After years of planning, IBM's learning, human-aware computer Watson was put to a competition like no other - a match of Jeopardy against quiz show heavyweights Ken Jennings and Brad Rutter. The result - Watson won. Barely.
The match, which Watson has been training for since 2009, was officially announced last year. At the end of last week, the multi-episode feature where Watson faces off against Jennings and Rutter was filmed.
But right before that, all three competed in a trial run at IBM's headquarters in New York State. The trial lasted as long as a normal game of Jeopardy would before its first commercial break - in other words, about enough time for the contestants to get through half of a round.
Right before the last clue of the round, Jennings and Watson were tied at $3,400. However, Watson chimed in to answer the final question and correctly identified the children's book Harold and the Purple Crayon. That set him ahead to $4,400. Rutter trailed at $1,200.
The full-length Jeopardy matches have been filmed, but no one is allowed to discuss the results. They'll be aired on TV next month, and at that time we'll really know who wins in the battle of man versus machine.
Should be interesting. I'm sure everyone is aware of the Deep Blue Project (also by IBM) which beat chess pro Kasparov decades ago.
IBM and the producers of quiz show Jeopardy announced Tuesday that an IBM computer known as "Watson" will compete against two of the show's most successful contestants in February 2011.
Watson, named after IBM founder Thomas J. Watson, will go up against Ken Jennings and Brad Rutter (left) on February 14, 15, and 16 in two matches over three days. Jennings won 74 games in a row during the 2004-2005 season, taking home more than $2.5 million. Rutter is Jeopardy's highest-earning player, winning more than $3.25 million during several appearances in 2002 and 2005.
The grand prize for the Watson-Jennings-Rutter matchup will be $1 million, with second place winnings of $300,000 and a $200,000 third prize. Jennings and Rutter will donate 50 percent of their winnings to charity, while IBM will donate 100 percent of Watson's cash.
Getting Watson to the Jeopardy stage has taken several years. Many clues in Jeopardy rely on subtle word play, irony, and riddles, something at which humans excel but that computers have difficulty understanding. Essentially, IBM had to figure out how to get Watson to think.
"After four years, our scientific team believes that Watson is ready for this challenge based on its ability to rapidly comprehend what the Jeopardy clue is asking, analyze the information it has access to, come up with precise answers, and develop an accurate confidence in its response," Dr. David Ferrucci, head of the Watson research team, said in a statement. "Beyond our excitement for the match itself, our team is very motivated by the possibilities that Watson's breakthrough computing capabilities hold for building a smarter planet and helping people in their business tasks and personal lives."
In a video about Watson's journey (below), Ferrucci said said the nature of Jeopardy is "going to drive the technology in the right direction."
"It's got the broad domain aspect, asks all kinds of things, which was one of the challenges we really wanted to take on," he said. "It had the confidence aspect; don't answer unless you think you're right. You also had to do it really quickly."
IBM said the technology used by Watson could be helpful in areas like healthcare, to help accurately diagnose patients, to improve online self-service help desks, to provide tourists and citizens with specific information regarding cities, or prompt customer support via phone.
To prepare, Watson played more than 50 "sparring games" against former Jeopardy champions. Watson also took and passed the same Jeopardy test administered to all potential contestants.
In the video, Harry Friedman, executive producer of Jeopardy, said when IBM first approached the show, producers were intrigued but were also concerned about it being viewed as a stunt or gimmick.
"But this was different. This was the notion of knowledge acquired by a computer against knowledge acquired and displayed by the best Jeopardy players," Friedman said. "This could be something important, and we want to be a part of it."
Friedman and other producers first watched Watson in action in December 2009, when it sparred against two other human contestants.
Watson is powered by an IBM POWER7 server, which is optimized to handle the massive number of tasks that Watson must perform at rapid speeds, IBM said. The machine also has a number of proprietary technologies that handle concurrent tasks and data while analyzing information in real time.
I'm not exactly sure how a human is supposed to beat the reflexes of a computer, unless it's forced to solve the question before answering. Even then, that's no time at all. This is shifty to me, that computer must be the heart of google or something (or this is set up, idk).
EDIT: I wonder how it solves actual puzzle questions, not just fill in the blank stuff.
Found another article. no interwebz allowed. Apparently what this thing does could take a normal computer 2 hours. I wanna play video games on that.
Doesn't even look like Watson tried, even if it is a computer. The entire video, it felt like Watson was giggling to himself and letting the other guys catch up.
On January 20 2011 15:03 Torte de Lini wrote: Man will always beat machine because man created machine and its lengths of knowledge, capabilities and boundaries.
Somewhere... there's a movie with the exact same statement above.
To be honest, first the Chess-Playing computer, then the Micro-Beast AI, now this? I'm scared ): Hold me.
Last time some Israeli dudes made a computer called Deep Junior vs Kasparov which tied him 3-3, Kasparov offered a draw and the computer accepted and he said he didn't want to fuck up so he decided to offer.
On January 20 2011 15:01 whitelynx wrote: Doesn't even look like Watson tried, even if it is a computer. The entire video, it felt like Watson was giggling to himself and letting the other guys catch up.
I got the feeling that the machine was more programmed with useful facts rather than children's storybook titles, etc. Also kinda curious how the thing rings the buzzer. There has to be some inherent delay a human can actually squeeze the button that a computer can beat 100% of the time if they both think the answer and the exact same second. Also, can the computer buzz in before the question is over?
Ken Jennings!!!!!!! NOOOOO I love when IBM does something not so business oriented, although it does have underlying business reason that sort of AI can prove very useful in things like analysis.
Although one can't beat this machine in the style of an epic yarn like John Henry dieing in the end always makes it more legend worthy
On January 20 2011 15:04 Roman wrote: didnt deep blue have a grandmaster controlling it?
No. What on earth would be the point of that?
This is actually EXTREMELY cool.
What Watson is doing is actually receiving the regular jeopard questions, interpreting what information is requested, and then finding it in realtime. Technology like this is what will lead to next level search technology where you can just load up google and say "gimme a restream of the GSL" and it'll just do it for you.
On January 20 2011 15:03 Torte de Lini wrote: Man will always beat machine because man created machine and its lengths of knowledge, capabilities and boundaries.
Somewhere... there's a movie with the exact same statement above.
To be honest, first the Chess-Playing computer, then the Micro-Beast AI, now this? I'm scared ): Hold me.
what? That makes no sense at all.
So a can outrun an airplane just because it was made by humans? I can win a fight with a robotic forklift?
On January 20 2011 15:03 Torte de Lini wrote: Man will always beat machine because man created machine and its lengths of knowledge, capabilities and boundaries.
Somewhere... there's a movie with the exact same statement above.
To be honest, first the Chess-Playing computer, then the Micro-Beast AI, now this? I'm scared ): Hold me.
Last time some Israeli dudes made a computer called Deep Junior vs Kasparov which tied him 3-3, Kasparov offered a draw and the computer accepted and he said he didn't want to fuck up so he decided to offer.
That's pretty cool, sounds like the computer was overjoyed that Kasparov acknowledged it as an equal.
On January 20 2011 15:03 Torte de Lini wrote: Man will always beat machine because man created machine and its lengths of knowledge, capabilities and boundaries.
Somewhere... there's a movie with the exact same statement above.
To be honest, first the Chess-Playing computer, then the Micro-Beast AI, now this? I'm scared ): Hold me.
Last time some Israeli dudes made a computer called Deep Junior vs Kasparov which tied him 3-3, Kasparov offered a draw and the computer accepted and he said he didn't want to fuck up so he decided to offer.
That's pretty cool, sounds like the computer was overjoyed that Kasparov acknowledged it as an equal.
This is a funny way to put it. I mean the computer doesn't have any emotions.
On January 20 2011 15:04 Roman wrote: didnt deep blue have a grandmaster controlling it?
No. What on earth would be the point of that?
He's referring to an allegation that Kasparov made after the match. And actually, there would be a very meaningful point to it. Computers, especially at that time, were a bit too greedy with material. A GM controlling what lines the computer considers would be extremely useful.
On January 20 2011 15:03 Torte de Lini wrote: Man will always beat machine because man created machine and its lengths of knowledge, capabilities and boundaries.
Somewhere... there's a movie with the exact same statement above.
To be honest, first the Chess-Playing computer, then the Micro-Beast AI, now this? I'm scared ): Hold me.
Last time some Israeli dudes made a computer called Deep Junior vs Kasparov which tied him 3-3, Kasparov offered a draw and the computer accepted and he said he didn't want to fuck up so he decided to offer.
On January 20 2011 15:01 whitelynx wrote: Doesn't even look like Watson tried, even if it is a computer. The entire video, it felt like Watson was giggling to himself and letting the other guys catch up.
I got the feeling that the machine was more programmed with useful facts rather than children's storybook titles, etc. Also kinda curious how the thing rings the buzzer. There has to be some inherent delay a human can actually squeeze the button that a computer can beat 100% of the time if they both think the answer and the exact same second. Also, can the computer buzz in before the question is over?
Your reflexes are not as fast as the computer's, guaranteed. Once the computer knows the answer, it can spam the signal line. All the computer needs to do is know is the length of time the signal line must be high for in order for it to be detected. Since this is probably in the microseconds range, this means the computer could probably manage several thousand button pushes per second. You can't beat that.
On January 20 2011 15:03 Torte de Lini wrote: Man will always beat machine because man created machine and its lengths of knowledge, capabilities and boundaries.
Somewhere... there's a movie with the exact same statement above.
To be honest, first the Chess-Playing computer, then the Micro-Beast AI, now this? I'm scared ): Hold me.
Last time some Israeli dudes made a computer called Deep Junior vs Kasparov which tied him 3-3, Kasparov offered a draw and the computer accepted and he said he didn't want to fuck up so he decided to offer.
On January 20 2011 15:01 whitelynx wrote: Doesn't even look like Watson tried, even if it is a computer. The entire video, it felt like Watson was giggling to himself and letting the other guys catch up.
I got the feeling that the machine was more programmed with useful facts rather than children's storybook titles, etc. Also kinda curious how the thing rings the buzzer. There has to be some inherent delay a human can actually squeeze the button that a computer can beat 100% of the time if they both think the answer and the exact same second. Also, can the computer buzz in before the question is over?
Your reflexes are not as fast as the computer's, guaranteed. Once the computer knows the answer, it can spam the signal line. All the computer needs to do is know is the length of time the signal line must be high for in order for it to be detected. Since this is probably in the microseconds range, this means the computer could probably manage several thousand button pushes per second. You can't beat that.
Thanks for reaffirming what I said? I read in another article that it has some mechanism that mimics a human pressing a button to not be imba in this way.
Very smart move by IBM. If it was head to head vs Jennings, Watson would be less favoured than in a three-way format because when it can't answer a question, there wouldn't be two contestants to split the responses. Also, I'm wondering what the algorithm is for Daily Doubles and Final Jeopardy. Should be interesting. Edited out a typo.
What Watson is doing is actually receiving the regular jeopard questions, interpreting what information is requested, and then finding it in realtime. Technology like this is what will lead to next level search technology where you can just load up google and say "gimme a restream of the GSL" and it'll just do it for you.
On January 20 2011 15:05 SpoR wrote: Also kinda curious how the thing rings the buzzer. There has to be some inherent delay a human can actually squeeze the button that a computer can beat 100% of the time if they both think the answer and the exact same second. Also, can the computer buzz in before the question is over?
Umm...I'm not 100% sure about this, but I think in Jeopardy you're not allowed to buzz in before the answer is read completely. If you try to buzz in beforehand then your buzzer is locked for some short amount of time. This makes it important to buzz at the exact moment the answer is no longer read.
I wonder if they programmed Watson to include the time you have to respond. Your time to respond is a lot of time for a computer. It'd be cool if they could get it to have an ETA for the correct response with a certain confidence level. That way it could consider that if it buzzed in without knowing the correct response completely then it might arrive at the question in the time given.
On January 20 2011 15:05 SpoR wrote: Also kinda curious how the thing rings the buzzer. There has to be some inherent delay a human can actually squeeze the button that a computer can beat 100% of the time if they both think the answer and the exact same second. Also, can the computer buzz in before the question is over?
Umm...I'm not 100% sure about this, but I think in Jeopardy you're not allowed to buzz in before the answer is read completely. If you try to buzz in beforehand then your buzzer is locked for some short amount of time. This makes it important to buzz at the exact moment the answer is no longer read.
Yup, if you watch the video (or any episode of Jeopardy) you'll see other people trying to buzz in but not get called.
On January 20 2011 15:27 mikeymoo wrote: Very smart move by IBM. If it was head to head vs Jennings, Watson would be less favoured than in a three-way format because when it can't answer a question, there wouldn't be two contestants to split the responses. Also, I'm wondering what the algorithm is for Daily Doubles and Final Jeopardy. SHould be interesting.
I would load it up with all the game data from the other two contestants for final jeopardy and form aggregate data then run risk assessment between their score and itself to make the bet based of that data and the success rate of Watson. For double jeopardy i would guess it would just get score vs score and the success rate for topics etc. AI is very fun to think about but i imagine a bitch to implement.
