The AI Problems
The Artificial Intelligence study earlier dealt with the tasks such as game playing or theorem proving. One of the primary reasons for choosing these fields was that it was thought that people who do well in these tasks are considered to be intelligent. It was assumed that as computers are fast at exploring various solution paths and also select the best out of them and these processes do not need large amount of knowledge and therefore what ever amount of knowledge is required therefore can be programmed easily. But it proved out to be completely wrong as the no. of solution paths available for AI problems is very large. No computer is that fast which can cover all the available combinations.
The AI problems can be broadly divided in to various categories like :
- Ordinary Tasks
- Formal Tasks
- Expert Tasks
Ordinary Tasks:
- Commonsense Reasoning: Commonsense reasoning i.e. developing computer systems which has some commonsense like if we let fall any thing on the floor it may break. In other words reasoning about relation between various physical objects. General Problem Solver is a technique which is applied to the problems involving common sense reasoning.
- Perception: Perception includes two basic properties what humans generally posses the i.e. Vision and Speech. Developing computer systems which can perceive about the world around us is difficult task as they involve analog signals as compared to digital one. People can see easily see lots of things at one time even those who are obscuring one another.
- Natural Language understanding: Communicating various ideas is perhaps the most important thing that differentiates humans from animals. Making machine understands spoken language is very hard nut to crack because of many reasons but some obvious ones are: different people have different way of speaking i.e. pitches may vary, accent may be different etc. If the problem is restricted to written language then that problem is called Natural Language Understanding. But even if we restrict ourselves to understanding written language it still difficult.
Formal Tasks
- Game Playing: Making computers playing games seems to be very interesting that is why many researchers have extensively contributed for computer game playing methods. As discussed above the development of computer games are not easy because of combinational explosion of solution paths. Take a simple example of playing chess there are 64 black and white square boxes available. Think about the number of possible moves a player can make during the game is estimated up to about 35100. Even if the fastest computer is available the amount of memory required will be very high. Hence one important property about AI problems can be noted down here is that it involves combinational explosion of various possible states. Chekers is another game which also had great deal of attention Arthur Samuel build the a chekers playing program which not only played but also learnt from the experience and the mistakes.
- Mathematics: Finding a proof for a theorem in mathematics is certainly is an intelligent task. The study of theorem proving play a significant part in development of Artificial Intelligence Methods. The formalization of deductive process using the language of predicate logic helps us to understand more clearly some of the components of reasoning. Mathematics becomes important as it provides the basic structure to a program called logic and once logic is known then putting it in a form of a program becomes a little easy.
Expert Tasks:
- Expert Systems: This area of AI deals in creation of computer systems which can perform those tasks which now a days is performed by experts. Expert systems are the expert programs that manipulate encoded knowledge to solve problem in a particular domain e.g. Medical, Military, Chemical composition, Scientific analysis etc. To develop these type of system great amount of knowledge can captured from magazines, journals and from domain specific experts e.g. DENDRAL is an expert system which can replace a doctor and diagnose a disease. Strange but true DENDRAL expert system proved to be 65% more reliable as compared to humans. One of the major reason for this is considered to be that human feel fatigue after some time and computers never gets tired.