Artificial Intelligence: An introduction [Archives:2005/846/Education]

Prof R K Jayaraman
Dept of English
Faculty of Education
Sana'a University
Sana'a
[email protected]
In an attempt to understand the concept 'Artificial Intelligence' (AI, for short), and also AI as a field of serious, scientific study, we shall look at the following definition of the term in some detail:

The concept 'Artificial or Machine Intelligence' may be provisionally defined as 'electronically simulated human intelligence'.

This may appear to be a fairly straightforward and easily comprehensible definition you have ever come across, but I must warn you that it is somewhat deceptive. For example, what do you understand by 'human intelligence' (HI, for short), to begin with? I am sure that you believe that it is one of those household words like, 'foolishness', 'sensitiveness', and so on. Synonyms such as 'shrewdness', 'cleverness', brightness of mind', etc., may also occur to you. The problem, however, is that while HI as it is used here is certainly not incompatible with the sense of 'cleverness' or 'brightness of mind', it is technically wrong and misleading to equate the word with these meanings. For example, even a fool is said to be endowed with HI, if he can spontaneously smile a smile of recognition, say, when he meets a friend after a gap of time. That smile is a sure indication of the fact that the person is capable of adjusting his behaviour to what happens around him and this capability is symptomatic of HI. Again, if an idiot can avoid fire when he sees it, with out having to come into bodily contact with it, he is 'intelligent', too. In other words, HI is that faculty of any normal (that is, not mentally retarded) human being, which accounts for all his actions, speech and behaviour – idiotic as well as informed.

Now, when a computer is programmed to simulate this faculty of human beings, it is said to be 'intelligent', and this capability of the machine is 'Artificial or Machine Intelligence'.

Let us, in the next place, turn to AI as a field of serious, scientific study. Here it is useful to make a distinction between AI theory and practice. As theory, AI is the engineering counterpart of 'cognitive science'. Cognitive science is a cover term for 'philosophy', 'linguistics', and 'psychology', each of which claims to account for certain complementary aspects of HI. Philosophy asks and answers questions about 'knowledge', and its nature and expression; it seeks to find out whether all knowledge is realized only in terms of language or whether there is a kind of knowledge which is language independent. The nature and expression of 'meaning' have also been the subject matter of almost any school of philosophical enquiry. Now, no serious discussion of 'intelligence', be it natural or artificial, can afford to ignore the philosophical enquiries concerning 'knowledge' and 'meaning'.

As for linguistics, it is the scientific study of language, its structure and function. It aims to relate the world of meanings to the world of sounds on the one hand, and the world of graphic symbols on the other. Our definition of HI laid down that all human actions and behaviour, including 'speech' are symptomatic of intelligence, and speech certainly belongs to the domain of linguistics.

Psychology, to be more precise, cognitive psychology is an equally important constituent of cognitive science, and is very closely related to intelligence. Of particular interest to us, here, are those aspects of psychology which attempt to characterize mental models of objects – captured as facts and reasoning of various kinds.

Thus, various aspects of HI are fruitfully studied from the view points of philosophy, linguistics and psychology. The viability of any AI theory directly depends on how effectively it can capture the essential features of HI for purposes of AI modeling and hence the suggested influence of cognitive science on AI theory.

Let us return to our definition of AI once again, before taking up AI practice. There are two parts to this definition: (a) human intelligence, and (b) electronic simulation. Our discussion of AI practice, which primarily stresses on the functional aspects of AI, will center on the electronic simulation part of the definition. Thus, functionally, AI is concerned with the design of a computer to simulate human intelligence. We said that all human action, including speech, is symptomatic of intelligence. In order to determine the scope of electronic simulation of HI, however, a greater understanding of HI than a mere analysis of symptoms becomes necessary. A clearer perception of what constitutes HI than what evidence we have of it is highly relevant to the simulation efforts of AI practice.

'Knowledge' and 'reasoning' are two important constituents of HI. Knowledge comes to us first or second hand. We see things, feel them, and in appropriate cases, smell and taste them and thereby, KNOW about them first hand. We hear about persons, places, and things; we read about them, and in appropriate cases, see 2-D or 3-D images of them, and thereby KNOW about them second hand. We develop intellectually perceived artifacts, talk about them, and thereby exchange KNOWLEDGE about them. We come to KNOW about many things without being aware of the process; we also acquire KNOWLEDGE through conscious training, and education. A large body of knowledge is merely inferred knowledge; there is no need for us to have personally experienced this body of knowledge. For example, it is intelligent behaviour to avoid coming into violent, head-on, bodily contact with a fast-moving object, like a speeding bus. The human ability to avoid this collision course is not necessarily based on personally experienced knowledge; nor does this ability come necessarily or conscious education or training; it is, in a large number of cases, based on inference or reasoning.

Now, having said that all actions of any normal human being are symptomatic of intelligence, how then do we account for variations in intelligent behaviour among people? Good question. Well, these variations are due to personal styles of behaviour, culturally conditioned styles of behaviour, and even misbehaviour, brought about by intellectual intervention. Thus, at the heart of any intelligent behaviour are knowledge and reasoning, which may be assumed to be located in the human brain. The application of this knowledge in a given situation and the activation of the reasoning capability may be substantially moderated, coloured, or influenced by the personal emotions and the impulses of the individual concerned. We know how the parents' judgment of their children's behaviour is often characterized by a degree of tolerance, which they may not be prepared to show when it comes to judging the behaviour of other children.

Our intelligent response in a given situation may be further moderated by 'cultural conditioning'. Thus people belonging to different cultures show different intelligent behaviours. For example, what is morally right and constitutes social justice in culture 'x' may cause a revolt in culture 'y'.

Against the background of this discussion of what constitutes HI, let us examine the principal concern of AI practice, namely, electronic simulation of HI, or the design of an intelligent software to programme a computer to behave like a human being to the extent possible. The concern may be expressed in the form of the following questions:

a) What aspects of HI are capable of being electronically simulated? Which of them defy any attempt at electronic simulation and why?

b) How to simulate what is to be simulated? What are the mechanical means of simulation?

c) What constraints operate on electronic simulation?, etc.

For reasons of space, I do not intend to answer these questions here. I may, however, briefly indicate that those aspects of HI, such as, personal emotions, moral overtones, cultural colouring etc., which defy quantification, defy any attempt at simulation. It must, however, be mentioned that AI practice is still in its infancy and that no great progress has been made in the matter of successful simulation, except in some areas like, robotics, expert systems, and so on. Even in these areas there is further scope for improvement.
——
[archive-e:846-v:13-y:2005-d:2005-05-30-p:education]