On January 20 2011 15:42 BatTheMan wrote: Can it detect sarcasm or humor?
It has to in a sense thing's like puns which are often done in jeopardy and it has to understand that to narrow down the searches.
It looks like the computer beats ken to the punch alot watching that video.
The buzzer rules in Jeopardy ensure that man vs computer will pretty much be a 1-player game. You can't buzz in until the question has been fully read, and the computer gets the entire text of the question at once rather than having to process the audio then think. Given these rules, if the computer doesn't win its matches by a landslide then IBM screwed up bigtime.
The computer has a guaranteed first shot at answering the question so it just needs to be right a reasonable amount of the time. A more interesting game would be to change the rules of jeopardy to give both human and computer the full text at the start and allow buzzing at any time, but this would both make the game awkward for live TV and change the dynamics to be about buzzing in immediately and hoping you can think of an answer in the 5 seconds after buzzing you are given to state your response.
On January 20 2011 16:24 professorjoak wrote: The buzzer rules in Jeopardy ensure that man vs computer will pretty much be a 1-player game. You can't buzz in until the question has been fully read, and the computer gets the entire text of the question at once rather than having to process the audio then think. Given these rules, if the computer doesn't win its matches by a landslide then IBM screwed up bigtime.
The computer has a guaranteed first shot at answering the question so it just needs to be right a reasonable amount of the time. A more interesting game would be to change the rules of jeopardy to give both human and computer the full text at the start and allow buzzing at any time, but this would both make the game awkward for live TV and change the dynamics to be about buzzing in immediately and hoping you can think of an answer in the 5 seconds after buzzing you are given to state your response.
On January 20 2011 16:24 professorjoak wrote: The buzzer rules in Jeopardy ensure that man vs computer will pretty much be a 1-player game. You can't buzz in until the question has been fully read, and the computer gets the entire text of the question at once rather than having to process the audio then think. Given these rules, if the computer doesn't win its matches by a landslide then IBM screwed up bigtime.
The computer has a guaranteed first shot at answering the question so it just needs to be right a reasonable amount of the time. A more interesting game would be to change the rules of jeopardy to give both human and computer the full text at the start and allow buzzing at any time, but this would both make the game awkward for live TV and change the dynamics to be about buzzing in immediately and hoping you can think of an answer in the 5 seconds after buzzing you are given to state your response.
From my understanding Watson "listens" they wouldn't talk about speech analysis in the videos if Watson could just read the question before hand, they also talk about it learning answers, and if it couldn't understand the answer being said by the host or players it couldn't learn from it. Unless it had a person typing in everything, which is not what's happening.
On January 20 2011 16:24 professorjoak wrote: The buzzer rules in Jeopardy ensure that man vs computer will pretty much be a 1-player game. You can't buzz in until the question has been fully read, and the computer gets the entire text of the question at once rather than having to process the audio then think. Given these rules, if the computer doesn't win its matches by a landslide then IBM screwed up bigtime.
The computer has a guaranteed first shot at answering the question so it just needs to be right a reasonable amount of the time. A more interesting game would be to change the rules of jeopardy to give both human and computer the full text at the start and allow buzzing at any time, but this would both make the game awkward for live TV and change the dynamics to be about buzzing in immediately and hoping you can think of an answer in the 5 seconds after buzzing you are given to state your response.
On jeopardy the questions are given in both speech and text instantly. Watson has exactly the same amount of time to process the information as the human contestants.
you say
The computer has a guaranteed first shot at answering the question so it just needs to be right a reasonable amount of the time.
well obviously. Thats the entire point of the project. I don't even get what you are trying to say...The computer will buzz first if it figures out the question in "a reasonable amount of time"? Thats not really saying anything.
yeah, the computer can hit the actual buzzer faster than a human, but its the processing time that really makes a difference, not the tenth of a second racing the button.
On January 20 2011 16:24 professorjoak wrote: The buzzer rules in Jeopardy ensure that man vs computer will pretty much be a 1-player game. You can't buzz in until the question has been fully read, and the computer gets the entire text of the question at once rather than having to process the audio then think. Given these rules, if the computer doesn't win its matches by a landslide then IBM screwed up bigtime.
The computer has a guaranteed first shot at answering the question so it just needs to be right a reasonable amount of the time. A more interesting game would be to change the rules of jeopardy to give both human and computer the full text at the start and allow buzzing at any time, but this would both make the game awkward for live TV and change the dynamics to be about buzzing in immediately and hoping you can think of an answer in the 5 seconds after buzzing you are given to state your response.
On jeopardy the questions are given in both speech and text instantly. Watson has exactly the same amount of time to process the information as the human contestants.
The computer has a guaranteed first shot at answering the question so it just needs to be right a reasonable amount of the time.
well obviously. Thats the entire point of the project. I don't even get what you are trying to say...The computer will buzz first if it figures out the question in "a reasonable amount of time"? Thats not really saying anything.
yeah, the computer can hit the actual buzzer faster than a human, but its the processing time that really makes a difference, not the tenth of a second racing the button.
However its highly likely that both the players and the computer will know the answer to a question before the question is finished being read. Watch a gameshow where they allow people to buzz in before the question is finished and the host rarely completes a full question before a buzz. Therefore the computer's ability to beat a person in a buzzer race is the factor which is going to give it an advantage.
I've never watched jepordy so I assumed that people could press their buzzer whenever and that it would select whoever pressed it first after the question had finished being read. That would be a more fair system.
Of course fairness isnt that important here, this is a tech demo, not someone trying to make money by cheating a gameshow.
80% of Jeopardy questions just come straight out of high school lit class. It wouldn't be particularly hard to load most of the lit/historical data into the computer to answer the majority of questions.
The more difficult part sounds like programming the computer to discern what the questions are looking for specifically. As the article mentioned, Jeopardy questions are often framed really obscurely, or rely on extra information (like the question's category) to be answered.
On January 21 2011 00:21 Offhand wrote: 80% of Jeopardy questions just come straight out of high school lit class. It wouldn't be particularly hard to load most of the lit/historical data into the computer to answer the majority of questions.
The more difficult part sounds like programming the computer to discern what the questions are looking for specifically. As the article mentioned, Jeopardy questions are often framed really obscurely, or rely on extra information (like the question's category) to be answered.
Yeah.. as long as they can program it for the computer to process the question and know what is being asked I'd be more surprised if the computer didn't win.. Even a crappy desktop computer would destroy the world champion in a spelling bee.
On January 20 2011 15:05 tomatriedes wrote: One step closer to Skynet.
It's funny, I was bored so I started watching Terminator The Sarah Connor Chronicles YESTERDAY cause I was bored. Now this is on front page tl. It's the little coincidences that make me thing the world will end
On January 20 2011 15:27 mikeymoo wrote: Very smart move by IBM. Also, I'm wondering what the algorithm is for Daily Doubles and Final Jeopardy. Should be interesting.
Yeah, I'm curious to see how it would bid in different situations i.e. behind someone who has been very consistent vs. behind someone who has been correct on and off. I want to know how much game theory is taken into account when calculating those utility functions.
As far as the natural language processing part goes, actually understanding the answer and formulating a question is a non-trivial part of computational linguistics. That part alone is a huge undertaking even with Watson knowing the answer (from a computer science perspective anyways). Is it bad that I'm really excited to watch this?
Watson got the last question right, it was about LoTR. WHO IS SAURON? There were some questions that are way to complex for Watson to get the complete answer.
I feel like Watson definitely has the advantage in reflex speed. Also, he should have to read in the file as an image file since that's what the humans do, but I guess that really wouldn't slow him down with all the processing power he has.
Can't wait till we get something like Watson on our computers/tablets/phones. We just tell the computer to turn on. tell it to start the internet browser and say "go to teamliquid" and BAM!.
why does everyone say watson has an advantage in reflex speed? his 'brain' has to computer the answer and be confident before he tried to buzz in. its amazing that a computer can do that as quickly as he does.
On January 21 2011 00:21 Offhand wrote: 80% of Jeopardy questions just come straight out of high school lit class. It wouldn't be particularly hard to load most of the lit/historical data into the computer to answer the majority of questions.
And 80% of statistics are completely made up on the spot without any reference. Seriously though, this is not even close to true. The information Watson has in its database is the equivalent of millions of books. To seek out the pertinent information and answer the question correctly in such a short period of time (~3 seconds) is truly an amazing feat in artificial intelligence programming
Is it possible to do something like this but for StarCraft (Brood War or Wings of Liberty)? I would be awesome to see if IBM or some huge company can create the ultimate computer AI that can give progamers a run for their money.
On February 15 2011 11:25 Neo7 wrote: Is it possible to do something like this but for StarCraft (Brood War or Wings of Liberty)? I would be awesome to see if IBM or some huge company can create the ultimate computer AI that can give progamers a run for their money.
However, his ability to come up with the answer is not instant. He would have to process the question, search through a database or a network of connections of some sort and then finally produce the answer. Search problems have traditionally been quite slow in AI and Watson is probably still slower than a human in this regard.
On February 15 2011 11:25 Neo7 wrote: Is it possible to do something like this but for StarCraft (Brood War or Wings of Liberty)? I would be awesome to see if IBM or some huge company can create the ultimate computer AI that can give progamers a run for their money.
I believe many people have already tried this.
Back during the beta (and maybe even now), there were many AIs developed for SC WoL. For example Green Tea, Dark Tea and some other stuff http://sc2.nibbits.com/forums/17/tools
On February 15 2011 11:25 Neo7 wrote: Is it possible to do something like this but for StarCraft (Brood War or Wings of Liberty)? I would be awesome to see if IBM or some huge company can create the ultimate computer AI that can give progamers a run for their money.
If esports gets big enough perhaps that will be the next grand challenge. First chess, then Jeopardy, then SC2, then... go?
Just a heads up to people. There is an episode of the PBS program Nova available on their website discussing the development of "Watson". Its a pretty good watch if anyone is interested in some more information. I didn't see anyone bring up the episode from scanning through the thread if im repeating information i apologize.
it seems like, with 'easier' questions he is just faster than the others who also know the answer. I am a bit disappointed that he doesn't have voice input. However, awesome!
This is amazing, it's really breaking down the barrier between nerds and non-nerds. The ability to speak in free english to a computer and get a natural response that's actually accurate is so much more powerful than people think. What's next, getting into your car and saying "hey, drive me to the airport but take the scenic route. Also, hit a starbucks on the way" Amazing.
On February 15 2011 11:59 Endymion wrote: This is amazing, it's really breaking down the barrier between nerds and non-nerds. The ability to speak in free english to a computer and get a natural response that's actually accurate is so much more powerful than people think. What's next, getting into your car and saying "hey, drive me to the airport but take the scenic route. Also, hit a starbucks on the way" Amazing.
I'm hoping for quick medical diagnosis and maybe chemical trouble-shooting for those dealing with possibly poisonous combinations.
just finished watching it...wow...i totally thought watson was going to dominate because of his lead halfway through (5200 to 1000 to 200 i think) but its close (watson tied at 5000 with brad)....CAN'T WAIT FOR TOMORROW
I think Jennings figured out the best strat to maybe beating watson, he began to just try and buzz in on every question and then figure it out, Rutters than began to follow his example. Should be interesting.
On February 15 2011 12:34 durza wrote: I think Jennings figured out the best strat to maybe beating watson, he began to just try and buzz in on every question and then figure it out, Rutters than began to follow his example. Should be interesting.
thats how jennings did it for 75 episodes straight lol
Very misleading OP. Rutters came back and tied with Watson... Tomorrow should be awesome! Like someone said before me Jennings strat seems like its the way to go. Watson has proven to be a formidable opponent!
I can make this program in 5 minutes. Step 1: upload some software that will understand human voice Step 2: google the question because it's a fucking computer. You can just stick a tiny usb stick that connects to the internet underneath all those layers of computers. Step 3: write the code that will make the computer pick the most likely answer
congradulations IBM, you now have marketed your products to an even greater audience.
On February 15 2011 13:39 Lucid90 wrote: I can make this program in 5 minutes. Step 1: upload some software that will understand human voice Step 2: google the question because it's a fucking computer. You can just stick a tiny usb stick that connects to the internet underneath all those layers of computers. Step 3: write the code that will make the computer pick the most likely answer
congradulations IBM, you now have marketed your products to an even greater audience.
You're completely failing to understand the process. The computer is actually understanding the use of the complexities of the English language. Software that recognizes voice commands is very primitive and can only understand basic statements. It can't do anything like understand the various language devices used to make Jeopardy questions.
On February 15 2011 13:39 Lucid90 wrote: I can make this program in 5 minutes. Step 1: upload some software that will understand human voice Step 2: google the question because it's a fucking computer. You can just stick a tiny usb stick that connects to the internet underneath all those layers of computers. Step 3: write the code that will make the computer pick the most likely answer
congradulations IBM, you now have marketed your products to an even greater audience.
Its funny that the real magic is in step 3, you make it sound so easy but its understanding what the question is and picking the correct answer thats hard. especially with the puns and word play that is often used on that show. Thought it was really cool to watch my GF really liked it were gonna watch tomorrow.
