Next Question
Email to a friend |
RSS
No Best Answer Selected, Tip Refunded
1 answerers thought this was unfair.
An AI solution to solve IQ tests would just be a further extension of the above. You set out to solve a problem and teach your AI bot to solve it, in this case it would be the IQ test. You will notice your article you liked to in the reply says they have AI which can solve part of an IQ test, which isn't the whole thing. I think what you want to know is if someone has created AI which could take an IQ test without knowing what it is and do well. The answer is sadly no.
Solving specific problems like your linked article using AI techniques isn't particularly difficult. Solving larger general problems such as doing the whole test is where it gets hard. Personally I think there will need to be a revolution in the way computers are built (perhaps a mix of organic and silicon) before we will see any convincing AI, but that wont come about until we really understand how the brain works. Much AI comes about by attempting to mimic the brain but how can you do that without knowing how it works?
Just recently I finished my Thesis, where I used AI to solve the general problem of pulling pixel text out of images and converting it to ASCII text. During my readings I looked at a lot of AI papers and nothing came even close to doing what you have asked. Within narrow scopes AI can work VERY well and solve very difficult problems, but when given general problems it tends to fall over.
Permalink | Report
e.g. real time translation / transcription of human language or guiding an engineer to diagnose a machine fault.
I think what you are asking is has an intelligent computer program ever been created whose intelligence can be tested. I think the answer is no but many people are working on creating 'machine intelligence' e.g. by trying to create computers that simulate the human brain:
http://news.nationalgeographic.com/news/2005/07/0720_050720_bluebrain.html
http://www.guardian.co.uk/technology/2007/dec/20/research.it
http://news.cnet.co.uk/desktops/0,39029662,39189792,00.htm
http://www.lanl.gov/news/index.php/fuseaction/home.story/story_id/13602
Permalink | Report
Question
M$1
December 22, 2008 07:12 AM
Have any artificial intelligences taken any IQ tests in their entirety and scored well?
I know about the one that did part of a WAIS related to duplicating block configurations.. but that was only part of the test.. most tests have some kind of verbal section too with analogies or some such..
Also, I know what a Turing test is and why most consider that to be a measure of strong AI. I'm just trying to find out if any researchers have managed to handle all sections of a common standardized IQ test.. I accept the Turing test as the best measure of strong or generalized AI. (Although I am also NOT asking for your opinions as to the best measures of this). (I actually believe that managing even verbal sections of an IQ test to be demonstrating a form of narrow AI.)
Anyway DO NOT bother explaining what a Turing test is or discussing related issues. Please just answer the question.
Also, I know what a Turing test is and why most consider that to be a measure of strong AI. I'm just trying to find out if any researchers have managed to handle all sections of a common standardized IQ test.. I accept the Turing test as the best measure of strong or generalized AI. (Although I am also NOT asking for your opinions as to the best measures of this). (I actually believe that managing even verbal sections of an IQ test to be demonstrating a form of narrow AI.)
Anyway DO NOT bother explaining what a Turing test is or discussing related issues. Please just answer the question.
Interesting Question?
Yes (0)
No (0)
- In Science & Mathematics |
- |
- Report |
-
Share
RSS
No Best Answer Selected, Tip Refunded
1 answerers thought this was unfair.
Answers (4)
December 22, 2008 11:33 AM
Not to my knowledge. The thing about AI as it currently stands is that you build AI to solve a specific problem. An AI solution to solve IQ tests would just be a further extension of the above. You set out to solve a problem and teach your AI bot to solve it, in this case it would be the IQ test. You will notice your article you liked to in the reply says they have AI which can solve part of an IQ test, which isn't the whole thing. I think what you want to know is if someone has created AI which could take an IQ test without knowing what it is and do well. The answer is sadly no.
Solving specific problems like your linked article using AI techniques isn't particularly difficult. Solving larger general problems such as doing the whole test is where it gets hard. Personally I think there will need to be a revolution in the way computers are built (perhaps a mix of organic and silicon) before we will see any convincing AI, but that wont come about until we really understand how the brain works. Much AI comes about by attempting to mimic the brain but how can you do that without knowing how it works?
Just recently I finished my Thesis, where I used AI to solve the general problem of pulling pixel text out of images and converting it to ASCII text. During my readings I looked at a lot of AI papers and nothing came even close to doing what you have asked. Within narrow scopes AI can work VERY well and solve very difficult problems, but when given general problems it tends to fall over.
It was unfair to choose no best answer
Not only were there several answers many of them provided evidence.
Not only were there several answers many of them provided evidence.
Permalink | Report
Reply
December 22, 2008 11:47 AM
Yep, AI is just a special type of computer programming that uses algorithms and knowledge models to solve problems of a specific type. e.g. real time translation / transcription of human language or guiding an engineer to diagnose a machine fault.
I think what you are asking is has an intelligent computer program ever been created whose intelligence can be tested. I think the answer is no but many people are working on creating 'machine intelligence' e.g. by trying to create computers that simulate the human brain:
http://news.nationalgeographic.com/news/2005/07/0720_050720_bluebrain.html
http://www.guardian.co.uk/technology/2007/dec/20/research.it
http://news.cnet.co.uk/desktops/0,39029662,39189792,00.htm
http://www.lanl.gov/news/index.php/fuseaction/home.story/story_id/13602
Permalink | Report
Reply
December 22, 2008 12:49 PM
Thanks, I know what AI is and I'm sorry but that's not what I asked. In fact we can and do test computer programs for intelligence routinely using the Turing test and so far (as per public knowledge?) have not detected any 'strong' or 'general' intelligence (but that does not stop people from their computer programs for intelligence). And in the other sense, in my question I described a computer program that was in fact being subjected to part of an IQ test, so that would be one 'whose intelligence ' _(g)_ ' can be tested'.
