AfterDawn: Tech news

Watson beats puny humans in last day of 'Jeopardy' event

Written by Andre Yoskowitz (Google+) @ 16 Feb 2011 21:11 User comments (21)

Watson beats puny humans in last day of 'Jeopardy' event The IBM supercomputer 'Watson' has finished handily defeating its human opponents on Jeopardy today, beating former champions Ken Jennings and Bruce Rutter.
At the end of the three-day special event, Watson finished with $77,147 to Jenning's $24,000 and Rutter's $21,600.

After the first day of the three-day event, Watson was tied with Rutter at $5000 each, with Jennings lagging behind at $2000. At the end of day 2, however, Watson had taken a commanding lead, with $35,734 to Rutter's $10,400 and Jennings' $4,800.

IBM will receive $1 million, which it is donating to World Vision. Second place gets $300,000 and third gets $200,000. Jennings and Rutter will donate half of their winnings.

Watson is able to calculate thousands of algorithms at the same time, while searching its massive database for the right answer. The behemoth computer runs on ninety 32-core IBM Power 750 Express servers and has 16TB of memory.

Previous Next  

21 user comments

116.2.2011 21:51

I'm not surprised at all. I think it was pointless of them event challenging the machine in the first place.

216.2.2011 22:47

The point was not the knowledge, but the understanding. Having the correct answer does you no good if you cannot understand the question.

Think of the possibilities here...we could have actual "Answer Engine"s instead of search engines. Type in a question and get the answer; not a list of websites that might have an answer; just the actual answer, immediately...for almost any subject, with almost any wording (including synonym searching). Actually, forget typing...it seems that the voice recognition works pretty well on Watson. I can't wait for Watson's children to hit the cloud.



317.2.2011 0:01

Ether way it breaks down to memory management skill, to have that knowledge and to get it quickly, yes it can compute a question as an answer its merely just a different way to look up something. Its not that huge of a deal.

417.2.2011 0:30

That is a bit like saying that a $50,000,000,000 automotive plant is just another way of building something and that it is essentially the same as a home workshop.



517.2.2011 1:12

Not really, the main difference is the search query, instead of seeking a result from a word or 2 it query's a larger word grouping and the answers the question with the best result.

I am pretty sure you can script something similar easy enough and have it run on any computer, speed will vary due to the ability of the script to find the best result and the power of the processing power of the computer itself.

Now mind you I am a hack but it seems like its just searching for something with a few more words, taking into account the sentence or word grouping as a whole rather than the individual keywords that everything else is built on.

/*brain smoking*

LOL


Copyright infringement is nothing more than civil disobedience to a bad set of laws. Lets renegotiate them.

---
Check out my crappy creations
http://zippydsmlee.deviantart.com/

617.2.2011 4:10

What search engine works like this? Do you know of any search engine that gives you an answer rather than a list of websites that may or may not have the answer? I know there are a few sites like Yahoo Answers, but those don't search other sites; they just search their own databases and they still usually can't find an answer. I would love to see a google where you can get the answer to the question, rather than 10,000,000 pages of crap with the answer hiding somewhere towards the bottom of page 5,385,242.



717.2.2011 7:39

Search engines don't offer answers because they don't look for answer they use web pages that use keywords that hopefully have what your after.

There are engines that do what your talking about they are mainly on mobiles these days where you enter an answer and it looks for direct answers.

The music search programs do the exact same thing.

817.2.2011 8:03

I think it is great achievement. I can't believe people compare Watson to Google, when in fact is more like Iron Man's Jarvis...


Quote:
"IBM researchers spent four years building Watson. The machine is capable of processing 80 trillion operations (teraflops) per second."
I am certain you'd be able to "script something similar easy enough and have it run on any computer", Zippy.

917.2.2011 11:16

Originally posted by cyprusrom:
I think it is great achievement. I can't believe people compare Watson to Google, when in fact is more like Iron Man's Jarvis...