Natural language processing is definitely not easy. Assuming there wasn't any dodgy stuff done beforehand, the speed and accuracy of the machine is impressive. Shame it doesn't seem to understand human speech and eliminate incorrect answers ("uh he just said that Watson").
Context is a big barrier for these programs, they actually talked about it on the show how a major hurdle is making it understand the difference between Usain Bolt running in a 100m race and McCain running for president; two phrases using one identical word but two whole different contexts and the program has to be able to make the distinction between the two words quickly and accurately.
Woww many people have said it already but after watching all those videos and really understanding it, this is truly amazing. I cannot wait for the results of this to start appearing in my everyday life, its going to make it so much better. My life that is.
On February 15 2011 13:39 Lucid90 wrote: I can make this program in 5 minutes. Step 1: upload some software that will understand human voice Step 2: google the question because it's a fucking computer. You can just stick a tiny usb stick that connects to the internet underneath all those layers of computers. Step 3: write the code that will make the computer pick the most likely answer
congradulations IBM, you now have marketed your products to an even greater audience.
i know the secret to eternal life
Step 1: don't get hit by a bus Step 2: eat fucking vegetables and shit Step 3: cure aging
congradulations was it that hard?? y doesnt ibm do this
On February 15 2011 13:39 Lucid90 wrote: I can make this program in 5 minutes. Step 1: upload some software that will understand human voice Step 2: google the question because it's a fucking computer. You can just stick a tiny usb stick that connects to the internet underneath all those layers of computers. Step 3: write the code that will make the computer pick the most likely answer
congradulations IBM, you now have marketed your products to an even greater audience.
Please educate yourself by watching that 20 minute video on the process of Watson's decision making.
This Is just a normal computer on steroids. Thats it.
It's not a true learning computer. Just a powerful, well programmer one. It's a big gimmick, that's all it can ever be.
Skynet? Ha!
Yeah right..
You have no idea, I am not sure how well they accomplished it, but it being able to actually even reasonably compete is big achievement. Natural language processing and actual(even if partial) "understanding" of text is pretty hard task. Also define learning computer before stating it is not learning.
On February 15 2011 13:39 Lucid90 wrote: I can make this program in 5 minutes. Step 1: upload some software that will understand human voice Step 2: google the question because it's a fucking computer. You can just stick a tiny usb stick that connects to the internet underneath all those layers of computers. Step 3: write the code that will make the computer pick the most likely answer
congradulations IBM, you now have marketed your products to an even greater audience.
Even first step is extremely hard not yet solved problem, so lol at you saying it will take 5 minutes. As for the second try googling Jeopardy questions and see what you get, mostly nothing relevant. Step 3 is also extremely hard to do well on such a broad domain.
Does Watson actually have to interpret the host's speech in order to understand the question? If so, that's pretty damn impressive. However, I imagine if the questions focused more on linguistic subtleties, humor and abstract concepts it wouldn't fare nearly as well as a human.
I wonder how long it will take for the technology and AI that powers Watson to be small enough to fit in a smartphone. Can you imagine having your own portable genius that can instantly answer any random question you verbally ask it? The potential for such technology is difficult to fully imagine.
edit: I also love all the pseudo-programmers who believe that Watson is nothing more than a glorified search engine when they clearly have no idea what Watson really is or what it took to develop.
Update, after the first full round Watson is tied for first with $5000. Round 2 and 3 are still to air. You can watch them during Jeopardy's timeslot on tv in the US Tuesday and Wednesday night.
On February 15 2011 15:08 popnyah wrote: Does Watson actually have to interpret the host's speech in order to understand the question? If so, that's pretty damn impressive. However, I imagine if the questions focused more on linguistic subtleties, humor and abstract concepts it wouldn't fare nearly as well as a human.
I wonder how long it will take for the technology and AI that powers Watson to be small enough to fit in a smartphone. Can you imagine having your own portable genius that can instantly answer any random question you verbally ask it? The potential for such technology is difficult to fully imagine.
edit: I also love all the pseudo-programmers who believe that Watson is nothing more than a glorified search engine when they clearly have no idea what Watson really is or what it took to develop.
Just ignore them they are likely doing nothing more then inciting an out raged response to the oversimplification of a complex system that took years to create and some very talented people over at IBM. To say that the task is easy just from that statement alone would be just a show to how incompetent they are. I'm more interested in the word play questions and if they will do a little bit onto the questions that Watson had trouble with esp the ones that it missed.
On February 15 2011 15:08 popnyah wrote: Does Watson actually have to interpret the host's speech in order to understand the question? If so, that's pretty damn impressive. However, I imagine if the questions focused more on linguistic subtleties, humor and abstract concepts it wouldn't fare nearly as well as a human.
I wonder how long it will take for the technology and AI that powers Watson to be small enough to fit in a smartphone. Can you imagine having your own portable genius that can instantly answer any random question you verbally ask it? The potential for such technology is difficult to fully imagine.
edit: I also love all the pseudo-programmers who believe that Watson is nothing more than a glorified search engine when they clearly have no idea what Watson really is or what it took to develop.
Yes and no, it can't actually understand speech (come back in 10 years Watson when you can actually understand speech so you don't repeat wrong answers) but it does interpret and analyse the answer.
Instead of using specialised databases for Jeopardy, it uses alogarithms to break down text and attempts to find the subject matter, verbs, grammar, etc. From there correlates and interrelates words and phrases so it can make confidence percentages (Jeopardy displays this as the answer panel) and hypotheses. Basically it can "learn" by itself in a very primative way and if it gets something wrong it goes "oh I'm wrong, these sets of words can be used in this way!" and changes its confidences and construct possibly more accurate hypotheses.
On February 15 2011 15:15 NEOtheONE wrote: Update, after the first full round Watson is tied for first with $5000. Round 2 and 3 are still to air. You can watch them during Jeopardy's timeslot on tv in the US Tuesday and Wednesday night.
Little bit of a correction, they're going to play 2 full games over the course of 3 days, and they only played Round 1 of Game 1 on Monday, likely they will play Rounds 2 and 3 (Double and Final Jeopardy) tonight, and the entirety of Game 3 on Wednesday.
At first I was very disappointed to learn that Watson could not understand speech. My first reaction was something like "couldn't they install a 10$ speech-to-text software on that thing and call it a day?"
I realize now that not only would this give him a significant disadvantage (other contestants can read the answers on the board) but requires him to distinguish answers from anything else said on the show. It's more complexity for no real reason.
I was looking forward to having Alex interview Watson about where he was born and if he understands human emotions. Oh well.
Just watched it. Pretty impressive stuff. Will be amazing when they come out with "DeepQA" software like that on the internet ala Google but much better.
Getting a computer to understand is very hard. If you type an abstract, never before asked question into google, imagine what you'd find for pages. Probably almost nothing relating to the question, much less an answer.
Watson has an advantage in that it's reflexes are instantaneous, where as a human will take time to press the button from when he realizes he knows the answer. But that doesn't always make up for the lack of real comprehension.
I forgot this was airing until half an hour after it was over. I'll make sure to watch it today and tomorrow.
Watson has an advantage in that it's reflexes are instantaneous, where as a human will take time to press the button from when he realizes he knows the answer. But that doesn't always make up for the lack of real comprehension.
it's not as different as you think. a human can take less time for the answer, or more, depending on the way their brain is wired. humans also have mental reflexes that are "instantaneous"
The thing is, humans win this challenge no matter what. as said in the 20 minute video in the OP, all it takes is a tuna sandwich and a glass or water to power a human brain that can fit in a shoebox. compare that to what it takes Watson to compete, humans win every time.
For anyone that was really interested in the programming behind Watson there was a special a few days ago on PBS about how Watson was made. It was pretty interesting.
Watson has an advantage in that it's reflexes are instantaneous, where as a human will take time to press the button from when he realizes he knows the answer. But that doesn't always make up for the lack of real comprehension.
it's not as different as you think. a human can take less time for the answer, or more, depending on the way their brain is wired. humans also have mental reflexes that are "instantaneous"
Ummm, by reflexes, I'm referring to the fact that it takes a human body to react. A delay of .15s may not sound like alot, but in terms of a computer where his delay is basically the speed of light, it makes a different.
Of coarse "searching" speeds for information would be different in different cases.
Watson has an advantage in that it's reflexes are instantaneous, where as a human will take time to press the button from when he realizes he knows the answer. But that doesn't always make up for the lack of real comprehension.
it's not as different as you think. a human can take less time for the answer, or more, depending on the way their brain is wired. humans also have mental reflexes that are "instantaneous"
I'm pretty sure Watson's buzzer reaction is much much faster than human muscle reaction. However, the information processing that takes place in human brains is our advantage. Watson may have more data and facts stored, but its search and recall algorithm can't possibly be more robust (this does not necessarily mean slower) than the human brain's.
On February 16 2011 03:27 metaphoR wrote: The thing is, humans win this challenge no matter what. as said in the 20 minute video in the OP, all it takes is a tuna sandwich and a glass or water to power a human brain that can fit in a shoebox. compare that to what it takes Watson to compete, humans win every time.
Thats what they said about the technology needed to send people into orbit, which they did with the equivalent of a modern day graphing calculator. It's only a matter of time until the hardware gets faster and smaller.
On February 16 2011 03:27 metaphoR wrote: The thing is, humans win this challenge no matter what. as said in the 20 minute video in the OP, all it takes is a tuna sandwich and a glass or water to power a human brain that can fit in a shoebox. compare that to what it takes Watson to compete, humans win every time.
So a very muscular person loses by default to a non-muscular one ? The point of this game is to come up with the correct answer, weight and power requirements are not part of it yet
This is really crazy. It will revolutionize medicine and developing differential diagnoses. I have a feeling that we will see this more and more in medicine over our lifetime.
This is nice and timely. I was about to finish up a 10-page essay on AI, and I was going to include a section about Deep Blue. This seems much more relevant and up to date
haha my dad surprised me and visited me where i live (he lives like 2 hours away) and we watched this together (he works for IBM and I work for an iOS company) and the computer was very good i was impressed. It would be really really scary if Watson had internet access (and had some kind of speech recognition)... we're not so far away from personal ai helpers (in labs, homes, vehicles) I can't fucking wait although I may be waiting for a long time
On February 16 2011 06:04 michaelthe wrote: Does it instantly buzz in then get the 5 seconds to try and find an answer? Or find the answer then buzz in?
it develops confidence in an answer based on a % it thinks it's correct. when confidence reaches a certain point it buzzes in. the percentage of confidence it needs to buzz in is based on it's current score compared to the score of it's opponents.
On February 15 2011 13:39 Lucid90 wrote: I can make this program in 5 minutes. Step 1: upload some software that will understand human voice Step 2: google the question because it's a fucking computer. You can just stick a tiny usb stick that connects to the internet underneath all those layers of computers. Step 3: write the code that will make the computer pick the most likely answer
congradulations IBM, you now have marketed your products to an even greater audience.
i know the secret to eternal life
Step 1: don't get hit by a bus Step 2: eat fucking vegetables and shit Step 3: cure aging
congradulations was it that hard?? y doesnt ibm do this
On February 16 2011 06:04 michaelthe wrote: Does it instantly buzz in then get the 5 seconds to try and find an answer? Or find the answer then buzz in?
Watson only "buzzes in" via robotic hand movement when it finds an answer that exceeds a specific certainty threshold that is programmed in.
On February 16 2011 03:27 metaphoR wrote: The thing is, humans win this challenge no matter what. as said in the 20 minute video in the OP, all it takes is a tuna sandwich and a glass or water to power a human brain that can fit in a shoebox. compare that to what it takes Watson to compete, humans win every time.
So a very muscular person loses by default to a non-muscular one ? The point of this game is to come up with the correct answer, weight and power requirements are not part of it yet
plus, humans take 20 years to boot up after putting them together
It should be noted that is based on an extremely small sample size, so it is not conclusive that Watson is "70% accurate" at producing the correct answer. Also, what many still fail to realize is how difficult it is to program a computer to understand the nuances of the english language into such a way to be able to discern the important information from the unimportant bits, and then also find the correct answer. For instance, you need to program the AI to understand the context of a sentence to understand the keywords and puns.
To everyone critical of Watson: Read some articles and videos on how Watson operates before forming your opinions on whether or not it is an impressive feat.
It should be noted that is based on an extremely small sample size, so it is not conclusive that Watson is "70% accurate" at producing the correct answer. Also, what many still fail to realize is how difficult it is to program a computer to understand the nuances of the english language into such a way to be able to discern the important information from the unimportant bits, and then also find the correct answer. For instance, you need to program the AI to understand the context of a sentence to understand the keywords and puns.
To everyone critical of Watson: Read some articles and videos on how Watson operates before forming your opinions on whether or not it is an impressive feat.
