This was a high point in the story of artificial intelligence in the early days. If simple programs could achieve so much, what would happen if one were to make a major concerted effort to achieve intelligence in computer programs? It seemed that the sky was the limit!

CAN THERE BE A SYSTEM TO SOLVE ALL INTELLIGENCE PROBLEMS?

Alan Newell, Simon and Shaw started work on an ambitious 'Generalized Problem Solver System (GPSS)'. This was to operate in a really wide variety of environments. Their approach, nevertheless, was quite simple. The idea was to study and incorporate methods similar to those followed by humans in solving everyday problems.

To achieve this, they observed human subjects solving such problems. The subjects were to do loud thinking while solving them. By analyzing their loudly articulated thoughts, the researchers evolved a general problem solving approach, which they felt would work in a variety of situations.

To illustrate this, let's examine a simple real life problem. To quote Newell, "I want to take my son to nursery school. What is the difference between what I have and what I want? One of distances. What changes distances? My automobile. My automobile will not work. What is needed to make it work? A new battery. What has new batteries? An automobile repair shop. I want the repair shop to put in a new battery, but the shop does not know that I need one. What is the difficulty? One of communication. What allows communication? A telephone…… " and so on.

What is the basic strategy that the GPSS incorporated?
Answer: "Achieve a goal by setting up sub-goals whose achievements lead to the achievement of the main initial goal.

" We have to admit that in a sense, this strategy has universal validity. It is quite another thing that the GPSS strategy did not turn out to be a panacea for all the problems of artificial intelligence.

Such programs use a heuristic, as opposed to an algorithmic, approach. A heuristic incorporates a trick or a rule of thumb - A provisional or plausible strategy which is likely but not guaranteed to work. In contrast, computer programs in general use an algorithmic approach. An algorithm is a systematic proven method which is guaranteed to work under the conditions specified for it.

next page>>>

TATA INFOTECH - TOTAL SOLUTIONS TOTAL COMMITMENT