Review: Vehicles by Valentino Braitenberg

Vehicles: Experiments in Synthetic Psychology by Valentino Braitenberg
Cambridge, MA: The MIT Press, 1996, x + 152 pp., $16.23, ISBN 0-262-52112-1

I first heard about Vehicles from the recommended reading list of Dr. Brian Wandell, Principal Investigator of the VISTA Lab at Stanford University in Menlo Park, CA. It is an excellent read that creatively challenged the way I approach studying the brain, and I highly recommend it.

The simple premise of Vehicles is beautifully encapsulated in the first chapter entitled "Let the Problem of the Mind Dissolve in Your Mind" and that is exactly what author Valentino Braitenberg helps the reader do. The first eighty-five pages are devoted to building a very abstract model of a mind, and the subsequent fifty pages take this model apart and examine how it relates directly to what we know about the biology of the brain.

The model consists of fourteen "vehicles" possessing increasingly more complex hardware. Through an incredibly intuitive and believable progression, Braitenberg works upward from a simple sensor attached to a motor to tiny animated machines with very human-like behavior. Although their internal workings are simple to understand, the final vehicles perform surprisingly complex actions indicating memory, foresight, and free will.

This part of the book is manageable even for someone with no background in science, yet is thought-provoking enough to inspire the imagination of even the most seasoned neurobiologist. It is, in a word, beautiful.

Following this seductive creation of a human-like mind out of simple circuits are fifty pages of biological notes. The biological explanations are a bit dense and technical and often inconclusive.

One of the recurring themes in this section is the lack of sufficient information to determine the validity of the Vehicles model. While direct biological evidence for the simpler circuits is easy to find, the more advanced circuitry is still largely hypothetical given our current understanding of the brain.

While this is definitely a disappointing ending, it is still possible to appreciate the primary message of the book: synthesis is vastly easier than analysis. For all our difficulty in figuring out the inner workings of the brain, the human mind may turn out to be simpler than we imagine.

Review: The Mind Within the Net by Manfred Spitzer

The Mind Within the Net: Models of Learning, Thinking, and Acting by Manfred Spitzer
Cambridge, MA: The MIT Press, 2000, xiv + 359 pp., $29.00, ISBN 0-262-19406-6

The Mind Within the Net was recommended to me by Professor Paul Kieffaber at the College of William & Mary in Virginia. The book provides an excellent review of neural networks and their applications that is both accessible and surprisingly thorough; I highly recommend it.

In The Mind Within the Net, German psychiatrist Manfred Spitzer guides the budding neuroscientist from the basic workings of a single neuron to an intuitive sense for the emergent properties of whole networks. Spitzer takes care to provide crucial background information at each step and each new concept emerges organically out of the previous sections. Yet despite its accessibility the book smoothly tackles many formidable concepts in neuroscience, including:

• mechanisms for selective attention
• declarative and working memory
• information storage in complex networks
• the development of language

The result is a lucid explanation of how structure relates to function at every level of the brain.

The first half of the book is designed to help the reader build a tool-kit of different types of basic network structures that can be combined to produce more complex brain functions. In addition to explaining the biological mechanism for each network, the book is peppered liberally with visual illustrations and examples from research. 

The tool-kit contains four basic types of networks:

• Hebbian Learning (Long-Term Potentiation)
• Kohonen networks (center-surround organization)
• Hopfield networks (attractor states)
• Elman networks (recurrent states)

Along the way Spitzer also takes time to demonstrate data generalization, type formation, and the importance of noise within the system. He uses these concepts and the network structures above as a solid foundation for the second half of the book, in which he explores applications of neural nets in understanding brain dysfunction.

The applications section covers a number of neurological disorders that have been the subject of significant study in recent years: schizophrenia, autism, learning disabilities, phantom limbs, Alzheimer's disease, and depression. In each case Spitzer examines how small changes in the parameters of normal networks can lead to large qualitative changes in the network's function as a whole.

Although the examples Spitzer provides are not intended as definitive etiologies of these diseases, they do give the reader a sense for for how network level disorders manifest and how they are studied.

By working both forward (creating networks that mimic biological structures) and backward (deducing network structure from DSM-IV symptoms) Spitzer successfully argues that neural networks are an integral tool in our quest to better understand the brain.