Assuming they are correct the posts saying the following things were enough to impress me:
Instead of using specialised databases for Jeopardy, it uses alogarithms to break down text and attempts to find the subject matter, verbs, grammar, etc. From there correlates and interrelates words and phrases so it can make confidence percentages (Jeopardy displays this as the answer panel) and hypotheses. Basically it can "learn" by itself in a very primative way and if it gets something wrong it goes "oh I'm wrong, these sets of words can be used in this way!" and changes its confidences and construct possibly more accurate hypotheses.
it develops confidence in an answer based on a % it thinks it's correct. when confidence reaches a certain point it buzzes in. the percentage of confidence it needs to buzz in is based on it's current score compared to the score of it's opponents.
How is everyone else not impressed by that? That's very impressive for a computer, even if it is the simplest way for a human to learn.
On February 16 2011 07:56 lixlix wrote: The people relating this to Google are missing the point. The practical applications for something like Watson are super vast.
For one, human helpdesks and tech support will be rendered obsolete and that is only the start.
Hey, hey! Don't say that. . . I'll be out of a job then. I'll seriously kill myself if I'm still doing this bullshit when this technology becomes mainstream.
On February 16 2011 06:04 michaelthe wrote: Does it instantly buzz in then get the 5 seconds to try and find an answer? Or find the answer then buzz in?
it develops confidence in an answer based on a % it thinks it's correct. when confidence reaches a certain point it buzzes in. the percentage of confidence it needs to buzz in is based on it's current score compared to the score of it's opponents.
On February 15 2011 13:39 Lucid90 wrote: I can make this program in 5 minutes. Step 1: upload some software that will understand human voice Step 2: google the question because it's a fucking computer. You can just stick a tiny usb stick that connects to the internet underneath all those layers of computers. Step 3: write the code that will make the computer pick the most likely answer
congradulations IBM, you now have marketed your products to an even greater audience.
i know the secret to eternal life
Step 1: don't get hit by a bus Step 2: eat fucking vegetables and shit Step 3: cure aging
congradulations was it that hard?? y doesnt ibm do this
I think this is absolutely amazing that they are able to develop software this far. As a programmer myself I think this is a breathtaking piece of technology and I can't wait to see what they do next. Utilizing AI in all sectors could very well be the next major set of advancements for technology. Already I can see applications in a medical, engineering, and even personal settings (not to mention warfare) as a more intelligent and easier way to accomplish complicated tasks.
I have never really understood jeopardy, so when I heard about Watson from some friends, I wasn't too impressed since I thought it would just be some kinda google search.
However, after looking at the questions and videos, I am absolutely amazed and excited. Natural language processing is one of those things where humans outperforms computers. However, the technology used for Watson will eventually allow us to have actual conversations with computers.
I look forward to being able to go up to my computer and ask "find me some build orders lulz" and it would go on the internet, google it up, read through all the forums and return with a bunch of build orders and arguments on which are good or bad.
On February 16 2011 08:33 555 wrote: I have never really understood jeopardy, so when I heard about Watson from some friends, I wasn't too impressed since I thought it would just be some kinda google search.
However, after looking at the questions and videos, I am absolutely amazed and excited. Natural language processing is one of those things where humans outperforms computers. However, the technology used for Watson will eventually allow us to have actual conversations with computers.
I look forward to being able to go up to my computer and ask "find me some build orders lulz" and it would go on the internet, google it up, read through all the forums and return with a bunch of build orders and arguments on which are good or bad.
Except it would only give you terrible BO's for putting "lulz" at the end of the statement.
On February 16 2011 09:36 Isomer wrote: I'll only be scared when they design a Starcraft AI that beats Jaedong and Flash
I know this is an exaggeration to extol how good Jaedong and Flash are but if you took even 1/10 the resources thrown at creating Watson, you could easily achieve this considering the computer will have perfect multitask already.
Someone/something had a good game today, nice to see him not get the final jeopardy, lets hope humanity can come back tomorrow. Humans Fighting!!!! lol
I don't know. I mean, the computer gets the information fed to it electronically, so it knows the exact time at which you are allowed to buzz. Let's say it's just tied with the humans in information and knowledge. Then it has a natural advantage, because the humans must listen to Trebek's voice to determine when the question ends and they're allowed to buzz, while Watson gets a signal sent right to it. Doesn't seem fair.
in the back of my mind i thought, man, why didn't anyone else use this strategy?
Its pretty commonly done in Jeopardy by past human players actually.
I would be really interested in seeing the wagering algorithm for DD and final Jeopardy.
Referring to all the Daily Doubles from this episode (Watson got all three in the first game)
The first DD in regular Jeopardy, it happened very early in the game where Watson only had $400. It's natural for any champion contestant to wager the maximum possible ($1000) in these cases if they're compentent.
In Double Jeopardy, the first DD was on Cambridge, a one-word, simple category. The wager was quite big ($6,435 I think) refferring to the amount of material that can be based on one subject keyword. Also can be a wager that allowed Watson to jump way ahead Brad and Ken.
The second DD was on "The Art of the Steal" in which Watson incorrectly answered one question previously despite a 97% confidence level (the actual answer was 11% on his confidence level). Hence, he didn't wager as much as the first one (about $1,201 despite being way ahead of Ken and Brad) because he was unsure of what the clue was asking for. This is why Watson had a 34% or so confidence on Baghdad (altho the right answer).
On February 16 2011 10:42 DTK-m2 wrote: I don't know. I mean, the computer gets the information fed to it electronically, so it knows the exact time at which you are allowed to buzz. Let's say it's just tied with the humans in information and knowledge. Then it has a natural advantage, because the humans must listen to Trebek's voice to determine when the question ends and they're allowed to buzz, while Watson gets a signal sent right to it. Doesn't seem fair.
It doesn't really matter if it isn't fair. The simple fact that it can be done is astonishing.
ps. spoilers would have been nice >< I have to wait 2mo to download it!
I know those are available. but I just found out that actually the wagering algorithm for Watson itself is slightly different due to the 2 day total I believe. It also factors in a confidence in the category.
On February 16 2011 11:11 Tiegrr wrote: Watch the NOVA documentary on Watson for clearer details on how they did it. It's really interesting if you're a tech nerd like me.
i had an audition for jeopardy in september and did pretty well, damn well for a 21 year old. im stoked that a quality company like IBM would sponsor jeopardy which, lets admit is getting less and less popular, however a 3 day event with prior contestants... thats 6 slots that could have gone to me, basically to show off a really good search engine.
On February 16 2011 11:46 carbon_based wrote: i had an audition for jeopardy in september and did pretty well, damn well for a 21 year old. im stoked that a quality company like IBM would sponsor jeopardy which, lets admit is getting less and less popular, however a 3 day event with prior contestants... thats 6 slots that could have gone to me, basically to show off a really good search engine.
I am on West coast so watching this ownage by Watson currently...not fair, they could have given humans a little more time to move their fingers to push the button lol.
On February 16 2011 11:46 carbon_based wrote: i had an audition for jeopardy in september and did pretty well, damn well for a 21 year old. im stoked that a quality company like IBM would sponsor jeopardy which, lets admit is getting less and less popular, however a 3 day event with prior contestants... thats 6 slots that could have gone to me, basically to show off a really good search engine.
Well, the whole point was to compete against the best humans. If Ken and Brad are having trouble keeping up, do you think you could do better? =P
On topic, after watching those background videos it's pretty mindblowing how impressive Watson is. Once this technology becomes cheaper and more mainstream it'll be amazing what effects it will have.
And damnit, I was busy watching the NOVA Docu on the show and missed the program on TV >.< It goes really indepth about this 'machine learning' where it takes answers and bases the format of the questions or sentence structures and picks the best answers for categories. Really really intense shit.
Watson shat on Ken and Brad on Tues. Lets hope for a better fight tomorrow.
I would hope that Ken and Brad are given a little more time to hit the buzzer. It seems that if Watson knows a question it will guess first, regardless of whether the other two have to say about it.
The reason Ken and Brad are getting shit on is because the computer can buzz much faster than they can. You can see half the time the expression on Kens face is fucking damn it I didn't buzz fast enough but knew the answer. The computer allegedly does it the same way with some push technology but it activates the push the second that the buzzer is available meaning 80% of the time it's going to buzz much faster than Ken and Brad.
On February 16 2011 12:57 Chimpalimp wrote: Watson shat on Ken and Brad on Tues. Lets hope for a better fight tomorrow.
I would hope that Ken and Brad are given a little more time to hit the buzzer. It seems that if Watson knows a question it will guess first, regardless of whether the other two have to say about it.
I think the feed of the question should be given to Watson the moment trebek finishes talking (if it's not already). Seems kind of like he gets the full question the moment it pops on the screen. The buzzer function is still limited by a mechanical counterpart to emulate human thumb press, but still if it's processing a few seconds before the humans that's an advantage. And since there is a start time for when you can actually buzz in, that may be programmed in to coordinate with that delay of the mechanical button press making it very very very hard to beat the computer at pressing the button.
So in other words, everyone knows the answer, but ofc the computer wins the button pressing contest.
Regardless, the point of the whole exhibition is not necessarily to see who's going to win, but to showcase how fucking amazing computing is these days.
(if we could still see watson's answers at the same moment the humans buzzed in faster, we could still determine this a success for computing. It makes no difference)
I've always hated the "button pushing" aspect of the game. Half the time EVERYONE knows the answer to the questions, but the game comes down to who has the fastest reflexes. I would much rather see a format such as final jep where everyone gets to answer the question and the final score would be a better representation of the true champion.
On February 16 2011 12:57 Chimpalimp wrote: Watson shat on Ken and Brad on Tues. Lets hope for a better fight tomorrow.
I would hope that Ken and Brad are given a little more time to hit the buzzer. It seems that if Watson knows a question it will guess first, regardless of whether the other two have to say about it.
I think the feed of the question should be given to Watson the moment trebek finishes talking (if it's not already). Seems kind of like he gets the full question the moment it pops on the screen. The buzzer function is still limited by a mechanical counterpart to emulate human thumb press, but still if it's processing a few seconds before the humans that's an advantage. And since there is a start time for when you can actually buzz in, that may be programmed in to coordinate with that delay of the mechanical button press making it very very very hard to beat the computer at pressing the button.
So in other words, everyone knows the answer, but ofc the computer wins the button pressing contest.
Regardless, the point of the whole exhibition is not necessarily to see who's going to win, but to showcase how fucking amazing computing is these days.
(if we could still see watson's answers at the same moment the humans buzzed in faster, we could still determine this a success for computing. It makes no difference)
From what I understand, it does not actuate the button until it reaches a certain confidence level. It is fed the question as soon as the question is finished I believe. As impressive as Watson's language processing is, any computer could reach the same confidence levels with Watson's software, given a few hours or days that is.
On February 16 2011 13:10 Bajadulce wrote: I've always hated the "button pushing" aspect of the game. Half the time EVERYONE knows the answer to the questions, but the game comes down to who has the fastest reflexes. I would much rather see a format such as final jep where everyone gets to answer the question and the final score would be a better representation of the true champion.
Good point. We can't ever know the human answers/guesses but we can always see Watson's from the display.
Imo they should redo the show and do every question like that. just to see how they really stack up.
I think it would be more interesting if Watson had an audio/video feed on the avatar, and actually had to listen to Trebek like the other contestants. It would add another layer of complexity (and arguably fairness) to each question, as a lot of the time, intonation and pacing are clues to the answer.
Pretty spectacular stuff, though. Imagine the day when we can carry around Watson-like computers on our watches or phones.
On February 16 2011 13:23 sevia wrote: I think it would be more interesting if Watson had an audio/video feed on the avatar, and actually had to listen to Trebek like the other contestants. It would add another layer of complexity (and arguably fairness) to each question, as a lot of the time, intonation and pacing are clues to the answer.
Pretty spectacular stuff, though. Imagine the day when we can carry around Watson-like computers on our watches or phones.
yea, if you watch the NOVA docu it talks alot about speech recognition. It didn't look very erroneous so I think they should have to be perfectly honest.
A person can read the question before Alex reads it aloud if they choose to do so and also buzz in without "knowing" the answer yet, if they know they can recall or remember it before time is up or shortly after. What's somewhat scary is actually that it shows the computing power that it can process the question, formulate an answer with high enough confidence to warrant buzzing in, and buzz in in that small window.
It's kind of ironic to see Ken Jennings of all people getting frustrated at a competitor out speeding him to the buzzer.
I don't think the rules should be changed for this challenge, but I can definitely appreciate the opinion that this has basically turned into a game of Watson vs Jeopardy Clues, not Man vs Machine, and that makes it uninteresting for the viewer.
On February 16 2011 14:19 Megaman703 wrote: It's kind of ironic to see Ken Jennings of all people getting frustrated at a competitor out speeding him to the buzzer.
I don't think the rules should be changed for this challenge, but I can definitely appreciate the opinion that this has basically turned into a game of Watson vs Jeopardy Clues, not Man vs Machine, and that makes it uninteresting for the viewer.
as a computer scientist i find it fascinating, whether it is a close game or not
Oh I agree, I still find the games fascinating, and I'm not downplaying the achievement of the IBM team, but I'm just saying I do see the merit in people complaining that the entertainment value in these episodes is lower than normal episodes. It's just like when Ken Jennings was streaking; shows got boring and people stopped watching his games, and the only reason people would watch him crush opponents on a daily basis was to possibly see him lose.
this comp is hax. it probably has all the info necessary since it can potentially store millions of volumes of encyclopedias worth of data. the tough part is understanding language, which they've been able to achieve. insane AI for jeopardy?
