Toward a Mechanics of Conceptual Machines

This paper suggest that conceptual machines|computational objects such as neural networks, fuzzy systems, and genetic algorithms|are being tamed by a kind of conceptual mechanics that brings a spectrum of models, ranging from raw intuition to complex equations of motion, to bear on crucial design problems. Beginning by questioning the cultural divide that separates material and conceptual machines, the paper shifts the debate to making a distinction between description and prescription. After arguing that the prescriptive enterprise should be built on economic foundations, the paper reviews how one class of conceptual machine|genetic algorithms|has succumbed to the integrated application of a number of little models|mid-range quantitative models built on dimensional or facetwise grounds. These results are still nding their way into practice, but they are paving the way to the speedy, accurate, and reliable solution of a large array of diicult problems.