Quote:
"IBM researchers spent four years building Watson. The machine is capable of processing 80 trillion operations (teraflops) per second."
I am certain you'd be able to "script something similar easy enough and have it run on any computer", Zippy.
when you take the long road to solve a problem you wind up doing to much.

Copyright infringement is nothing more than civil disobedience to a bad set of laws. Lets renegotiate them.

---
Check out my crappy creations
http://zippydsmlee.deviantart.com/

1017.2.2011 11:22

Originally posted by KillerBug:
What search engine works like this? Do you know of any search engine that gives you an answer rather than a list of websites that may or may not have the answer? I know there are a few sites like Yahoo Answers, but those don't search other sites; they just search their own databases and they still usually can't find an answer. I would love to see a google where you can get the answer to the question, rather than 10,000,000 pages of crap with the answer hiding somewhere towards the bottom of page 5,385,242.
As I said current search engines use key word searches and not moire broader terms like questions and answers.

Copyright infringement is nothing more than civil disobedience to a bad set of laws. Lets renegotiate them.

---
Check out my crappy creations
http://zippydsmlee.deviantart.com/

1117.2.2011 12:01

I'm actually rather amazed by it.

As a way of developing a system that has an intuitive grasp of language nuance for communication it's impressive. Still, I'm not keen on a world where answers to complex questions are spoon fed me.

In some ways Zip isn't far wrong. Goggle seems highly intuitive at times. Often I don't need to go past the search page's example text to the site for the answer to a simple question a la Jeopardy.


Its a lot easier being righteous than right.


DSE VZ300-
Zilog Z80 CPU, 32KB RAM (16K+16K cartridge), video processor 6847, 2KB video RAM, 16 colours (text mode), 5.25" FDD

1217.2.2011 12:16

Originally posted by Jemborg:
I'm actually rather amazed by it.

As a way of developing a system that has an intuitive grasp of language nuance for communication it's impressive. Still, I'm not keen on a world where answers to complex questions are spoon fed me.

In some ways Zip isn't far wrong. Goggle seems highly intuitive at times. Often I don't need to go past the search page's example text to the site for the answer to a simple question a la Jeopardy.
Ya what its doing is dealing with grammar/structure rather than key words, I think if you built a database and used script to look through it using a non key word phraseing/queying/whatever.

Dose anyone know if its looking through a database of some kind or is it really looking at words in a sentence structure?

Copyright infringement is nothing more than civil disobedience to a bad set of laws. Lets renegotiate them.

---
Check out my crappy creations
http://zippydsmlee.deviantart.com/

1317.2.2011 18:00

The amazing thing about this is that it can understand jeopardy style questions and get the right answer. It is not just key words because the questions use many puns, innuendos, vague references, plays on words, etc. It processes all this, finds the real question, searches for the answer, finds it, and buzzes in. All in a matter of a few seconds or less.


I wanna devise a virus and bring dire straights to your environments. Crush your corporations with a mild touch, trash the whole computer systems and revert you to papyrus - Deltron 3030

1417.2.2011 18:47

Originally posted by klassic:
The amazing thing about this is that it can understand jeopardy style questions and get the right answer. It is not just key words because the questions use many puns, innuendos, vague references, plays on words, etc. It processes all this, finds the real question, searches for the answer, finds it, and buzzes in. All in a matter of a few seconds or less.
its a word processor doing word gruope processing unlike normal search stuff which dose key word processing.

Copyright infringement is nothing more than civil disobedience to a bad set of laws. Lets renegotiate them.

---
Check out my crappy creations
http://zippydsmlee.deviantart.com/

1518.2.2011 0:18

This is pretty impressive. Probably the first time we've actually seen an almost real AI, not the type of "AI" we see nowadays in gaming, automated messages, etc. All he needs now are feelings and a personality. Let's put Watson brain in a robot! But first we gotta cut down on the component size.

1618.2.2011 12:11

Originally posted by ZippyDSM:
Originally posted by cyprusrom:
I think it is great achievement. I can't believe people compare Watson to Google, when in fact is more like Iron Man's Jarvis...