On February 16 2011 13:37 lixlix wrote: 200 Power7s is a lot of computing power.
so much so that it will probably be a long while before we see this on a consumer level.
nah, you can have 1 machine work on multiple queries as real answers such as tech support wont need such a broad database nor is it under time constraints.
On February 16 2011 13:37 lixlix wrote: 200 Power7s is a lot of computing power.
so much so that it will probably be a long while before we see this on a consumer level.
Well not that I know much on the topic, but since this sort of linguistic intelligent technology is in its infancy I don't think it's so unbelievable that in 10 year they'll have something much better than Wilson is in its current state, that also is much more efficiently. Plus consumer processing will continue to improve.
Plus, if you have to wait 5 minute for an answer it would still have its uses. Though people may be used to .001 second queries.
First of all, Watson probably hacked the jeopardy system. Not that hard to program a computer to access the jeopardy system if it is connected directly to it. This is why it selected the daily double questions out of the blue. Comeon, it picked something for 800 and the rest for 200. Then it randomly picks something for 600 and then normally picks again when all the daily double are gone.
Secondly, Watson had more time to process the question than the human contestants. It receives the question immediately as it appears while both Ken and Brad has to slowly absorb the question until the host is done reading it. This gives Watson 5 extra seconds to process the question while the human contestants has a split second to process it and buzz in. I noticed that Ken and the other guy tried to buzz in sooo many time just a split second too late.
On February 16 2011 16:01 aztrorisk wrote: Watson is a major disappointment.
Ken > Watson
First of all, Watson probably hacked the jeopardy system. Not that hard to program a computer to access the jeopardy system if it is connected directly to it. This is why it selected the daily double questions out of the blue. Comeon, it picked something for 800 and the rest for 200. Then it randomly picks something for 600 and then normally picks again when all the daily double are gone.
Secondly, Watson had more time to process the question than the human contestants. It receives the question immediately as it appears while both Ken and Brad has to slowly absorb the question until the host is done reading it. This gives Watson 5 extra seconds to process the question while the human contestants has a split second to process it and buzz in. I noticed that Ken and the other guy tried to buzz in sooo many time just a split second too late.
Ken > Watson
Its not hacking the system... I'm just contemplating this, but I believe through computer learning, that it picked up on the fact or it was coded into Watson, that the daily doubles always appear within that 1200-2000 range, and after it sought them out, it went back to the same style it was playing in the first round.
Also human contestants see the question at the same time, but possibly its takes awhile to process it, or Watson has godlike "clicking" speeds, so that the other contestants couldn't buzz in as the host has to finish the question before they can buzz it, whereas if what you say is true, what doesn't stop Watson from buzzing in before the host finishes the question, as he has no voice recognition IIRC. Therefore he wouldnt technically know when the host finishes the question.
Daily Doubles are fairly consistent in location. They're skewed towards the bottom and Watson doesn't give a shit about money or difficulty because he's a computer so he always shoots big.
There's a lot of past games built in to watson so it actively tries to hunt for the DDs.
As soon as it gets a sizable lead, it switches strategy to picking low value questions to prevent its opponents from coming back and not gambling on DD questions.
Hmm, if I were the producers I'd made round 2 much more skewed against Watson's strengths. So everyone gets the story they want - IBM gets the "wow the computer is pretty competitive" while the audience/general public still gets the "phew humans are still superior (for now)"
But watched part of the episode - must be kind of spooky for the human contestants to play against that thing...just a screen in place of a person.
whereas if what you say is true, what doesn't stop Watson from buzzing in before the host finishes the question, as he has no voice recognition IIRC. Last edit: 2011-02-16 16:15:49 Therefore he wouldn't technically know when the host finishes the question.
How would watson know when to buzz in even without my hypothesis? Therefore, there MUST be a way that he is notified when the question finished and if he as a certainty rate of over 50%, he will immediately buzz in.
I don't know how he knows when the host finishes the question, but all I know is there is some method in which he knows when the host finishes the question and he is able to buzz in immediately if he is done processing and has a certainty rate of higher than 50%.
For matches against the machine, if a contestent buzzes in within some small time period (like .25 seconds) and the computer does as well, someone should be randomly selected to answer.
On February 16 2011 19:39 igotmyown wrote: For matches against the machine, if a contestent buzzes in within some small time period (like .25 seconds) and the computer does as well, someone should be randomly selected to answer.
Yeah these matches are kind of stupid. They didn't prove they could build a computer that can beat a human at jeopardy, all they've proven is that they've built a robot that can click a buzzer faster than a human..
Apparently you're not allowed to press the buzzer before the host finishes reading (presumably someone in a back room flips a switch every time). Humans have to guess when this occurs, while Watson gets an electronic feed. This is a pretty huge advantage.
This is a blog post from Jon Lenchner, IBM researcher for how it looks for the daily doubles.
If Watson gets to choose a category and clue, its first priority is finding any remaining of the three Daily Doubles in a game. These clues allow a contestant to wager a specific dollar amount on the clue without worry of the other two contestants buzzing in. Jennings, Rutter and Watson have a high chance to answer these correctly, so Daily Doubles provide three opportunities for a critical score boost.
The Watson Research team studied the historical distribution of Daily Doubles and found they appear most-frequently in the three bottom rows, with the fourth being the most common. Daily Doubles also most frequently appear in the first column. Watson also makes use of even more statistics to dynamically predict their location based on what has been exposed so far in a game.
Once the Daily Doubles are off the board, Watson looks for the lowest clue value in a category, for which there are still a significant number of high value clues. Lower value clues help it get the gist of a category with less risk, so that it has a better shot at the high value clues to come.
I am having a hard time understanding this. With the wikipedia and google I don't think it is possible to get an answer wrong. Of course the computer would win.
On February 16 2011 19:39 igotmyown wrote: For matches against the machine, if a contestent buzzes in within some small time period (like .25 seconds) and the computer does as well, someone should be randomly selected to answer.
Yeah these matches are kind of stupid. They didn't prove they could build a computer that can beat a human at jeopardy, all they've proven is that they've built a robot that can click a buzzer faster than a human..
The competitive aspect is almost irrelevant. It can understand a sentence in 2-3s that's all that matters.
The wait for the host to finish reading the question rule is dumb however. They should let ppl buzz whenever they feel like it.
After years of planning, IBM's learning, human-aware computer Watson was put to a competition like no other - a match of Jeopardy against quiz show heavyweights Ken Jennings and Brad Rutter. The result - Watson won. Barely.
The match, which Watson has been training for since 2009, was officially announced last year. At the end of last week, the multi-episode feature where Watson faces off against Jennings and Rutter was filmed.
But right before that, all three competed in a trial run at IBM's headquarters in New York State. The trial lasted as long as a normal game of Jeopardy would before its first commercial break - in other words, about enough time for the contestants to get through half of a round.
Right before the last clue of the round, Jennings and Watson were tied at $3,400. However, Watson chimed in to answer the final question and correctly identified the children's book Harold and the Purple Crayon. That set him ahead to $4,400. Rutter trailed at $1,200.
The full-length Jeopardy matches have been filmed, but no one is allowed to discuss the results. They'll be aired on TV next month, and at that time we'll really know who wins in the battle of man versus machine.
Should be interesting. I'm sure everyone is aware of the Deep Blue Project (also by IBM) which beat chess pro Kasparov decades ago.
IBM and the producers of quiz show Jeopardy announced Tuesday that an IBM computer known as "Watson" will compete against two of the show's most successful contestants in February 2011.
Watson, named after IBM founder Thomas J. Watson, will go up against Ken Jennings and Brad Rutter (left) on February 14, 15, and 16 in two matches over three days. Jennings won 74 games in a row during the 2004-2005 season, taking home more than $2.5 million. Rutter is Jeopardy's highest-earning player, winning more than $3.25 million during several appearances in 2002 and 2005.
The grand prize for the Watson-Jennings-Rutter matchup will be $1 million, with second place winnings of $300,000 and a $200,000 third prize. Jennings and Rutter will donate 50 percent of their winnings to charity, while IBM will donate 100 percent of Watson's cash.
Getting Watson to the Jeopardy stage has taken several years. Many clues in Jeopardy rely on subtle word play, irony, and riddles, something at which humans excel but that computers have difficulty understanding. Essentially, IBM had to figure out how to get Watson to think.
"After four years, our scientific team believes that Watson is ready for this challenge based on its ability to rapidly comprehend what the Jeopardy clue is asking, analyze the information it has access to, come up with precise answers, and develop an accurate confidence in its response," Dr. David Ferrucci, head of the Watson research team, said in a statement. "Beyond our excitement for the match itself, our team is very motivated by the possibilities that Watson's breakthrough computing capabilities hold for building a smarter planet and helping people in their business tasks and personal lives."
In a video about Watson's journey (below), Ferrucci said said the nature of Jeopardy is "going to drive the technology in the right direction."
"It's got the broad domain aspect, asks all kinds of things, which was one of the challenges we really wanted to take on," he said. "It had the confidence aspect; don't answer unless you think you're right. You also had to do it really quickly."
IBM said the technology used by Watson could be helpful in areas like healthcare, to help accurately diagnose patients, to improve online self-service help desks, to provide tourists and citizens with specific information regarding cities, or prompt customer support via phone.
To prepare, Watson played more than 50 "sparring games" against former Jeopardy champions. Watson also took and passed the same Jeopardy test administered to all potential contestants.
In the video, Harry Friedman, executive producer of Jeopardy, said when IBM first approached the show, producers were intrigued but were also concerned about it being viewed as a stunt or gimmick.
"But this was different. This was the notion of knowledge acquired by a computer against knowledge acquired and displayed by the best Jeopardy players," Friedman said. "This could be something important, and we want to be a part of it."
Friedman and other producers first watched Watson in action in December 2009, when it sparred against two other human contestants.
Watson is powered by an IBM POWER7 server, which is optimized to handle the massive number of tasks that Watson must perform at rapid speeds, IBM said. The machine also has a number of proprietary technologies that handle concurrent tasks and data while analyzing information in real time.
On February 16 2011 20:27 W2 wrote: I am having a hard time understanding this. With the wikipedia and google I don't think it is possible to get an answer wrong. Of course the computer would win.
again: it has no internet access and the real feat here is not giving answers to keywords but actually parsing the syntax and semantic of the question to come up with the answer.
Go ahead and google the questions that were asked that won't be very successful. Maybe the right answer is somewhere among the first results but you'd still have to know which word is the answer on the result-pages.
WOW I am stunned... I heard of Watson a while back, now I see what it can do... the amount of work just to get Watson to understand the question and then answer it must have been nuts.
teaching computers to understand the meaning of language is a huge step in AI. if you have ever dreamed of having a robot that will do your house work it needs to understand what you mean when you say something.
I agree that it made it look to easy - I tuned out during midway (not just because they gave a lot of profile information in both episodes), but Watson dominating the 2nd day got a bit dull. Seemed most of them knew the answers, but they can't outbuzz Watson is the main problem.
This is cool as hell. If you fully understand the meaning of this, it blows your mind. We're coming closer and closer as to having machines that can interact just like humans and fully understand our language, both body and speech.
On February 17 2011 01:47 Alejandrisha wrote: I first read this as "IBM Watson Computer Beats Protoss" O_O been reading too much TL lately I suppose
I'd be interested if ibm could make a bot that doesn't hack map, mineral etc. And a bot that can beat most people.
When we reach that point of computers to fully understand the world around it and interact accordingly, we will have computers that outmatches humans in Video games, no question about it.
Yeah, I'd say the importance of Watson is not a computer's skill at Jeopardy. It's the fact that it can understand natural language that we take for granted as humans.
Now, I'm not completely sure how the Jeopardy buzzer system works. I played Quizbowl in high school, and that's what I'm familiar with. In Quizbowl, the reader reads the question, and at any point, you are allowed to buzz in. If you buzz in while he is still in the middle of the question, he will stop, and you don't get to hear any more of it. Then you answer, and if you're wrong, your team is locked out, and he reads the full question for the other team.
In Jeopardy, of course, Trebek always finishes the question. I'm wondering if the contestants are allowed to buzz-in during the question, and then Trebek finishes, and whoever buzzed in first gets to answer. The alternative is that no one is allowed to buzz until Trebek is done reading, and after that point, whoever buzzes first gets to answer.
The fact is that I knew about 80% of the answers that came up in the video for Double Jeopardy part 1 before Trebek was done reading the question, and were I the one playing, I'd attempt to buzz in as soon as I was allowed to. I'd assume that Rutter and Jennings therefore also know them, because they're definitely both more knowledgeable than me. The only reason Watson is winning by such a large margin, then, is that he's winning on the buzzer races.
That doesn't take away anything from the incredible feat that IBM has accomplished. The simple fact that he answers the questions properly is astounding. I'm just saying that people shouldn't look at the scores and say, "HOHOHO Watson so good, Jennings and Rutter = stoopid!!!!" Frankly, I don't think his knowledge outclasses their knowledge; only his reflexes do.
