Michael A. Arbib,Editors, ,The Metaphorical Brain: An Introduction to Cybernetics as Artificial Intelligence and Brain Theory (1972) Wiley-Interscience,New York 243+xii pp. £6.25.
✍ Scribed by R.L. Gregory
- Publisher
- Elsevier Science
- Year
- 1974
- Weight
- 137 KB
- Volume
- 6
- Category
- Article
- ISSN
- 0020-7373
No coin nor oath required. For personal study only.
✦ Synopsis
The strength of this interesting book is like glass fibre--given by intricate combining of dissimilar threads--from cybernetics and from neuroanatomy. The author started his career as a mathematician and programmer, and developed, partly through the influence of that fascinating person the late Warren McCulloch, a detailed knowledge of the nervous system as studied by biologists. This book sets out both to look at nerve cells and their connections, and also at ways in which they may be functionally organized to yield the complex patterns of behaviour. Arbib is looking essentially for models of neural function; but (unusually) he pays due regard to the components and pathways which, somehow, are intelligent.
Implicit in this search for theoretical models is the notion that functional principles of the brain can, at least in principle, be described and discussed apart from actual brains. Further, that the principles of function might, at least in principle, be applied to making intelligent machines, though with components very different from brain cells. So the bringing together of neuroanatomy and cybernetics, while being philosophical, is yet a kind of applied engineering--in which the answer is known before we even know how to ask the questions, let alone answer them. An "answer", of course, is ourselves.
Engineering models of brain function have changed in emphasis over the last two or three decades, from the cybernetics of Norbert Wiener which was mainly concerned with control systems employing feedback; to a far greater emphasis on computing techniques, and data retrieval and classification, with the hope of designing autonomous intelligent devices. Hopefully these will accept data directly from the world with sensors such as T.V. cameras, and act upon it effectively with limbs or special tools. This, Arbib calls the Artificial Intelligence Approach--the point being that intelligence might be produced with devices quite unlike organisms to look at, including their detailed structure--but essentially similar in key features of the logic of their design. It is an open question how far knowledge of the structure and physiology of organisms will help to reveal models necessary for understanding intelligence; but it seems a good bet to have a look at examples of "answers" provided by nature. (This may be rather like looking, almost guiltily, at "worked examples" at the end of a text book, but this can be helpful!)
As an example of the balance of the arguments presented: in a discussion, near the start, of logical difficulties inherent in locating function by cutting out bits of a system and seeing what happens Arbib states the problem clearly, with examples of circuits where an "ablationist" engineer would be misled, but then he rightly says that much has in fact been learned from brain ablation studies by neurologists. 729