The History Of Knowledge Management Part I: The Search for Artificial Intelligence
The first stage in the development of knowledge management focused on replacing human intelligence with artificial intelligence. Artificial intelligence had been a subject for debate long before computers were available. Philosophers such as Descartes debated whether people could be thought of as machines. With the advent of computers, the debate focused on whether they could be made to think. An early attempt at an `intelligent' program was made by Newell, Shaw and Simon in the 1950s. The aim of the work was to produce a general problem solver (GPS). The generality of the GPS meant it was inefficient and superficial in its analysis.
The lesson from early systems such as the GPS was the realization that the successful encapsulation of knowledge is most easily attained within a specific and narrow problem domain. This led to the concept of expert systems as systems which model expertise within a narrow and specific domain. Within such domains, some notable successes were achieved; for example XCON was used by DEC to configure computers for many years from the 1970s onwards. It is noticeable, however, that the same case-studies are constantly quoted.
Today, there is still a mystique surrounding the subject of AI and consequent confusion about what can be achieved in practice. So let's address these issues. Computers do not think for themselves. An AI application is a program like any other. It is a set of procedures, which the computer follows, with inputs and outputs. AI programs use a symbolic representation of the world. All computer programs are based upon a model of the problem they seek to solve. In conventional information systems, this model is constructed in terms of numbers. In an AI program, the model is constructed in terms nearer to a human view using symbols, such as text or pictures. The computer's view is still a model, but is simply expressed in different terms.
Machine intelligence is not human intelligence. Human intelligence is a very broad and varied thing. Some of the characteristics of intelligent behaviour are reasoning, deduction, learning and adaptability. Machine intelligence seeks to model this behaviour, but is only able to do so in limited domains and in limited areas. Many machine intelligence techniques focus on only one or two aspects of human intelligence.