Not for novices, this sophisticated text by a computer designer--Aleksander is head of the British Cybernetic Society--explores the state of the art of robotics, and goes beyond. Early chapters cover the usual background from Babbage to von Neumann, from RUR to Asimov. What interests the writers, however, is how to go in the future. Truly superior machines--perceptive, language-adept, ""thinking""--will depend on basic shifts in structural concepts. Instead of preprogramming machines in a ""top-down"" mode--building in elaborate memories that enable the machines to choose alternatives from a multi-branched decision tree--the emphasis should be on a ""bottom-up"" approach. This approach enables the computer to learn from its inputs and change its state, so it can essentially adapt to and learn from experience. This might be done, most elaborately, by mimicking the neuronal circuitry of the brain. Here, the authors take off from earlier seminal papers on neuronal nets to describe Aleksander's development of WISARD, a computer program which employs a subset of interconnected random access memories (RAMs). In principle, the combined analysis of these submemories enables a discriminator to detect a shape--say, a wheel--allowing for a variety of deviations from the ideal shape. Thus, the book moves from a general survey, to a critique of different philosophies of artificial intelligence (and other theories governing the design of robots or machine translation), to arguments for the authors' approach. Those with hands-on experience will find much stimulating and discussable material here.