Quote:
"IBM researchers spent four years building Watson. The machine is capable of processing 80 trillion operations (teraflops) per second."
I am certain you'd be able to "script something similar easy enough and have it run on any computer", Zippy.
when you take the long road to solve a problem you wind up doing to much.
LOL, but you what? Most of our problems are created by taking short cuts :)
It's not about taking the long road, it's about the experiences learned while trudging on that road.

Peace!

1718.2.2011 12:50

Machine against machine. Have all the contestants be computers with the three major operating systems: Windows, Apple, and Linux.

1818.2.2011 13:50

Originally posted by Gnawnivek:
Originally posted by ZippyDSM:
Originally posted by cyprusrom:
I think it is great achievement. I can't believe people compare Watson to Google, when in fact is more like Iron Man's Jarvis...


Quote:
"IBM researchers spent four years building Watson. The machine is capable of processing 80 trillion operations (teraflops) per second."
I am certain you'd be able to "script something similar easy enough and have it run on any computer", Zippy.
when you take the long road to solve a problem you wind up doing to much.
LOL, but you what? Most of our problems are created by taking short cuts :)
It's not about taking the long road, it's about the experiences learned while trudging on that road.

True true.


Copyright infringement is nothing more than civil disobedience to a bad set of laws. Lets renegotiate them.

---
Check out my crappy creations
http://zippydsmlee.deviantart.com/

1918.2.2011 18:46

Originally posted by ZippyDSM:
Originally posted by Jemborg:
I'm actually rather amazed by it.

As a way of developing a system that has an intuitive grasp of language nuance for communication it's impressive. Still, I'm not keen on a world where answers to complex questions are spoon fed me.

In some ways Zip isn't far wrong. Goggle seems highly intuitive at times. Often I don't need to go past the search page's example text to the site for the answer to a simple question a la Jeopardy.
Ya what its doing is dealing with grammar/structure rather than key words, I think if you built a database and used script to look through it using a non key word phraseing/queying/whatever.

Dose anyone know if its looking through a database of some kind or is it really looking at words in a sentence structure?
There is more to this than meets the eye. I been thinking about this a bit. Sure it uses a database... it may even use Google's itself (or at least Wikipedia). Think... google's readymade database is peer popularity rated. That counts for a lot especially language idiosyncrasy verses rules. The machine's database is crucial and 4 years in not enough to build one from scratch. Google or Wikipedia's DB can be re-cashed and reindexed according to their own algorithms.

I tried a Jeopardy type search on Google... "He was Hitler's architect". Top instant answer: "Albert Speer" - Wikipedia, the free encyclopaedia. So rewrite to, "Who was Albert Speer". This experiment demonstrates that it's maybe not so outrageous.

I am not discounting an "understanding" of the question but without a cogent database it would be rooted (an Aussie expression for f*cked).

I sure would like to see the shows themselves.

Originally posted by mike.m:
This is pretty impressive. Probably the first time we've actually seen an almost real AI, not the type of "AI" we see nowadays in gaming, automated messages, etc. All he needs now are feelings and a personality. Let's put Watson brain in a robot! But first we gotta cut down on the component size.
Cute. Still intelligence is seen as a pure Cartesian entity in itself. If you just have enough of it you will get self-awareness. Intelligence is built of many evolved parts and conscience is another processing method. Machines with AI will be what we make them to be. If we want them to have personality and feelings we will have to build that into them. Rather than rebel they will be high-functioning autistic idiot savants cheerfully telling you what day of the week you were born on. Being enslaved is a human issue. After all, we are not made we are born.


Originally posted by Gnawnivek:
LOL, but you what? Most of our problems are created by taking short cuts :)
Indeed, but taking shortcuts is what characterises human thinking. It is at once our bane and our greatness. We would not be successful without them. Yet mental shortcuts are what is considered behind the prevalence of religiosity in our species. In this case a tendency to "find" (invent) meaning in everything and anything. That, and a leaning towards teleological Cartesian dualism. Oh yeah, and vanity.

