i am designing a back propagation neural network with MLP method for recognizing handwritten digit please tell me what the main function can be written in MATLAB for doing it?
Arificial Intelligence is a bad name for an interdisciplinary field involving several areas such as philosophy, computer science, mathematics, logics, etc... In my classes I use to say that AI is neither artificial nor intelligence. I prefer to envisage AI as a different way to designing computer programs whose behavour is somehow "similar" to human thinking. There is a big controversy in the definition of AI and still there is not a gold standard definition. Richard Feyman used to call AI as the field of advanced computer applications. I would suggest the readind of John R. Searle and Roger Penrose.
I respect all your answers, but How an AI program will learn from its past mistakes or observing surrounding? Is it like program is writing a program for its own development or progress? Are The Humanoids based on the same principle?
again there is no single solution to machine learning (cf. http://en.wikipedia.org/wiki/Machine_learning). Approaches reach from learning appropriate parameters of an parameterized algorithm to neuronal networks ("human like approach"), genetic algorithms, etc. Usage depends on the application in mind, symbolic vs. sub-symbolic representation, probabilistic vs. possibilistic, etc.
There are at least two schools of thought when it comes to AI. The first is the idea of training and learning. The second is a feed it with all the knowledge in the universe and see what it can do.
This first method is Computational Intelligence techniques, ANN, Evolutionary Algorithms etc.
The second method is PROLOG, Wilson, Chess Playing Machines. etc.
Programs writing programs has been just one implementation of an AI concept. PROLOG and LISP are good languages for the implementations for such ideas, you might be interested in those basics.
You are basically still asking extremely general questions to which there are no complete short answers too, but your on the right track in term of your understanding.
You now need to refine what your trying to get out of AI for us? Are you looking to get a general feel for the subject due to a course your taking, are you looking to implement something?
Joseph and Dieter thank you for your response. Basically what I know about AI, maybe it is not sufficient for the application that I have in my mind.
I am pursuing postgraduate course in Automotive Technology. Looking at the global market of automobiles, the future is "Driverless Car".
I am thinking of introducing AI in the Driverless Car (right now only for prototype), so that it can learn from its surrounding and might behave like humans.
Driverless Car has a variety of issues and each of them involves AI to some extend:
The car has to process the sensor data (e.g. laser scanner) to produce a model of its environment. This goes back to image processing and is known as SLAM (Simultaneous localization and mapping).
The car has to plan its route to its destination. This involves classical AI-planning algorithms.
The car has to drive safely. So it has to watch for other vehicles and predict their motions in time. This gives rise to the problem of recognizing intensions.
...
This is not an exclusive list of AI topics involved in Driverless Cars; just to give you a quick glimpse of the various issues each asking for different techniques.