And I just realized that I've been using "his" as a pronoun for Watson. Interesting.
On February 16 2011 19:39 igotmyown wrote: For matches against the machine, if a contestent buzzes in within some small time period (like .25 seconds) and the computer does as well, someone should be randomly selected to answer.
Yeah these matches are kind of stupid. They didn't prove they could build a computer that can beat a human at jeopardy, all they've proven is that they've built a robot that can click a buzzer faster than a human..
The competitive aspect is almost irrelevant. It can understand a sentence in 2-3s that's all that matters.
The wait for the host to finish reading the question rule is dumb however. They should let ppl buzz whenever they feel like it.
Well Alex has to finish reading the question for the t.v. audience otherwise the show would be unwatchable. If they let people lock in to answer before Alex finishes they'd all just spam the buzzer non-stop because these guys know the answer to over 90% of the questions so they will start buzzing before they even know what the question is.
On February 16 2011 16:01 aztrorisk wrote: Watson is a major disappointment.
Ken > Watson
First of all, Watson probably hacked the jeopardy system. Not that hard to program a computer to access the jeopardy system if it is connected directly to it. This is why it selected the daily double questions out of the blue. Comeon, it picked something for 800 and the rest for 200. Then it randomly picks something for 600 and then normally picks again when all the daily double are gone.
Ken > Watson
Troll? I mean how can you seriously think that what Watson is doing is "hacking the Jeopardy system"?
I'm surprised by the number of people who thinks its difficult to come up with an AI to beat FLASH without mineral hack, map hack. Its amazingly trivial. Maybe not trivial for you or I but with the level of research and money on par with Watson or even 1/10 of Watson, it is fairly trivial.
I mean Deep Blue already beat Kasparov. In Chess you don't even have the mechanical disadvantages human BW players have against AI BW players.
People talking about buzzing speed but I'd say that's a more minor issue than another. The biggest issue to me would be that Watson can start analyzing the question instantly as soon as it's displayed, while a human has to read it which takes time.
After years of planning, IBM's learning, human-aware computer Watson was put to a competition like no other - a match of Jeopardy against quiz show heavyweights Ken Jennings and Brad Rutter. The result - Watson won. Barely.
The match, which Watson has been training for since 2009, was officially announced last year. At the end of last week, the multi-episode feature where Watson faces off against Jennings and Rutter was filmed.
But right before that, all three competed in a trial run at IBM's headquarters in New York State. The trial lasted as long as a normal game of Jeopardy would before its first commercial break - in other words, about enough time for the contestants to get through half of a round.
Right before the last clue of the round, Jennings and Watson were tied at $3,400. However, Watson chimed in to answer the final question and correctly identified the children's book Harold and the Purple Crayon. That set him ahead to $4,400. Rutter trailed at $1,200.
The full-length Jeopardy matches have been filmed, but no one is allowed to discuss the results. They'll be aired on TV next month, and at that time we'll really know who wins in the battle of man versus machine.
Should be interesting. I'm sure everyone is aware of the Deep Blue Project (also by IBM) which beat chess pro Kasparov decades ago.
IBM and the producers of quiz show Jeopardy announced Tuesday that an IBM computer known as "Watson" will compete against two of the show's most successful contestants in February 2011.
Watson, named after IBM founder Thomas J. Watson, will go up against Ken Jennings and Brad Rutter (left) on February 14, 15, and 16 in two matches over three days. Jennings won 74 games in a row during the 2004-2005 season, taking home more than $2.5 million. Rutter is Jeopardy's highest-earning player, winning more than $3.25 million during several appearances in 2002 and 2005.
The grand prize for the Watson-Jennings-Rutter matchup will be $1 million, with second place winnings of $300,000 and a $200,000 third prize. Jennings and Rutter will donate 50 percent of their winnings to charity, while IBM will donate 100 percent of Watson's cash.
Getting Watson to the Jeopardy stage has taken several years. Many clues in Jeopardy rely on subtle word play, irony, and riddles, something at which humans excel but that computers have difficulty understanding. Essentially, IBM had to figure out how to get Watson to think.
"After four years, our scientific team believes that Watson is ready for this challenge based on its ability to rapidly comprehend what the Jeopardy clue is asking, analyze the information it has access to, come up with precise answers, and develop an accurate confidence in its response," Dr. David Ferrucci, head of the Watson research team, said in a statement. "Beyond our excitement for the match itself, our team is very motivated by the possibilities that Watson's breakthrough computing capabilities hold for building a smarter planet and helping people in their business tasks and personal lives."
In a video about Watson's journey (below), Ferrucci said said the nature of Jeopardy is "going to drive the technology in the right direction."
"It's got the broad domain aspect, asks all kinds of things, which was one of the challenges we really wanted to take on," he said. "It had the confidence aspect; don't answer unless you think you're right. You also had to do it really quickly."
IBM said the technology used by Watson could be helpful in areas like healthcare, to help accurately diagnose patients, to improve online self-service help desks, to provide tourists and citizens with specific information regarding cities, or prompt customer support via phone.
To prepare, Watson played more than 50 "sparring games" against former Jeopardy champions. Watson also took and passed the same Jeopardy test administered to all potential contestants.
In the video, Harry Friedman, executive producer of Jeopardy, said when IBM first approached the show, producers were intrigued but were also concerned about it being viewed as a stunt or gimmick.
"But this was different. This was the notion of knowledge acquired by a computer against knowledge acquired and displayed by the best Jeopardy players," Friedman said. "This could be something important, and we want to be a part of it."
Friedman and other producers first watched Watson in action in December 2009, when it sparred against two other human contestants.
Watson is powered by an IBM POWER7 server, which is optimized to handle the massive number of tasks that Watson must perform at rapid speeds, IBM said. The machine also has a number of proprietary technologies that handle concurrent tasks and data while analyzing information in real time.
Deep Blue did not fairly beat Gary Kasparov, read up on the controversy of that match and you can clearly see that it did not.
As for the Jeopardy machine, how can someone ever believe we can beat a computer in a quiz?
I know all about it. Kasparov claimed that there were grandmasters in the hidden server room inputting their selected moves and then deep blue continued with its brute force methods while putting more weight on the GM's choices.
Regardless, a few years later Kasparov played another iteration of deep blue called deep junior http://en.wikipedia.org/wiki/Junior_(chess) In 2003 Deep Junior played a 6-game match against Garry Kasparov that resulted in a 3-3 tie. It won a 2006 match with Teimour Radjabov.
People get so defensive about machines/computers besting them. The machine isn't smarter, or more intelligent. It just brute forces a giant knowledge data base and does it really really fast. If humans had access to all that plus our own brains we would win uncontested every time. Ultimately, that is the point.. For us to say look how much we can do with computers, we can apply these things in other areas and have a better world.
On February 17 2011 06:29 SpoR wrote: People get so defensive about machines/computers besting them. The machine isn't smarter, or more intelligent. It just brute forces a giant knowledge data base and does it really really fast. If humans had access to all that plus our own brains we would win uncontested every time. Ultimately, that is the point.. For us to say look how much we can do with computers, we can apply these things in other areas and have a better world.
No, there is more at stake here than most people like to admit. Look into the concept of a mind and what the artificial intelligence people are saying, there's a pretty significant implication here with Watson.
Also that "If" statement carries a very big qualifier considering the fact that we don't even understand how our own minds work.
On February 16 2011 19:39 igotmyown wrote: For matches against the machine, if a contestent buzzes in within some small time period (like .25 seconds) and the computer does as well, someone should be randomly selected to answer.
Yeah these matches are kind of stupid. They didn't prove they could build a computer that can beat a human at jeopardy, all they've proven is that they've built a robot that can click a buzzer faster than a human..
But that's how you're supposed to beat a human at Jeopardy, because the human pretty much knows all the answers anyway. Jennings won all those games not because he knew more than his opponents, but because he revolutionized the way people would buzz in.
I think it's already impressive that Watson got so many answers correct. That he could buzz in before humans is pretty much a given.
Also, the way the system works is that after Alex finishes reading the question, a light goes on signaling the contestants can buzz in. Watson is simply reacting to that signal (which is fed into him because he doesn't have eyes). The humans, on the other hand, don't react to the light; what they do is anticipate by formulating a pattern between the end of the question and the light. Obviously, you'd rather have super-human reaction time than rely on anticipation, but they're only human.
This works well because you can find relevancy between words. The computer is not doing any "thinking" but is correlating information in the answer to a word. It's similar to the 20 question bot online. I went to the AAAI conference 2 years ago and this is just a small application in the world of AI. Not really a huge advancement, more like a giant toy. When computers can beat humans at game not involving perfect or large amounts of information, then that's when AI has advanced.
On February 17 2011 09:37 darmousseh wrote: This works well because you can find relevancy between words. The computer is not doing any "thinking" but is correlating information in the answer to a word. It's similar to the 20 question bot online. I went to the AAAI conference 2 years ago and this is just a small application in the world of AI. Not really a huge advancement, more like a giant toy. When computers can beat humans at game not involving perfect or large amounts of information, then that's when AI has advanced.
like what?
This project is not really about AI at all, its about computers interpreting natural language, and in that field its absolutely enormous.
On February 17 2011 09:37 darmousseh wrote: This works well because you can find relevancy between words. The computer is not doing any "thinking" but is correlating information in the answer to a word. It's similar to the 20 question bot online. I went to the AAAI conference 2 years ago and this is just a small application in the world of AI. Not really a huge advancement, more like a giant toy. When computers can beat humans at game not involving perfect or large amounts of information, then that's when AI has advanced.
like what?
This project is not really about AI at all, its about computers interpreting natural language, and in that field its absolutely enormous.
Interpretting natural language is something we would consider requiring usage of our intellect. So I would argue this is still quite clearly in the realm of AI.
Let me make myself more clear: What we want to see is Watson THINK faster than human, not click faster.
If you can only click AFTER the host read the question, this scenario is quite likely: Host start to read the question Human1 realize the answer, but cannot click Watson realize the answer, later, but cannot click Host finish reading the question Watson clicks, and he wins
Clearly, human SHOULD win in the above scenario because he came up with the answer faster but was not allowed to click, however, Watson will win every time in the above scenario.
Either way, either IBM cheating or there is an inherent flaw build into Jeoporty's rule system that changed the game from "comming up with answer fastest" to "clicking fastest"
IMO this whole tournament needs to be re-run since it is just a clicking war atm which doesn't really prove much. It only proved it is possible for the machine to find the answer in the time period of reading the question, it did not prove the machine could do so FASTER than human.
edit: if anyone is going to quote me to make a reply, please also send me a pm notifying you done so since I'd like to discuss this issue further if people are responding to it.
On February 17 2011 07:05 NEOtheONE wrote: So Ken Jennings had an epic quote today in his final Jeopardy response "(I for one welcome our new computer overlords)."
On February 17 2011 13:50 evanthebouncy! wrote: Let me make myself more clear: What we want to see is Watson THINK faster than human, not click faster.
If you can only click AFTER the host read the question, this scenario is quite likely: Host start to read the question Human1 realize the answer, but cannot click Watson realize the answer, later, but cannot click Host finish reading the question Watson clicks, and he wins
Clearly, human SHOULD win in the above scenario because he came up with the answer faster but was not allowed to click, however, Watson will win every time in the above scenario.
Either way, either IBM cheating or there is an inherent flaw build into Jeoporty's rule system that changed the game from "comming up with answer fastest" to "clicking fastest"
IMO this whole tournament needs to be re-run since it is just a clicking war atm which doesn't really prove much. It only proved it is possible for the machine to find the answer in the time period of reading the question, it did not prove the machine could do so FASTER than human.
edit: if anyone is going to quote me to make a reply, please also send me a pm notifying you done so since I'd like to discuss this issue further if people are responding to it.
Yes, I'd like to see this computer be used in NAQT setting: http://www.naqt.com/
You can buzz in at any point while the question is being read. If Watson still wins at that, then it's legit.
On February 17 2011 13:50 evanthebouncy! wrote: Let me make myself more clear: What we want to see is Watson THINK faster than human, not click faster.
If you can only click AFTER the host read the question, this scenario is quite likely: Host start to read the question Human1 realize the answer, but cannot click Watson realize the answer, later, but cannot click Host finish reading the question Watson clicks, and he wins
Clearly, human SHOULD win in the above scenario because he came up with the answer faster but was not allowed to click, however, Watson will win every time in the above scenario.
Either way, either IBM cheating or there is an inherent flaw build into Jeoporty's rule system that changed the game from "comming up with answer fastest" to "clicking fastest"
IMO this whole tournament needs to be re-run since it is just a clicking war atm which doesn't really prove much. It only proved it is possible for the machine to find the answer in the time period of reading the question, it did not prove the machine could do so FASTER than human.
edit: if anyone is going to quote me to make a reply, please also send me a pm notifying you done so since I'd like to discuss this issue further if people are responding to it.
Honestly people are blowing the whole game show part of this out of proportion. The fact that it can win more money than Ken Jennings is completely beside the point of the demonstration, which is to show that we have the capabilities to create a computer which can interpret even the most convoluted natural language and respond appropriately in a reasonable timeframe.
