The Computational Brain (Computational Neuroscience)

This book can be viewed as one of the first attempts to use results from psychology, neuroscience, computer science, and philosophy with the intent of gaining an understanding of how the mind/brain works, but all of this is done within the "computational mind" paradigm. The approach taken by the authors is one of the most honest of those in the literature, for throughout the book they are careful to note just how much evidence there is to support their position(s), and to what extent further work is to done. Philosophically speaking, the authors are clearly in the materialist camp, believing that Cartesian dualism does not cohere with current scientific knowledge. But they state that materialism is not an established fact, allowing the possibility, but not the probability, that dualism may in fact be true. They reject early on though any "arguments from ignorance" in their assertion that just because neuroscience does not have an explanation of consciousness, that such an explanation is impossible. The authors call the failure to be able to think of consciousness in terms of neuronal activity "intuition dissonance", and reject completely its efficacy in establishing the truth of the nature of the mind/brain. The underlying theme in the book is to explain emergent properties as "high-level" effects that are dependent on "lower-level" phenomena, hence rejecting the thesis that they are "nomologically autonomous", i.e. that such a dependence cannot be done and is outside the domain of science. The science in this book recognizes its historical origins, and it is clear that the authors will not accept explanations of the mind/brain that do not involve scientific experimentation and analysis. Much has been done experimentally in neuroscience since this book was published, especially using the techniques of magnetic resonance imaging (MRI). A brief discussion of MRI is given in the Appendix of the book, but no doubt if the book were updated there would be a lengthy overview of it. The current experimental situation in neuroscience has led some to predict a total "reverse engineering" of the brain in the upcoming decades. This prediction is an optimistic one, but no doubt detailed knowledge of the brain will continue to accelerate, this being a sign of what the authors call "a remarkable time in the history of science". The authors devote an entire chapter to the computational modeling of the brain, mostly of course dealing with the mathematics of neural networks. The approach in this chapter though is still at a level that would allow a general audience to follow it. Readers with a background in physics, especially statistical physics, will appreciate more the discussion on Hopfield networks and Boltzmann machines. Experimental results are inserted as graphs throughout the book, with detailed explanation. As a whole the discussion of the biology of the brain is purely descriptive, and the line drawings could stand some improvement. The chapter on neuronal plasticity is the most interesting in the book, the authors viewing the brain as an entity that is continuously undergoing modification. Their stated goal in the chapter is to explain how the "local" property of plasticity can result in the "global" property of learning. Clearly intelligence to the authors is an emergent property, i.e. an object or device may be characterized as intelligent without its components being intelligent. Particularly interesting in this chapter was the discussion of the amnesia of a patient who underwent bilateral resection of mesial temporal lobe structures. The time scales of the patient's memory are striking: he remembered things before the surgery but could not remember things that happened a few minutes or hours ago, but could remember things within a minute in his past. The authors also mention the fascinating work of Antonio Damasio and his collaborators, this research being even more important at the present time. The scientific study of consciousness is just beginning and no doubt this study will give many surprises as it develops throughout the twenty-first century.

There is an argument that this is a book of its time. It is nearly fifteen years since it was put together and a great deal of neural water has flowed under the bridge. The thematic enthusiasm for computationalism that dominates the book has not been convincingly proved in the meantime. If anything, the computational properties of models have been shown to entertain many unpleasant complexity results. Moreover, the localisation of brain functions grounded in naive interpretations of lesion effects has come under greater scrutiny due to detailed MRI results. Given twhat was known at the time, it is unsurprising that the book focuses on the visual system - a focus also found in Christof Koch's recent book. Acknowleding all that and more, it would be hard to find a better condensation of science, computationalism, and philosophical speculation than in this book.Leaving aside downsides arising from recent discoveries that the authors could not have anticipated, the book can be frustrating to read at times. In particular, there is a tendency to introduce technical concepts and descriptors into accounts without prior definition. For example, very early on in a brief account of monkey vision there is mention of V4, MT, etc. The terms are neither defined nor explained. Strangely, in the introduction to networks, the inner product of two vectors is explained while the outer product is not. Small points but the oversight recurs.The philosophical content in the book is light, but the assumptions driving the work are among the most contentious. There is no point reaming off a list but the book does not shirk supporing the brain-as-a-computer hypothesis.All in all a stimulating work, if in need of updating.

good in its domain albeit long

I bought this book in 1995 and read it cover-to-cover in 2009. It is an enormously valuable tour of many neuroscience topics, but the general (reasonably sophisticated) non-specialist reader must be tolerant of the very numerous gaps and discontinuities. In other words, this book is not self-contained in the sense that a non-specialist can pick it up and understand it.In order to wade through all of this book, one must first know enough to be able to fill in the very large gaps, and to bridge the discontinuities. I think that at least half of the technical terms were not defined at all. I felt it was a bit like picking up the proceedings of a research conference. One does not expect continuity in conference proceedings; nor does one expect everything to be defined. But generally in a book, one does expect definitions and continuity. The authors have given copious references parenthetically in the text, and background reading suggestions at the end of each chapter. But footnotes with definitions of terms would have been more useful. If one needs to look up too many things, one may as well read a different book. (Of course, there is a substantial bibliography at the end of the book also.)Still, it's good for those who are curious about the neurosciences to be thrown in at the deep end to read some snippets about the functional hierarchy of the brain, the basic facts in each level of the hierarchy, the functional capabilities and interrelationships between regions and sub-regions of the brain, numerous specific mechanisms for transmitting information from one part of the brain to another, numerous methods of encoding sensory inputs for transmission within the brain, some intriguing functions of the hippocampus, and the coupling between sensory inputs and motor outputs.The impression I had after reading this book is that the brain is no more nor less than a big associative engine. In other words, it is totally unlike the von Neumann architecture or any other kind of human-made computer. I also had the impression that an amazing amount is already known, and I look forward to the day when enough pieces of the jigsaw have been put in place to see an overall picture emerge for how the brain does what it does. (My own principal interest in the matter is to know the physical basis of the "seat of consciousness", for some meaningful definition of that naive concept.)My conclusion is that this book is an excellent sampler of research in computational neuroscience. Since it covers a broad area, it gives some perspective on the relations between the different areas of the subject. So it's a good place to start, to get some ideas for what to read next. However, don't expect an integrated, self-contained textbook-style presentation of the subject.

This book is comprehensive in its detailing the background and theory of computational neuroscience but is by no means inaccessible.