Book jacket: In The Brain from 25,000 Feet, Mark A. Changizi defends a non-reductionist philosophy and applies it to a variety of problems in the brain sciences. Some of the key questions answered are as follows. Why do we see visual illusions, and why are illusions inevitable for any finite-speed vision machine? Why aren't brains universal learning machines, and what does the riddle of induction and its solution have to do with human learning and innateness? The author tackles such questions as why the brain is folded, and why animals have as many limbs as they do, explaining how these relate to principles of network optimality. He describes how most natural language words are vague and then goes on to explain the connection to the ultimate computational limits on machines. There is also a fascinating discussion of how animals accommodate greater behavioral complexity. This book is a must-read for researchers interested in taking a high-level, non-mechanistic approach to answering age-old fundamental questions in the brain sciences.
Beginning of preface: Since brains are not "impossibility engines," they cannot do the logically impossible. Since brains are not infinite in size or speed, they cannot do the computationally impossible. Since brains are found in the universe, rather than in some fantasy world, they cannot do the physically impossible. Brains have constraints. And not simply the garden variety Earthly constraints like having to work well at the human body temperature, or having some upper limit in working memory. Brains have "high level" constraints, by which I mean constraints to which nearly any other possible kind of brain will also be subject. Such constraints are typically more at a physics or mathematics level, rather than depending on the particular contingencies of the ecology encountered by any specific kind of brain.
To understand the brain one must, I believe, understand the limits and principles governing all possible brain-like things--objects that somehow instantiate minds. For example, if I tell you I want to know how this computer on which I am typing works, and you tell me about the details of just this kind of computer, you really would have missed the point. That is, unless I already knew all about how computers worked generally, and just wanted to know the specifics about this kind of computer. But if I did not understand how any kind of computer works, then to really explain to me how this computer works will require telling me how computers work in general. That is what is interesting about computers and computation: the fundamental questions in computer science are about how computer-like things work generally, not about how this or that computer actually happens to work. Similarly, what is most fascinating about the brain is not that it is our brain (although that helps), but that it happens to be one of presumably an infinite class of brain-like things, and I want to know how brain-like things work.
To do this, one must back sufficiently far away from the brains we find here on Earth so as to lose sight of these brains' distracting peculiarities, and consequently to gain a focus on what is important about these Earthly brains. That is, we must view the brain from 25,000 feet up--or, from very high up. At this height, the details of the brain are lost, whether they be ion channels, intricate neural connectivity patterns, or pharmacological effects. The background required of a researcher who wishes to study the brain from this high level is therefore not traditional neurobiology, neuroanatomy, psychology or even computational neuroscience (the latter which almost always focuses on modeling specific, relatively lower-level mechanisms thought to occur in the brain). What is needed is a training in mathematics, computer science and physics, and even an appreciation of conceptual limits from philosophy.
This book serves two purposes. One aim is to illustrate a number of high-level approaches to brain science. These range from explanations for why natural language is vague, to solutions to the riddle of induction and applications to issues of innateness, to the use of probability and decision theory in modeling perception, to the inevitability of visual illusions for any animal with a non-instantaneous brain, to the morphology and scaling behavior of many aspects of nervous systems, and finally to issues concerning universal laws governing hierarchical complexity in behaviors exhibited by brains. The second aim is to both encourage others to approach brain science from a high level and to provide, along the way, an introduction to some of the mathematical, computational and conceptual principles needed to be able to think about brains at a higher level.