(Crikey, who would want to build that sort of AI into a robot?)

Its a lot easier being righteous than right.


DSE VZ300-
Zilog Z80 CPU, 32KB RAM (16K+16K cartridge), video processor 6847, 2KB video RAM, 16 colours (text mode), 5.25" FDD

2018.2.2011 21:15

Originally posted by Jemborg:
Originally posted by ZippyDSM:
Originally posted by Jemborg:
I'm actually rather amazed by it.

As a way of developing a system that has an intuitive grasp of language nuance for communication it's impressive. Still, I'm not keen on a world where answers to complex questions are spoon fed me.

In some ways Zip isn't far wrong. Goggle seems highly intuitive at times. Often I don't need to go past the search page's example text to the site for the answer to a simple question a la Jeopardy.
Ya what its doing is dealing with grammar/structure rather than key words, I think if you built a database and used script to look through it using a non key word phraseing/queying/whatever.

Dose anyone know if its looking through a database of some kind or is it really looking at words in a sentence structure?
There is more to this than meets the eye. I been thinking about this a bit. Sure it uses a database... it may even use Google's itself (or at least Wikipedia). Think... google's readymade database is peer popularity rated. That counts for a lot especially language idiosyncrasy verses rules. The machine's database is crucial and 4 years in not enough to build one from scratch. Google or Wikipedia's DB can be re-cashed and reindexed according to their own algorithms.

I tried a Jeopardy type search on Google... "He was Hitler's architect". Top instant answer: "Albert Speer" - Wikipedia, the free encyclopaedia. So rewrite to, "Who was Albert Speer". This experiment demonstrates that it's maybe not so outrageous.

I am not discounting an "understanding" of the question but without a cogent database it would be rooted (an Aussie expression for f*cked).

I sure would like to see the shows themselves.

Originally posted by mike.m:
This is pretty impressive. Probably the first time we've actually seen an almost real AI, not the type of "AI" we see nowadays in gaming, automated messages, etc. All he needs now are feelings and a personality. Let's put Watson brain in a robot! But first we gotta cut down on the component size.
Cute. Still intelligence is seen as a pure Cartesian entity in itself. If you just have enough of it you will get self-awareness. Intelligence is built of many evolved parts and conscience is another processing method. Machines with AI will be what we make them to be. If we want them to have personality and feelings we will have to build that into them. Rather than rebel they will be high-functioning autistic idiot savants cheerfully telling you what day of the week you were born on. Being enslaved is a human issue. After all, we are not made we are born.


Originally posted by Gnawnivek:
LOL, but you what? Most of our problems are created by taking short cuts :)
Indeed, but taking shortcuts is what characterises human thinking. It is at once our bane and our greatness. We would not be successful without them. Yet mental shortcuts are what is considered behind the prevalence of religiosity in our species. In this case a tendency to "find" (invent) meaning in everything and anything. That, and a leaning towards teleological Cartesian dualism. Oh yeah, and vanity.

(Crikey, who would want to build that sort of AI into a robot?)
Looks like someone really got into this way too much. People just take things way too seriously. You must have a lot of time on your hands to feel the need to criticize everyone's harmless comments. "Cute"? Haha, yes you're so awesome.
This message has been edited since its posting. Latest edit was made on 18 Feb 2011 @ 21:20

2119.2.2011 7:58

Chill out mike,m... I'm just having fun. :)


And that was cute... make mine a robo-sex bomb! :P

This message has been edited since its posting. Latest edit was made on 19 Feb 2011 @ 8:41

Its a lot easier being righteous than right.


DSE VZ300-
Zilog Z80 CPU, 32KB RAM (16K+16K cartridge), video processor 6847, 2KB video RAM, 16 colours (text mode), 5.25" FDD

Comments have been disabled for this article.

News archive