Artificial intelligence (AI) is, in theory, the ability of an artifact built by humans, such as a computer, to demonstrate
intelligent behavior. It is also the name of the field of science and technology in which such artifacts and underlying mechanisms are developed and studied. Such a definition is somewhat circular, in that there is no generally accepted definition for what "
intelligent behavior" is. In general, intelligent behavior can be viewed either in terms of how human beings behave, or in terms of ideal, or "rational," behavior that is somehow mathematically and logically optimal. Both views of what intelligent behavior should be are widely held and useful. The term AI itself, and the phenomena actually observed, invite philosophical speculation about what constitutes the mind or intelligence (see behaviorism; consciousness; reasoning). These kinds of
questions can be considered separately, however, from a description of the various approaches to constructing increasingly sophisticated mechanisms that exhibit a greater range and depth of "intelligent" behavior. Research into all aspects of AI is vigorous. However, many workers in AI have expressed caution about overstating the progress and short-term expectations of the fieldÑa problem that occurred frequently in the late 1970s and early 1980s. AI programs are often quite primitive when compared to the broad, general kinds of intuitive reasoning of which human brains or even the brains of less advanced organisms are capable. Nevertheless, AI has indeed shown impressive results in many narrower, well-defined application areas with techniques, such as expert systems (knowledge-based systems): computer programs that can answer questions in specific domains but that cannot learn new knowledge or answer questions on any other subjects. Examples of such systems are those capable of answering questions and explaining their reasoning in domains such as medical diagnosis, mineral prospecting, equipment configuration, and chemical structure analysis. Other successful approaches have led to computer systems capable of some degree of speech understanding, natural-language processing, visual interpretation, complex problem solving, legal reasoning, planning and scheduling, signal interpretation, reasoning with uncertainty, and learning. Although examples of such systems have proved valuable in solving practical, real-world, but limited problem domains, the field is still far from producing artifacts with general intelligent behavior.