Book Buzz

Darwin Among the Machines: The Evolution of Global Intelligence
by George B. Dyson; Perseus Books, 1997
Reviewed by Edward Piou, 8/22/1999

"This is a book about the nature of machines," George Dyson writes in his preface. As such, Darwin Among the Machines serves as a useful tool in imagining the world of computers which we seem to spending more and more time living in. It is also a wake-up call, to recognize the signs of intelligence among computers which have already appeared. From chapter to chapter, the author jumps across disciplines and time frames, to sometimes disorienting effect. In the end, when it all comes together, an enlightening perspective on human, societal, and computer development emerges.

Dyson manages to cover a lot of territory. He begins on a large scale, with Thomas Hobbes' theories of human societies as intelligent entities, and the evolutionary arguments which raged from the time of Erasmus Darwin and his contemporaries to that of Erasmus' more-famous grandson, Charles. He uses these arguments to put forth the theory that the evolutionary process exhibits intelligent behavior. It is intelligent in that it solves particular problems in particular environments - the problems of ensuring survival of certain codes (DNA), or of the hardware that carries those codes (bodies), depending on your point of view.

He then moves on to the evolution of computers and calculating machines - from Leibniz's basic calculator of the 17th century, through the well-known history of ENIAC earlier this century, up through the digital computers of today. Over time, the problems posited for the machines became more and more complex; as a result, the computers become more and more advanced. Though they don't build themselves (yet), Dyson demonstrates that there is still an evolutionary process at work here. Those software and hardware schemas which worked to solve problems gave rise to new, improved ones, while those that didn't work either disappeared completely or languished in barns and cellars. Dyson's entertaining narrative demonstrates that whether the computers themselves fit any definition of alive or intelligent, the process by which they change over time is shown to be an intelligent one, as is the evolution of any species.

After dealing with natural evolution and the history of computers, Dyson turns to the creation of networks. This is a subject he already touched on, in describing the nature of the brain, and in explaining theories that the first "real" AIs will be created using neural networks. It should come as no surprise, then, when he talks of both the networks of work and money, that make up an economy; and the networks of computers, that make up the Internet; as intelligent, and evolving - constantly improving, in the sense that they assure their own survival. The real difference between these artifical networks and the networks of cells that make up our bodies lies in the approach. Nature took a bottom-up approach in constructing life; we are taking a top-down approach in our creation of artificial life.

It's hard to do justice to the scope and depth of Dyson's theories in a one-page review. (If it were possible, Dyson wouldn't have taken an entire book to explain them.) The book is not always engrossing - it suffers a bit from the way it seems to jump from one topic to a seemingly unrelated other. But, just as seemingly random events in the natural world gave rise to some very valuable emergent behavior, the numerous puzzle pieces Dyson puts together give rise to a coherent, thought-provoking whole. This book probably won't help you do your job better; but it will help you understand the world you're living in.

(Full disclosure: the reviewer sometimes wonders why people think that machines can't, and why we think that other people can.)