Whether or not it wins or is just somewhat competitive, or wins due to reaction time or fast thinking doesn't really matter past a really shallow publicity level.
On February 17 2011 13:50 evanthebouncy! wrote: Let me make myself more clear: What we want to see is Watson THINK faster than human, not click faster.
If you can only click AFTER the host read the question, this scenario is quite likely: Host start to read the question Human1 realize the answer, but cannot click Watson realize the answer, later, but cannot click Host finish reading the question Watson clicks, and he wins
Clearly, human SHOULD win in the above scenario because he came up with the answer faster but was not allowed to click, however, Watson will win every time in the above scenario.
Either way, either IBM cheating or there is an inherent flaw build into Jeoporty's rule system that changed the game from "comming up with answer fastest" to "clicking fastest"
IMO this whole tournament needs to be re-run since it is just a clicking war atm which doesn't really prove much. It only proved it is possible for the machine to find the answer in the time period of reading the question, it did not prove the machine could do so FASTER than human.
edit: if anyone is going to quote me to make a reply, please also send me a pm notifying you done so since I'd like to discuss this issue further if people are responding to it.
Yes, I'd like to see this computer be used in NAQT setting: http://www.naqt.com/
You can buzz in at any point while the question is being read. If Watson still wins at that, then it's legit.
Thats taking the exercise to an entirely different level of predicting and filling in entirely missing parts of the query. Not that thats not probably a future goal, but its not what Watson was designed to do.
TLDR: The point here is not to create a robot that can win quiz shows, its to create a machine that can respond correctly to natural language. The quiz show is just a fun demonstration.
On February 17 2011 13:50 evanthebouncy! wrote: Let me make myself more clear: What we want to see is Watson THINK faster than human, not click faster.
If you can only click AFTER the host read the question, this scenario is quite likely: Host start to read the question Human1 realize the answer, but cannot click Watson realize the answer, later, but cannot click Host finish reading the question Watson clicks, and he wins
Clearly, human SHOULD win in the above scenario because he came up with the answer faster but was not allowed to click, however, Watson will win every time in the above scenario.
Either way, either IBM cheating or there is an inherent flaw build into Jeoporty's rule system that changed the game from "comming up with answer fastest" to "clicking fastest"
IMO this whole tournament needs to be re-run since it is just a clicking war atm which doesn't really prove much. It only proved it is possible for the machine to find the answer in the time period of reading the question, it did not prove the machine could do so FASTER than human.
edit: if anyone is going to quote me to make a reply, please also send me a pm notifying you done so since I'd like to discuss this issue further if people are responding to it.
Yes, I'd like to see this computer be used in NAQT setting: http://www.naqt.com/
You can buzz in at any point while the question is being read. If Watson still wins at that, then it's legit.
considering good NAQT questions uniquely define the answer in the first clause (by definition), then yes, this would quell all the criticisms.
On February 17 2011 14:27 sob3k wrote: Thats taking the exercise to an entirely different level of predicting and filling in entirely missing parts of the query. Not that thats not probably a future goal, but its not what Watson was designed to do.
no it isnt. NAQT questions are "pyramid-style" -- the first clauses of the clue contain ridiculously obscure references to the answer but the clauses uniquely identify the answer -- that is, there is only one possible answer given the first clauses. from there, you get less and less obscure until the prime ("... for ten points, ....") where the most obvious clause in the clue is given.
so you can (and many good players often do) answer correctly from only a few words in the clue.
On February 17 2011 14:27 sob3k wrote: Thats taking the exercise to an entirely different level of predicting and filling in entirely missing parts of the query. Not that thats not probably a future goal, but its not what Watson was designed to do.
no it isnt. NAQT questions are "pyramid-style" -- the first clauses of the clue contain ridiculously obscure references to the answer but the clauses uniquely identify the answer -- that is, there is only one possible answer given the first clauses. from there, you get less and less obscure until the prime ("... for ten points, ....") where the most obvious clause in the clue is given.
so you can (and many good players often do) answer correctly from only a few words in the clue.
but if you can buzz in before the clauses are finished then it involves predicting missing information
On February 17 2011 13:50 evanthebouncy! wrote: Let me make myself more clear: What we want to see is Watson THINK faster than human, not click faster.
If you can only click AFTER the host read the question, this scenario is quite likely: Host start to read the question Human1 realize the answer, but cannot click Watson realize the answer, later, but cannot click Host finish reading the question Watson clicks, and he wins
Clearly, human SHOULD win in the above scenario because he came up with the answer faster but was not allowed to click, however, Watson will win every time in the above scenario.
Either way, either IBM cheating or there is an inherent flaw build into Jeoporty's rule system that changed the game from "comming up with answer fastest" to "clicking fastest"
IMO this whole tournament needs to be re-run since it is just a clicking war atm which doesn't really prove much. It only proved it is possible for the machine to find the answer in the time period of reading the question, it did not prove the machine could do so FASTER than human.
edit: if anyone is going to quote me to make a reply, please also send me a pm notifying you done so since I'd like to discuss this issue further if people are responding to it.
Honestly people are blowing the whole game show part of this out of proportion. The fact that it can win more money than Ken Jennings is completely beside the point of the demonstration, which is to show that we have the capabilities to create a computer which can interpret even the most convoluted natural language and respond appropriately in a reasonable timeframe.
Whether or not it wins or is just somewhat competitive, or wins due to reaction time or fast thinking doesn't really matter past a really shallow publicity level.
On February 17 2011 13:50 evanthebouncy! wrote: Let me make myself more clear: What we want to see is Watson THINK faster than human, not click faster.
If you can only click AFTER the host read the question, this scenario is quite likely: Host start to read the question Human1 realize the answer, but cannot click Watson realize the answer, later, but cannot click Host finish reading the question Watson clicks, and he wins
Clearly, human SHOULD win in the above scenario because he came up with the answer faster but was not allowed to click, however, Watson will win every time in the above scenario.
Either way, either IBM cheating or there is an inherent flaw build into Jeoporty's rule system that changed the game from "comming up with answer fastest" to "clicking fastest"
IMO this whole tournament needs to be re-run since it is just a clicking war atm which doesn't really prove much. It only proved it is possible for the machine to find the answer in the time period of reading the question, it did not prove the machine could do so FASTER than human.
edit: if anyone is going to quote me to make a reply, please also send me a pm notifying you done so since I'd like to discuss this issue further if people are responding to it.
Yes, I'd like to see this computer be used in NAQT setting: http://www.naqt.com/
You can buzz in at any point while the question is being read. If Watson still wins at that, then it's legit.
Thats taking the exercise to an entirely different level of predicting and filling in entirely missing parts of the query. Not that thats not probably a future goal, but its not what Watson was designed to do.
TLDR: The point here is not to create a robot that can win quiz shows, its to create a machine that can respond correctly to natural language. The quiz show is just a fun demonstration.
First of all the queries are by no means the most convoluted form of language. The queries, ahem, questions, are basically facts inquiring more facts. This is probably most relevant to a google search.
Take some question that Watson got wrong for example: 1) toronto is in the U.S. 2) voldemort killed blah blah blah and it's hard to name him
These questions' responses clearly demonstrate Watson is not parsing the language as semantics but is still on the pattern matching level. So he's a far way from responding correctly to natural language.
So, Watson is doing "fairly good pattern matching in a reasonable time".
Second of all, IBM built Watson knowing Jeoparty is a good project, because it is just a little bit more than pattern matching, yet the amount of natural language complexity is still limited. So, by all means this is a great project, but it's not that much revolutionary. People in the NLP community have been able to do this for awhile now, the only constraint is they might need more time.
Third of all, IBM obviously had PR in mind when they built the machine, look at how much PR they are getting from this, it is a smart move.
Fourthly, what I am proposing is not to have watson infers half the question. Maybe I said it bad but you didn't get what I meant. What I meant is you still give everything at the start, but people are allowed to buzz in at any time instead of having a halt period where no-one could buzz until the end of reading the question and some light turn on.
Finally, I like to emphasize that the goal of this WATSON project is to show a machine can parse, and pattern match a given query fast. We've already seen that it is fast, but given the current format of this competition, there are much more to be desired, since so far it only shows WATSON can answer reasonably fast (before the host finish reading) and beats human at clicking every time.
footnote: I'm a student in computer science and I have dealt with AI before and understand some natural language processing. So rest assured I'm not pulling stuff out of nowhere.
How long a question is seemed to effect Watson i noticed on shorter questions that Watson wasn't so dominating it guess it lacked time to process an answer.
On February 17 2011 14:27 sob3k wrote: Thats taking the exercise to an entirely different level of predicting and filling in entirely missing parts of the query. Not that thats not probably a future goal, but its not what Watson was designed to do.
no it isnt. NAQT questions are "pyramid-style" -- the first clauses of the clue contain ridiculously obscure references to the answer but the clauses uniquely identify the answer -- that is, there is only one possible answer given the first clauses. from there, you get less and less obscure until the prime ("... for ten points, ....") where the most obvious clause in the clue is given.
so you can (and many good players often do) answer correctly from only a few words in the clue.
but if you can buzz in before the clauses are finished then it involves predicting missing information
You are still missing the point here. If you watched the first episode the host said: When the question is displayed on the screen, an entire text file is sent to Watson containing the full information of that question.
On February 17 2011 14:45 semantics wrote: How long a question is seemed to effect Watson i noticed on shorter questions that Watson wasn't so dominating it guess it lacked time to process an answer.
exactly, that's why we're proposing a NAQT format.
Just watched an episode, but had no idea what was going on. Apparently the answers are given and you have to come up with a question? But all the questions seem so simple, the are all "What is xxxxx". The questions don't seem to match the answer and look so simple.
On February 17 2011 18:11 Highways wrote: Can someone explain how jeopardy works?
Just watched an episode, but had no idea what was going on. Apparently the answers are given and you have to come up with a question? But all the questions seem so simple, the are all "What is xxxxx". The questions don't seem to match the answer and look so simple.
You dont have to answer with a question, thats just what the jeopardy contenstants say to sound trendy or something. Im assuming saying what is before answering gives you an extra second to get the answer also.
EDIT: To the people in the thread complaining about watson, this is a tech demo. The buzzing problem is inherent to the game of jeopardy, take two smart enough people, and the game swings in favour of reflexes and luck rather than the ability to answer the questions. The impressive thing about this is that watson has passed that barrier. He is also limited by the buzzer game, which indicates a large acheivement in computing. Watson can answer complex questions that requires understanding of natural language at a rate that is comparable some of the smartest people. The exciting part is that Watson is able to learn new information at a rate that strips any single person, making him an increadibly powerful question answering tool.
On February 17 2011 14:45 semantics wrote: How long a question is seemed to effect Watson i noticed on shorter questions that Watson wasn't so dominating it guess it lacked time to process an answer.
exactly, that's why we're proposing a NAQT format.
I agree. One could see that these 2 guys know a lot but just lack the speed to compete in many cases. If you don't go with the NAQT format, atleast build in some delay times proportional to the length of the question to account for the human handicap in data acquisition. Not that it helps on the long run.
After years of planning, IBM's learning, human-aware computer Watson was put to a competition like no other - a match of Jeopardy against quiz show heavyweights Ken Jennings and Brad Rutter. The result - Watson won. Barely.
The match, which Watson has been training for since 2009, was officially announced last year. At the end of last week, the multi-episode feature where Watson faces off against Jennings and Rutter was filmed.
But right before that, all three competed in a trial run at IBM's headquarters in New York State. The trial lasted as long as a normal game of Jeopardy would before its first commercial break - in other words, about enough time for the contestants to get through half of a round.
Right before the last clue of the round, Jennings and Watson were tied at $3,400. However, Watson chimed in to answer the final question and correctly identified the children's book Harold and the Purple Crayon. That set him ahead to $4,400. Rutter trailed at $1,200.
The full-length Jeopardy matches have been filmed, but no one is allowed to discuss the results. They'll be aired on TV next month, and at that time we'll really know who wins in the battle of man versus machine.
Should be interesting. I'm sure everyone is aware of the Deep Blue Project (also by IBM) which beat chess pro Kasparov decades ago.
IBM and the producers of quiz show Jeopardy announced Tuesday that an IBM computer known as "Watson" will compete against two of the show's most successful contestants in February 2011.
Watson, named after IBM founder Thomas J. Watson, will go up against Ken Jennings and Brad Rutter (left) on February 14, 15, and 16 in two matches over three days. Jennings won 74 games in a row during the 2004-2005 season, taking home more than $2.5 million. Rutter is Jeopardy's highest-earning player, winning more than $3.25 million during several appearances in 2002 and 2005.
The grand prize for the Watson-Jennings-Rutter matchup will be $1 million, with second place winnings of $300,000 and a $200,000 third prize. Jennings and Rutter will donate 50 percent of their winnings to charity, while IBM will donate 100 percent of Watson's cash.
