In summary, this two volume set is an excellent overview of AI as theorized and practiced at MIT and one may easily get acquainted with "more AF by simply following the references at the end of each paper. Although all the articles in Volume 1 (and similarly in Volume 2) are individually available in AI journals and symposium proceedings, their structured and coherent presentation under general headings make these two volumes indispensable. I strongly believe that every serious AI researcher should have this set available for reference and for reading. I will be very pleased if Winston and Brown possibly spare some time to update and enlarge this beautiful work with new contributions once in a while. (Surely, things have changed quite a bit since 1979). I can even hope that his work of Winston and Brown will be the origin of a much wider and larger set of books embracing all memos published by the MIT Lab since its inception. Then, one can call that work the, "MIT Encyclopedia of Artificial Intelligence." One might, perhaps, question whether hedonism is an appropriate metaphor for brain research. Viewers of the NATURE series on animal behavior may tend to be a bit cautious about using any human characteristic as a general metaphor for behavior, and they certainly have been given evidence that unadulterated behaviorism is not necessarily "the way" of animal behavior. Nevertheless, one scientist should never attempt to tell another scientist what his hypotheses should be, particularly if that latter scientist has a well-formed plan of experimental investigation. One can only wonder, however, if the author will be able to provide adequate support for his research plan unless he is doing research in the best of all possible worlds. Given that I am not very well versed in neurophysiology, I am not in a position to analyse the author's arguments in great detail. As a computer scientist, however, I must admit to a certain degree of disappointment. The initial discussion is concerned with the development of a neurophysiological model, and this model is built in such a way as to accommodate the myriad features discussed in the quoted paragraph. Unfortunately, the model is so big and so general that a reader who is armed with little more than commen sense cannot really develop a feel for what it actually does. It would be nice to see some sort of concrete demonstration of the model, something which would justify the author's faith in it. After all, the one danger of the book is that the reader will forget the author's own disclaimer. The line between a valid proposal and an idle speculation is often hard to draw. One might also fault the author for setting up outdated research in artificial intelligence research as a straw man. For example, he feels that a major fault of AI is that it has been concerned with systems built on von Neumann-style architectures. This may have been true when he first wrote that sentence, but it is certainly no longer true. He also seems to feel that the concept of "knowledge base" involves "a passive approach to learning and memory," an assertion which is also questionable as of the writing of this review. In summary, one should not expect too much from this book. It neither promises nor delivers a panacea. It does offer a lot of provocation. Given that it entails less than 200 pages of reading matter, one could do far worse than devote one's attention to it. THE H E D O N I S T I C N E U R O N : A T H E O R Y OF M E M O R Y , LEARNING, AND INTELLIGENCE Stephen W. Smoliar Sehlumberger-Doll Research P. O. Box 307 Ridgefield, CT 06877 (A. Harry Klopf, Hemisphere Washington, 1982, 140+xvii pp.) Publishing Corporation, Science is a wonderful thing, but it has not succeeded in maximizing pleasure and minimizing pain, and that's all we asked of it. Notes and Comment, The New Yorker, June 13, 1970 cited by Paul Dickson, in The Official Rules In an age in which verbose obfuscation becomes more fashionable with every new government appointment, one has to appreciate an author who can give a cogent summary of his entire book in the second paragraph of his Preface: The book examines a theory of brain function based on the postulate that neurons "seek" excitation and "avoid" inhibition, this goal being pursued within well-defined limits that preclude an epileptic outcome (except under pathological circumstances that, indeed, we know do arise). In the development of the theory, it will be seen that a plausible adaptive mechanism exists that can account, in mechanistic terms, for the postulated neuronal behavior. Also, the notion of an excitation-seeking, inhibitionavoiding neuron will be shown to be consistent with experimentally observed neuronal behavior. Neuronal and cortical polarization studies, the mirror focus, and epileptic foci appear to be understandable in light of the proposed neuronal model. At a psychological level, habituation, dishabituation, classical and operant conditioning, and extinction can be shown to be straightforward consequences of a goal-seeking neuron like the kind proposed. None of this is to suggest that the validity of the proposed theory is demonstrated here. Far from it. Difficult experiments will have to be performed at the neuronal level to test the theory rigorously. Until these experiments are accomplished, the significance of the theory lies in its offering a fundamentally new view of brain function, a view that suggests alternative, and perhaps more productive, experiments. Artificial Intelligence: An MIT Perspective, vol. 2 Reviewed by: Ram Nevatia Dept. of Electrical Engineering University of Southern California Los Angeles, CA 90089-0272 (Edited by P. H. Winston and R. H. Brown MIT Press 1982 (paperback edition)) This is a paperback edition of the original version published in 1979. Vol. 2 is divided in three sections dealing with vision, manipulation and machine design. (Vol. 1 deals with expert problem solving, natural language, and representation and learning.) Each section contains chapters written by various MIT researchers. In general, each chapter describes recent (as of the original date of publication) MIT research in the particular areas. Each section also provides a nice overview and summary by the editors. However outrageous one finds the thesis (and if nothing else, the thesis is certainly unorthodox), one must give the author credit for his honesty. If he seriously intends to pursue the experimental program proposed at the end of the above paragraph, then one must be fair enough to wait for the results to come in. In the meantime, this little book provides an interesting source of provocation for those whose views of cognition may have fallen into a rut.

/lp/association-for-computing-machinery/review-of-the-hedonistic-neuron-a-theory-of-memory-learning-and-dukLkws47n