I specifically asked you not to discuss general issues related to machine intelligence in your answer because I do not want or need any such general information as I am already well informed.
Report
I specifically asked you not to discuss general issues related to machine intelligence in your answer because I do not want or need any such general information as I am already well informed.
December 23, 2008 12:30 AM
Thats not true. Visual problems are actually where most AI techniques have been applied sucessfully and commercially. Things like postcode/zipcode readers use AI.
Report
Answer this Question
Related Questions
If one of your children were diagnosed with childhood diabetes and recommended to abs...
is there a web sight that will help you keep track of books you have read and suggest...
My previous employer did not pay into my SEP a small business version of a 401K. He ...
What percent of scientists are pro and con on climate change?
is there a web sight that will help you keep track of books you have read and suggest...
My previous employer did not pay into my SEP a small business version of a 401K. He ...
What percent of scientists are pro and con on climate change?
Ask a Question
Buy Mahalo Dollars with Credit Card or PayPal
Top Members
Most Popular Tags
Categories
- Anonymous
-
Arts & Design
-
Beauty & Style
-
Books & Authors
-
Business
-
Cars & Transportation
-
Consumer Electronics
- Coupons Deals
-
Education
-
Entertainment
-
Environment
-
Fitness
-
Food & Drink
- From Email
- From Iphone
- From Twitter
-
Health
-
History
-
Hobbies
-
Home & Garden
-
How Tos
-
Humor
-
Jobs
-
Legal
-
Local
-
Love & Relationships
-
Mahalo Answers Community
-
Money
-
Music
-
News
-
NSFW
-
Parenting
-
Pets
-
Science & Mathematics
-
Services
-
Shopping
-
Social Science
-
Society & Culture
-
Sports
-
Technology & Internet
-
Travel
-
Video Games
Welcome New Members
- marybethlittle, December 11, 2009 06:53 AM
- mblittle, December 11, 2009 06:52 AM
- richardmustered, December 11, 2009 06:42 AM
- mohamedjafeer, December 11, 2009 06:34 AM
- julan, December 11, 2009 06:34 AM
Mahalo Dollars are the currency of Mahalo Answers.
Each Mahalo Dollar costs $1.
Once you earn more than 40 Mahalo Dollars, you can request to be paid via PayPal. Each Mahalo Dollar is currently worth $0.75 when paid out via PayPal. Learn More
I am NOT looking for an AI that could a just approach an arbitrary IQ test and complete each type of task. If that were the case then either you would have solved several narrow AI tasks and built a system that could recognize which to apply or you would have (more likely) actually achieved strong AI as you suppose. I am quite aware of the situation with strong AI and certainly not expecting anyone to make a breakthrough announcement by way of answering my question.
All I am asking for is an AI that is able to solve all of the specific types of problems in any specific standardized IQ test. I wonder if there are any types of problems that are present on most IQ tests that really haven't been solved with narrow AI. I guess the reason it seems interesting is that even though such an AI isn't strong or generalized it does manage a test we use to measure "general intelligence" which I supposed might reflect on the nature of IQ tests if nothing else.
Anyway, since your thesis was on AI and you mention a revolution in the way computers are built, you are probably already familiar with these: http://www.stanford.edu/group/brainsinsilicon/approach.html and http://www.corticaldb.com/ccortex.asp?id=1 and http://news.bbc.co.uk/2/hi/technology/6600965.stm and http://www.opencog.org/wiki/The_Open_Cognition_Project .
Based on what you have said, and if there is AI that is able to solve all of the specific types of problems, the answer is still no. As you pointed out what you would do is tailor several AI solutions to each of the problems, then build another AI solution that determines which solution should be applied to each problem. I suppose you could still argue that it is one “coconscious”.
I think the real issue is that IQ isn’t really a measure of general intelligence. It is a measure of how well you can take IQ tests. If you look at it that way then yes you can teach AI to succeed at them, probably without too much difficulty.
Yes those links are pretty interesting. My issue with current CPU’s is that in comparison to that awesome parallelism of the brain even an 80 core CPU doesn’t even register as a blip.
This may just piss you off more, but I will try to explain anyway why I chose No Best Answer.. in case it helps.
I think that we just weren't able to communicate nearly as precisely as I needed. Both before and after I attempted to clarify myself, you seem to have interpreted my question as being about whether strong AI had been achieved (with the aim of applying it specifically to IQ tests). This really was not what I was asking.
I was also extremely irritated, in fact, by your (and the other responder's) lectures on the nature of AI and qualification of intelligence. Not only did I not get credit for the level of knowledge I displayed in my question and reply, but you also ignored my specific request -- "DO NOT bother explaining what a Turing test is or discussing related issues. Please just answer the question".
So, after all our efforts, it was all around an irritating and futile exercise for myself and apparently unfair to your viewpoint.
Sorry for irritating you, but I felt that a simple "No" might not have sufficed. Besides that dosn't help anyone else who might read this later. You will notice my answer started with "Not to my knowledge." to provide your yes no answer instantly.
I also didn't discuss the Turing test only what my experience and readings have let me know about AI and solving problems using it.
In any case no harm done. I have returned your tip, I didn't really answer for that reason :) Have a good one :)