Key points are not available for this paper at this time.
Successes with expert systems and other specialized systems have revived hopes for factory automation and productivity growth. A full realization of these potentials will require conscious effort to overcome obsolete rigidities, to develop unified and adaptive methods for integrating complex systems, and to increase our understanding of these systems (understanding which is vital to human productivity in developing software). How adaptive systems may be built and understood by extending control theory and statistics is discussed. Adaptive systems, like human infants, are less agile than young monkeys but have something important to contribute as they mature. It argues that the old dream of understanding intelligence in generalized terms, permitting a unified understanding of adaptive systems and of the human mind, was not incorrect; rather, the early attempts in that direction failed because they did not make full use of research possibilities in statistics, control theory, and numerical analysis (many of which are still unexploited) and were limited by hardware costs which are now coming down. A basic adaptive system derived from this approach fits the fundamental, qualitative empirical facts of human brain physiology in some detail (unlike the usual "general neuron models," which rarely even discriminate between basic components of the brain), and offers opportunities for further research; it can even translate certain fundamental ideas of Freud into something more mathematical and scientific. The mathematics of the basic system, and the fit to the brain, are described in detail.
Paul J. Werbos (Thu,) studied this question.