How Nature is Transforming Our Technology and Our Lives
272 pages, ISBN 0-7432-0447-6, Simon & Schuster, New York, 2001.www.simonsays.com
Reviewed by J. M. Haile, Macatea Productions, http://www.macatea.com/
In Digital Biology Peter Bentley describes how computer scientists are using behaviors in the natural world as guides to developing new kinds of computational software. This is an interesting and provocative instance of role reversal: instead of our trying to use computers and other devices to study Nature, the strategy is to use Nature to help design, develop, and apply new kinds of software. Why might we want to deploy such a strategy? Because there is a class of behaviors that we comprehend yet hardly understand, but which Nature has developed and employed in a variety of situations: the dynamics of complex systems. A complex system is composed of many separate entities that can interact and adapt; adaptations arise though feedback from interactions and response to external perturbations.
One of the surprises is that the rules for direct interactions can be (in fact, usually are) very simple; nevertheless, feedback ensures that even simple rules are implemented nonlinearly. So, complexity arises not from complicated rules, but from the numbers of possible interactions among entities and from the strength of the feedback that is created in response to interactions. Computationally, therefore, we can create a set of digital entities, network them together so they can interact, provide for feedback, and then let the system evolve. The resulting dynamics might carry the system to a stationary state, in which nothing interesting happens, or to chaos, in which nothing coherent happens, or to the edge of chaos, in which complex behavior emerges.
The bulk of Digital Biology is divided into six chapters in which Bentley first describes the behavior of a complex system in the physical world, then he shows how selected features from those systems are used to design, develop, and apply software in the digital world. Here is a summary of the major connections Bentley makes between the natural and digital worlds:
|Natural Process||Software Analog|
|Evolution||Evolutionary algorithms, such as genetic algorithms, which evolve solutions to problems|
|Brain Function||Neural networks|
|Insect Behavior and Swarm Intelligence||Distributed artificial intelligence, which uses collections of agents to seek solutions to problems|
|Plant Growth||Chaotic control systems; creation of fractal patterns using L-system grammars|
|Immune System Response||Anti-virus software; intrusion detection systems|
|Human Development||Cellular automata, Boolean networks, DNA computing|
For me, the more interesting parts of each chapter are those devoted to Nature, perhaps because I was already familiar with several of their analogs in the digital world. A particular strength is that in describing the physical systems, Bentley is consistently careful to articulate the rules of interaction. Here are some examples.
Bentley starts (in Chapter 2) by arguing that complex systems created in software are in fact universes; this then raises the question as to the definition of a universe. Bentley offers this two-part definition: (1) A universe is a set of rules that determine outcomes, and (2) At least one of the rules must rely on the existence of some other universe. To illustrate rule (2), note that any explanation of our physical universe relies on something external to our universe: a deity or energy/matter before the big bang or "turtles all the way down." Under his definition, complex digital systems are certainly universes. Whether this definition extends to physical universes, I'm not prepared to say, but I am prepared to accept these ideas as a working hypothesis.
Chapter 3 features evolution as the complex natural dynamics to be mimicked. Underlying evolution are three fundamental processes: (1) reproduction with inheritance, (2) selection, and (3) variation. These are of course familiar, but Bentley does a good job of explaining and illustrating each. And he is careful to point out logical pitfalls to be avoided. For example, slow incremental change is not necessarily evolution, evolution does not necessarily lead to improvement, and in evolution the only things that change are genes.
For this review, let's skip over Chapter 4 on the human brain. The brain is certainly a complex dynamic system—the most complex in the known universe—and under Bentley's definition it is itself a universe. Much has been written about the workings of the brain, about the distinctions between brain and mind, and about the software analog: the neural net. So the content of this chapter may be familiar to you.
Less familiar might be the content of Chapter 5: insects and in particular, insects that build nests or hives. Now a single ant can hardly be said to have any intelligence: it's little more than a few neurons on legs. But when several thousand ants get together, they self-organize into a collective intelligence that directs construction and maintenance of the nest, minimizes the path length to known food sources, and locates new food sources. How does such a collective intelligence function? Bentley postulates four processes: multiple interactions, positive feedback, negative feedback, and amplification of fluctuations. To illustrate the last of these, consider one ant searching for a new food source: she proceeds by a random walk. If a new source is found, she returns to the nest, laying down a pheromone trail so that other ants can find the new source and exploit it. The random fluctuation of one ant exploring on its own is amplified so the result can benefit the nest.
Apparently even simpler are the rules for swarm intelligence which governs the movements of flocks of birds, schools of fish, herds of buffalo, crowds on city streets, traffic on interstate highways. Only two rules are needed: (1) Individuals are attracted to one another, forming a group, and (2) when the group moves, don't collide with your neighbor. Wondrously, these two rules are sufficient to explain, for example, the complex, high-speed, synchronized motions displayed by schools of tropical fish.
The last natural system I'll mention here is the body's immune system, which Bentley discusses in Chapter 7. The human immune system is composed of a vast number of different organisms that interact chemically to protect the body from alien pathogens. The successful functioning of the immune system depends on four characteristic features: diversity, specifity, memory, and tolerance. Diversity is needed to counter the many different kinds of pathogens that might be encountered; specifity is needed to concentrate defenses when a particular pathogen is identified; memory is needed to recognize when a new threat is actually another instance of an previously encountered pathogen; tolerance is needed to allow antibodies for new pathogens to remain in the body to protect against future attacks. The behaviors of the immune system are so sophisticated and versatile that, says Bentley, we might categorize the immune system as a second brain. And given the complexity of its behaviors, perhaps we should not be surprised that the system sometimes fails by over-reacting, producing various autoimmune diseases.
Digital Biology is written in an informal, conversational style and it is filled with original, surprising metaphors and analogies. Substantial thought and talent are required to present highly technical material in such an accessible way. But a danger is that, unguarded, a breezy presentation can slide into glibness, as happens when Bentley tries to distinguish equilibrium from chaos: "… consider water. If there is insufficient energy and change in the molecules of the water, it settles into equilibrium—it solidifies to ice" (p. 112). But anyone who has had an undergraduate course in general physics or chemistry should know that equilibrium is independent of phase: ice, liquid water, steam may all be in equilibrium with their surroundings and each of these phases may be in equilibrium with one of the other phases.
More seriously, Bentley presents a wholly one-sided view, full of enthusiasm for the potential benefits that can be expected to arise from the digital universes he discusses. But we have learned that every technology has a downside: benefits are balanced at least by accompanying disadvantages and in some cases by real dangers. Here Bentley is disappointingly silent.
These criticisms aside, Digital Biology is a fine book that can expand your awareness of how our natural and digital universes are becoming entangled.
(jmh 04 October 06) © 2006 by J. M. Haile. All rights reserved.