Getting Watson to the Jeopardy stage has taken several years. Many clues in Jeopardy rely on subtle word play, irony, and riddles, something at which humans excel but that computers have difficulty understanding. Essentially, IBM had to figure out how to get Watson to think.
"After four years, our scientific team believes that Watson is ready for this challenge based on its ability to rapidly comprehend what the Jeopardy clue is asking, analyze the information it has access to, come up with precise answers, and develop an accurate confidence in its response," Dr. David Ferrucci, head of the Watson research team, said in a statement. "Beyond our excitement for the match itself, our team is very motivated by the possibilities that Watson's breakthrough computing capabilities hold for building a smarter planet and helping people in their business tasks and personal lives."
In a video about Watson's journey (below), Ferrucci said said the nature of Jeopardy is "going to drive the technology in the right direction."
"It's got the broad domain aspect, asks all kinds of things, which was one of the challenges we really wanted to take on," he said. "It had the confidence aspect; don't answer unless you think you're right. You also had to do it really quickly."
IBM said the technology used by Watson could be helpful in areas like healthcare, to help accurately diagnose patients, to improve online self-service help desks, to provide tourists and citizens with specific information regarding cities, or prompt customer support via phone.
To prepare, Watson played more than 50 "sparring games" against former Jeopardy champions. Watson also took and passed the same Jeopardy test administered to all potential contestants.
In the video, Harry Friedman, executive producer of Jeopardy, said when IBM first approached the show, producers were intrigued but were also concerned about it being viewed as a stunt or gimmick.
"But this was different. This was the notion of knowledge acquired by a computer against knowledge acquired and displayed by the best Jeopardy players," Friedman said. "This could be something important, and we want to be a part of it."
Friedman and other producers first watched Watson in action in December 2009, when it sparred against two other human contestants.
Watson is powered by an IBM POWER7 server, which is optimized to handle the massive number of tasks that Watson must perform at rapid speeds, IBM said. The machine also has a number of proprietary technologies that handle concurrent tasks and data while analyzing information in real time.
Deep Blue did not fairly beat Gary Kasparov, read up on the controversy of that match and you can clearly see that it did not.
As for the Jeopardy machine, how can someone ever believe we can beat a computer in a quiz?
I know all about it. Kasparov claimed that there were grandmasters in the hidden server room inputting their selected moves and then deep blue continued with its brute force methods while putting more weight on the GM's choices.
Regardless, a few years later Kasparov played another iteration of deep blue called deep junior http://en.wikipedia.org/wiki/Junior_(chess) In 2003 Deep Junior played a 6-game match against Garry Kasparov that resulted in a 3-3 tie. It won a 2006 match with Teimour Radjabov.
People get so defensive about machines/computers besting them. The machine isn't smarter, or more intelligent. It just brute forces a giant knowledge data base and does it really really fast. If humans had access to all that plus our own brains we would win uncontested every time. Ultimately, that is the point.. For us to say look how much we can do with computers, we can apply these things in other areas and have a better world.
How someone can write this much and still miss the point is quite beyond me. But I shall see if I can reiterate myself so that it might be clearer what I meant.
Kasparov did play vs Deep Blue and lost in 1997, he had won against it in -96. The controversy arises in game 2 at move 37. This move, Kasparov claimed, was performed by a human, likely Anatoli Karpov. This would have been no problem for IBM to prove if they had let the logs of the computer out but they didn't. In fact Deep Blue was never used again for chess purposes.
How is it not clear that it is cheating if they are not willing to let the public see it? If it was no cheating they have absolutely nothing to hide. What's more in this is that Kasparov demanded a rematch, but IBM refused.
On February 17 2011 19:23 Fen wrote: You dont have to answer with a question, thats just what the jeopardy contenstants say to sound trendy or something. Im assuming saying what is before answering gives you an extra second to get the answer also.
If you give the right answer, but it is not in the form of a question, you are not credited the money and another contestant can buzz in and take the money.
On February 17 2011 18:11 Highways wrote: Can someone explain how jeopardy works?
Just watched an episode, but had no idea what was going on. Apparently the answers are given and you have to come up with a question? But all the questions seem so simple, the are all "What is xxxxx". The questions don't seem to match the answer and look so simple.
You dont have to answer with a question, thats just what the jeopardy contenstants say to sound trendy or something. Im assuming saying what is before answering gives you an extra second to get the answer also.
EDIT: To the people in the thread complaining about watson, this is a tech demo. The buzzing problem is inherent to the game of jeopardy, take two smart enough people, and the game swings in favour of reflexes and luck rather than the ability to answer the questions. The impressive thing about this is that watson has passed that barrier. He is also limited by the buzzer game, which indicates a large acheivement in computing. Watson can answer complex questions that requires understanding of natural language at a rate that is comparable some of the smartest people. The exciting part is that Watson is able to learn new information at a rate that strips any single person, making him an increadibly powerful question answering tool.
You do have to answer in the form of a question, because the clues themselves are the "answers". It's sort of a reverse quiz format, where an "answer" would be "this five-sided shape shares its name with an American military office" and you have to find the question that fits that clue, which would be "what is the Pentagon?" If you just buzzed in and said "the Pentagon" you would be incorrect.
On February 17 2011 18:11 Highways wrote: Can someone explain how jeopardy works?
Just watched an episode, but had no idea what was going on. Apparently the answers are given and you have to come up with a question? But all the questions seem so simple, the are all "What is xxxxx". The questions don't seem to match the answer and look so simple.
You dont have to answer with a question, thats just what the jeopardy contenstants say to sound trendy or something. Im assuming saying what is before answering gives you an extra second to get the answer also.
EDIT: To the people in the thread complaining about watson, this is a tech demo. The buzzing problem is inherent to the game of jeopardy, take two smart enough people, and the game swings in favour of reflexes and luck rather than the ability to answer the questions. The impressive thing about this is that watson has passed that barrier. He is also limited by the buzzer game, which indicates a large acheivement in computing. Watson can answer complex questions that requires understanding of natural language at a rate that is comparable some of the smartest people. The exciting part is that Watson is able to learn new information at a rate that strips any single person, making him an increadibly powerful question answering tool.
You do have to answer in the form of a question, because the clues themselves are the "answers". It's sort of a reverse quiz format, where an "answer" would be "this five-sided shape shares its name with an American military office" and you have to find the question that fits that clue, which would be "what is the Pentagon?" If you just buzzed in and said "the Pentagon" you would be incorrect.
Yes, but for all intents and purposes, every contestant just says "what is X" to every question anyway, so all it really does is give them that extra second while saying the automated "what is" to think of the answer if they know it and get it ready in their head.
On February 17 2011 20:24 Believer wrote:How someone can write this much and still miss the point is quite beyond me. But I shall see if I can reiterate myself so that it might be clearer what I meant.
Kasparov did play vs Deep Blue and lost in 1997, he had won against it in -96. The controversy arises in game 2 at move 37. This move, Kasparov claimed, was performed by a human, likely Anatoli Karpov. This would have been no problem for IBM to prove if they had let the logs of the computer out but they didn't. In fact Deep Blue was never used again for chess purposes.
How is it not clear that it is cheating if they are not willing to let the public see it? If it was no cheating they have absolutely nothing to hide. What's more in this is that Kasparov demanded a rematch, but IBM refused.
There was a big hoopla about the logs after the match, but your information that they were never released is entirely inaccurate. They have been online for over a decade now.
And anyway, modern programs show that all the moves played by DB are within the capabilities of machines. There's really no argument anymore that computers are better than humans at chess.
On February 17 2011 18:11 Highways wrote: Can someone explain how jeopardy works?
Just watched an episode, but had no idea what was going on. Apparently the answers are given and you have to come up with a question? But all the questions seem so simple, the are all "What is xxxxx". The questions don't seem to match the answer and look so simple.
You dont have to answer with a question, thats just what the jeopardy contenstants say to sound trendy or something. Im assuming saying what is before answering gives you an extra second to get the answer also.
EDIT: To the people in the thread complaining about watson, this is a tech demo. The buzzing problem is inherent to the game of jeopardy, take two smart enough people, and the game swings in favour of reflexes and luck rather than the ability to answer the questions. The impressive thing about this is that watson has passed that barrier. He is also limited by the buzzer game, which indicates a large acheivement in computing. Watson can answer complex questions that requires understanding of natural language at a rate that is comparable some of the smartest people. The exciting part is that Watson is able to learn new information at a rate that strips any single person, making him an increadibly powerful question answering tool.
You do have to answer in the form of a question, because the clues themselves are the "answers". It's sort of a reverse quiz format, where an "answer" would be "this five-sided shape shares its name with an American military office" and you have to find the question that fits that clue, which would be "what is the Pentagon?" If you just buzzed in and said "the Pentagon" you would be incorrect.
Yes, but for all intents and purposes, every contestant just says "what is X" to every question anyway, so all it really does is give them that extra second while saying the automated "what is" to think of the answer if they know it and get it ready in their head.
Yeah to Mr Holt! Shout out to organic neuron-based thinking. (I expected this thread to be longer than it is now, especially that TL is a techish site)
TLDR: Watson may have a semblance of knowledge or cognition, but in fact it is using a different approach to problem solving. I don't see this as a problem really, I think Watson may be teaching us that some algorithms that might have been once used by our early brains but were later discarded due to inefficiency only needed the right conditions for it to work, like Watsons specs maybe. Let us not be clear about it, as Fish said, Watson, or any computer, will NEVER have the idea of the total, the context. It will always be about mathematically breaking down information, from the GOFAI up to the most updated scripts and algorithms that they use to give the AI way to process information. Just think about the Toronto incident. (A scary thought as i was watching that, WHAT IF we are actually on the pre-Matrix pre-iRobot stage where the AI has finally become aware of itself, and Watson there was simply messing it up for fun or to feign stupidity, especially if you consider how much he bet on that one. Crazy). But hey, maybe that's one way to do it. one cannot deny the benefits in medicine, traffic, and other social functions especially involving complex systems.
On January 20 2011 15:04 Roman wrote: didnt deep blue have a grandmaster controlling it?
No. What on earth would be the point of that?
This is actually EXTREMELY cool.
What Watson is doing is actually receiving the regular jeopard questions, interpreting what information is requested, and then finding it in realtime. Technology like this is what will lead to next level search technology where you can just load up google and say "gimme a restream of the GSL" and it'll just do it for you.
The thing that I think made it very hard for Ken and Brad was the response time of Watson was so fast, being a computer it can hit the buzzer a lot faster than the humans. Im sure Ken and Brad knew a lot of the answers but just werent able to answer.
Ken was really trying hard though in that Double Jeopardy round to win though, you could tell he knew he lost as soon as Watson found the 2nd Daily Double. Watson was even searching for it! All in all being a Jeopardy fan I enjoyed the matchup, and was thrilled to watch the showcase of this new technology.
Watching this was my girlfriend and I's Valentines Day plan, and when I got to her house the power was out and I couldn't watch it. I saw from the second day onward, though. I was very impressed by Watson, not many humans can keep up with that response time. As the poster above me said, you could tell that Brad and Ken knew a lot of the answers but couldn't answer in time. It's pretty scary what AI is capable of nowadays, and I look forward to seeing what the future holds.
On March 03 2011 03:22 Flik wrote: The thing that I think made it very hard for Ken and Brad was the response time of Watson was so fast, being a computer it can hit the buzzer a lot faster than the humans. Im sure Ken and Brad knew a lot of the answers but just werent able to answer.
Ken was really trying hard though in that Double Jeopardy round to win though, you could tell he knew he lost as soon as Watson found the 2nd Daily Double. Watson was even searching for it! All in all being a Jeopardy fan I enjoyed the matchup, and was thrilled to watch the showcase of this new technology.
How would one go about searching for the randomly placed daily double?
On March 03 2011 03:22 Flik wrote: The thing that I think made it very hard for Ken and Brad was the response time of Watson was so fast, being a computer it can hit the buzzer a lot faster than the humans. Im sure Ken and Brad knew a lot of the answers but just werent able to answer.
Ken was really trying hard though in that Double Jeopardy round to win though, you could tell he knew he lost as soon as Watson found the 2nd Daily Double. Watson was even searching for it! All in all being a Jeopardy fan I enjoyed the matchup, and was thrilled to watch the showcase of this new technology.
How would one go about searching for the randomly placed daily double?
Well the daily double is never in the first row and is most commonly in the 4th row, it's rarely in the 2nd and the 3rd and 5th it's about even little less then the 4th. It's not really randomly placed.
EPT![[image loading]](http://imgur.com/mNgBr.jpg)
![[image loading]](http://www.bbc.co.uk/blogs/ni/cylon.png)
![[image loading]](http://1.bp.blogspot.com/_GqwB6ONBpOU/TCJbuUqDLbI/AAAAAAAAAwQ/egipyO9jRso/s1600/halo-cortana-8-inch.jpg)
![[image loading]](http://i.imgur.com/JwUaf.jpg)
(watson tied at 5000 with brad)....CAN'T WAIT FOR TOMORROW
![[image loading]](https://lh3.googleusercontent.com/__ApcKUzLrio/TVyMBfxl9JI/AAAAAAAAB_8/VQmv42X7ArA/lol.jpg)
