Browse Books

Reviewer: Jaak Tepandi

A task beyond the capabilities of today's most powerful computers can be performed by the brain in a hundred steps. The approach taken in this book to explore such issues is similar to that often pursued in modern physics and could be expressed as “See what it does, write down its observable constraints or parameters, and try to predict its structure or underlying theory.” So the brain performs certain functions, and it includes certain components. How can they be connected, and which algorithms can make the functions happen__?__ The first three chapters after the introduction are devoted to the constraints of the brain (the neocortex and pyramidal neurons), the computational laws (how they can bridge the gap between the neurobiological level and the level of cognitive behavior), and the cognitive functions (specifications of the behavior exhibited by the brain), respectively. In chapters 5 and 6, a definition of the neuroidal model of the brain is given, specifying the components, their connections, and algorithms on the model. The next four chapters are devoted to modeling various tasks that formalize memorization and inductive learning, among others. Chapters 11 to 13 deal with more complex functions, such as the representation of relations among several objects and simple reasoning. The neuroidal model presented is relatively simple but introduces some severe constraints for implementation. The last chapter shows that these constraints can be relaxed (to better approximate biological reality) without the model losing its computational capabilities. The book is written in a clear style, with a sufficient number of figures illustrating the algorithms. A picture of a pyramidal neuron in section 2.3 would have clarified the text. The appendices include notes, exercises, references, an index of notation, and a general index. All of these are useful—as they complement the text and allow the reader to view it from a different angle—and mostly adequate. Especially welcome are the exercises and index of notation, which are not always present in a research text. The notes occasionally include references that would have been better placed in the main text. It is difficult to say whether after reading this book we are closer to understanding the power of the brain—for <__?__Pub Fmt nolinebreak>example,<__?__Pub Fmt /nolinebreak> to scene recognition in a hundred steps. The functions modeled are the simplest ones available for ordinary computers. The model proposed deals with random access tasks, delegating all other tasks—such as early vision and the interpretation of the three-dimensional world—to various peripheral devices. It seems possible that the essential difficulties in understanding the functioning of the brain may lie in these tasks, which are not part of the model. The main result seems to be in a clearly formulated methodological approach to investigating the brain. In general, the book provides a fresh look, even if what is said is not new, as in chapter 3 on computational laws. This new insight into complex problems of the brain, as well as the proposed methodology, makes the book highly readable and interesting. Time will show if the other potential contributions of the book—the functions modeling real cognition, the proposed knowledge representation, and the algorithms—represent a step forward in understanding the brain. It is even more difficult to say if one should wish to be closer to the kind of understanding mentioned above. The problems of the brain have been subject to many debates, often interesting, witty, and useless. Cases like this, where some real progress in invas<__?__Pub Caret>ion of a sensitive area seems to have been made, force one to ask what the existential, ethical, and moral consequences of reaching this knowledge are. The audience for this research book could be those interested in how the brain works, as well as those who teach or are engaged in research in artificial intelligence, neuroscience, or cognitive psychology. The applications people should be aware (presumably, the military is already). Besides these typical interest groups, the book might be valuable for a computer scientist specializing in new models of computation or in distributed computing.