like it was said above, human contestants often click before they know the answer, Watson cannot do that. It has to complete the analysis before clicking. I have seen it happen several times that it did have the right answer, but one of the other contestants was faster to click.
Quote: weaselmanIt's not how fast you can click, it's how fast you can figure out the answer. Like it was said above, human contestants often click before they know the answer, Watson cannot do that.
That's why wrong questions earn a deduction in the score. Otherwise who clicks in faster would have a huge advantage. In fact, not clicking without a question would help preserve the bankroll as it were.
Quote: NareedThat's why wrong questions earn a deduction in the score. Otherwise who clicks in faster would have a huge advantage. In fact, not clicking without a question would help preserve the bankroll as it were.
Sometimes it makes sense to click before you know though. For example, remember the clue about Voldermot? I knew I will know the question as soon as I heard the "Mad Eye Moody" line (I am a big Harry Potter fan), well before I had time to take in and understand what exactly the question was - the first thing, that popped in my mind was Harry Potter, but a couple of seconds later, when I finally took in the clue, I knew it was about Voldermot, but by then it was all over.
Quote: weaselmanIt's not how fast you can click, it's how fast you can figure out the answer. Like it was said above, human contestants often click before they know the answer, Watson cannot do that. It has to complete the analysis before clicking.
This does not appear to be exactly correct. If I have it right, all the contestants have to wait until Trebeck finishes talking, and evidently there is a penalty for clicking before you are supposed to [a time elapses before you can click again]. Perhaps only for this match? that I don't know.
Quote: weaselmanI have seen it happen several times that it did have the right answer, but one of the other contestants was faster to click.
I saw that too, but it usually went the other way the night I saw it.
Quote: NareedThat's why wrong questions earn a deduction in the score. Otherwise who clicks in faster would have a huge advantage. In fact, not clicking without a question would help preserve the bankroll as it were.
Jennings says clicking in faster is in fact just what you want to do. Below from Wikipedia:
[from an earlier trial]Quote:Originally Watson buzzed in electronically, but Jeopardy! requested that it physically press a button, as the human contestants would. Even with a robotic "finger" pressing the buzzer, Watson remained faster than its human competitors. Jennings noted, "If you're trying to win on the show, the buzzer is all," and that Watson "can knock out a microsecond-precise buzz every single time with little or no variation. Human reflexes can't compete with computer circuits in this regard." Also, Watson could avoid the time-penalty for accidentally buzzing in too early, because it was electronically notified when to buzz, while humans had to anticipate the right moment
PS: wikipedia answers another question I had
Quote:Watson is "deaf" and doesn't utilize speech recognition
Quote: weaselmanSometimes it makes sense to click before you know though. For example, remember the clue about Voldermot?
I haven't seen the show in months :) It's on when I'm at work.
When I do watch I don't play along with a clicker. I just say the questions aloud if I know them or can think of them. Usuaully I read faster than Trebek, so I know the right question before I could click. Sometimes I need to think it over. Sometimes I don't even understand what response is wanted.
And I'd likely get any Harry Potter questions wrong. I heard there's some book based on the movie... :P
It takes Watson 4 milliseconds to press the button after the question is read. It cannot lock itself out. If you press the button before the question is read, you are locked out. The 2 contestents last night locked themselves out 9 times. He also alluded that the results tonight will be substantially different than last night.
So, essentially, if the contestents and Watson know the answer, the contestent has a 4 millisecond length of time to respond before Watson will ring in, which is just about impossible. It did happen a couple of times last night however.
So, the only way that the contestents get the question is that Watson doesn't get it or gets it wrong. If I were the contestents, I would just let Watson ring in and let it try.
Quote: odiousgambit
This does not appear to be exactly correct. If I have it right, all the contestants have to wait until Trebeck finishes talking, and evidently there is a penalty for clicking before you are supposed to [a time elapses before you can click again]. Perhaps only for this match? that I don't know.
Yes, I knew that, sorry if I made it sound differently somehow. The penalty for clicking before the clue is read is not being able to click again for a quarter of a second AFAIK.
Quote:I saw that too, but it usually went the other way the night I saw it.
Yes, the second night was brutal. But on the first game, there were many more cases like that - Watson had the answer but was "outclicked" by one of the other contestants.