 |
|
|
 |
|
|
|
|
abhishek |
| | | | | | posted 4/16/2010 10:33 |   | | | | | | | | | My question to everyone working in the field of AI is whether we are working in the write direction or not....
I believe the idea, that was coined by Allen Turing in his 1950 Minds article, of AI was misinterpreted and we are heading a wrong direction on creating machines that will just imitate humans.
The question that arises in my mind is "what are we creating a new race of intelligent machines or just a thing that will imitate us....."
Regards,
Abhishek singh
My blog address is: http://abhishek-abhishekblog.blogspot.com/
| | |  My blog on Artificial intelligence | | |
| | |
tkorrovi |
| | | | | | posted 4/16/2010 15:31 | | | | | | | | | | The Turing idea was not described precisely enough, many questions remain open. Is a mere impression that a system is intelligent, enough for a positive result, or should such test include learning more about the system? Turing had not proposed that one has to find out what the system knows and what not, but the system's abilities cannot be objectively assessed without finding out more about the system itself. And sure the Turing idea was misinterpreted, Turing likely did not want to propose a test that is only based on subjective opinion, but it seems that this is the most common interpretation.
| | | Last edited by tkorrovi @ 4/16/2010 9:32:00 PM | | |
| | |
abhishek |
| | | | | | posted 4/18/2010 08:07 | | | | | | | | | | I have proposed a new test for intelligence based on an improvised model for intelligence. Kindly go through my blog....
If anything not clear kindly aproach me....
Help me impovise the idea so that we can achieve intelligence in machines.
Thanking you
Regards,
Abhishek Singh
http://abhishek-abhishekblog.blogspot.com/
| | | Intelligent Systems......... | | |
| | |
tkorrovi |
| | | | | | posted 4/18/2010 20:18 | | | | | | | | | | I have proposed an objective and statistical test.
Objective means that we test whether the system has certain abilities based on objective criterion, like can the system learn. Such testing is based on the knowledge about the system, like how much is pre-programmed or pre-determined, based on that and the results of the testing we can decide how much ability to learn the system has.
Statistical may not seem so good, but when the statistics is applied strictly and rigorously, then it is objective. Certain self-developing AI systems cannot be tested otherwise than statistically, and also it is not correct to test some other systems such as neural networks, otherwise than statistically. An example of statistical testing is training a system to play a simple game, each time when playing a geme, record the result (win/lose), and then calculate the z score from these records. The z score must be more than 2 for the test to have positive result.
This is the only scientifically valid way which i know, of testing AI systems.
| | | Artificial Consciousness ADS-AC project | | |
| | |
toborman |
| | | | | | posted 4/19/2010 01:11 | | | | | | | | | | | |
abhishek wrote @ 4/18/2010 8:07:00 AM:
I have proposed a new test for intelligence based on an improvised model for intelligence. Kindly go through my blog....
If anything not clear kindly aproach me....
Help me impovise the idea so that we can achieve intelligence in machines.
Thanking you
Regards,
Abhishek Singh
http://abhishek-abhishekblog.blogspot.com/
| |
In your blog you say:
"If I talk about our mind, there are three possible answers to a question 'yes', 'no' or 'may be'."
I think it would be more appropriate to say "I don't know" instead of "may be", since the answer may also mean "maybe not".
When the answer is "I don't know" then the following methods may be used to narrow the answers.
Abductive reasoning should narrow the search space to those answers that are "possible."
Inductive reasoning should provide the experiments, which would determine the probability for truth.
Deductive reasoning should narrow the answer even more by determining what is not possible, using the "square of opposition".
| | | Human Mind Map | | |
| | |
toborman |
| | | | | | posted 4/19/2010 01:22 | | | | | | | | | | | |
abhishek wrote @ 4/16/2010 10:33:00 AM:
My question to everyone working in the field of AI is whether we are working in the write direction or not....
I believe the idea, that was coined by Allen Turing in his 1950 Minds article, of AI was misinterpreted and we are heading a wrong direction on creating machines that will just imitate humans.
The question that arises in my mind is "what are we creating a new race of intelligent machines or just a thing that will imitate us....."
Regards,
Abhishek singh
My blog address is: http://abhishek-abhishekblog.blogspot.com/
| |
People in AI are working on ideas that they deem important and technically viable. This means they are not collectively working on one goal, but individually towards their own goal.
The Turing test is not their guiding light, nor will it be yours. The test you choose should satisfy your own objectives.
| | | Human Mind Map | | |
| | |
abhishek |
| | | | | | posted 4/19/2010 12:03 | | | | | | | | | | | |
toborman wrote @ 4/19/2010 1:22:00 AM:
People in AI are working on ideas that they deem important and technically viable. This means they are not collectively working on one goal, but individually towards their own goal.
The Turing test is not their guiding light, nor will it be yours. The test you choose should satisfy your own objectives.
| |
Working independently to achieve same goal does not seem to be a perfect idea.
Instead we should all work on a standardized definition.
As far as Turing Test is considered... I would say that it can be modified to get a test which could be a criteria for Artificial intelligence...
| | |
| | |
lordjakian |
| | | | | | posted 5/3/2010 11:39 | | | | | | | | | | I say the turing test is COMPLETELY about impressions. To me its like someone saying ", Hey, you know what, if you, a perfectly intelligent person could be fooled by the computer program, then the computer program is displaying a property of intelligence.
You could describe a physical trap that fools a human as being an intelligent trap.
Or akin to a poorly prepared meal served to a starving man as being called GOOD, as related to a funny email incident where a poorly made chatbot acting as a poor english speaking russian girl-bot had a three month relationship with some sad american who thought he was talking to a real person till he became suspious as started to ask about her family. Ahhahaha, read about that somewhere.
"Imitating humans" is a wrong direction, and I think most people believe that. Amazing is the kind of natural handicaps people will accept when they feel they are talking to someone who they know does not speak thier own natural language. Fun, fun.
| | |
| | |
hunt |
| | | | | | posted 5/3/2010 20:03 | | | | | | | | | | | |
lordjakian wrote @ 5/3/2010 11:39:00 AM:
You could describe a physical trap that fools a human as being an intelligent trap.
| |
He he he--this made my day. I agree that a Turing-style test may not be a reasonable approach to test intelligence.
However, it is the best approach to building a useful (and profitable) human-computer interface. Because when it comes to turning the average person's thoughts into actionable commands, their must be translation from natural language.
| | |
| | |
abhishek |
| | | | | | posted 5/4/2010 07:47 | | | | | | | | | | Well, I think, Turing Test after a bit modification and putting some constraints can act as a perfect test to test the intelligence of a thing.
I have mentioned the version of Turing Test developed by me on my blog.....
Kindly go through that and post your replies...
the blog address is: http://abhishek-abhishekblog.blogspot.com/
| | | For my blog click here | | |
|
|
|
|
|
