This review is forthcoming from the American Philosophical Association's Newletter on Philosophy and Computers.
Review of George Dyson,
Darwin Among The Machines
Peter Suber, Philosophy Department, Earlham College
George B. Dyson, Darwin Among the Machines: The Evolution of Global Intelligence, Perseus Books, 1997.
Dyson's book is an argument disguised as an intellectual history. The argument is that all intelligence is collective, in the way that human intelligence emerges from the collection of unintelligent neurons, and that a global collective intelligence is now emerging from the growing interconnections among human beings and their machines. The history traces the rise of computation and thinking about machine intelligence from Hobbes to the present. The history is fascinating and detailed. The thesis about collective intelligence is fascinating but lacking the detail which would make it more than merely suggestive.
Each chapter spends most of its time on figures and developments in the history of computation and AI. Some are major figures in every telling of this story, such as Leibniz, Babbage, Boole, Turing, Gödel, and von Neumann. Some are major figures in other fields whose contribution to computing or AI has not been widely recognized, such as Hobbes, Robert Hooke, Samuel Butler, Charles Peirce, and H.G. Wells. And some are nearly forgotten figures with prescient or eccentric ideas, or both, whom Dyson has painstakingly resurrected to fill in the story, such as Alfred Smee, Allan Marquand, Lewis Fry Richardson, Julian Bigelow, Nils Barricelli, William Ross Ashby, Olaf Stapleton, Frederic Myers, and Paul Baran. There are other histories of computation which tell the main story well, perhaps with more technical detail than Dyson, but they omit the quirky minor figures Dyson includes here. I know of no other history of pre-20th century thinking about machine intelligence as rich, detailed, or surprising as this one.
As parts of the history, Dyson eventually discusses many examples of collectives with emergent problem-solving abilities which exceed the abilities of their separate components. Apart from the example of brains, he considers Hobbes' Leviathan, networks of human calculators, free-market economies, and the evolution of multi-celled organisms.
At the end of each chapter, Dyson draws lessons for his growing thesis about evolution, collectives, emergence, life, and intelligence. Because these chapter-closing reflections explicate the book's title and subtitle better than the historical story, one concludes that Dyson is telling the history for the sake of this thesis, and not the other way around.
These closing reflections are intriguing, plausible, and rich with implications, but always brief, sometimes oracular, and not closely tied to evidence in the chapter. As a colleague of mine likes to write in the margins of student papers, these propositions are "important if true." But Dyson does not give them the elaboration and support they deserve, and which he does give to his history. One gets the sense, with some frustration, that the primary thesis gets secondary attention.
This is my only complaint about the book. The thesis is too interesting and important to leave in the impressionistic state in which he leaves it. And unlike impressionistic authors, Dyson can do better, as we know from the other 98% of the book. This is the complaint of a reader who was persuaded to take the idea seriously. I'd like to see the author do the same.
This criticism may be unfair. It's possible that Dyson's thesis cannot yet be given the detailed exposition and argument to which philosophers are accustomed. Dyson has done a service to marshal enough evidence to make the thesis plausible and to make us vigilant for clues relevant to its truth. But I'm afraid that philosophers will be impatient with the brevity of Dyson's treatment of his own thesis, and be impatient roughly to the extent that they find his thesis worth exploring. Nevertheless I recommend this book to philosophers, in part to increase the chance that others will pick up this thesis where Dyson left off.
A large part of the appeal of this book, apart from the very engaging history and very suggestive thesis, is its nuanced, alternative vision of machine intelligence, which makes the sci-fi scenarios still present in the philosophical literature look superficial. Dyson gives no attention to the common view of stand-alone artificial intelligence, which might pass a Turing Test, be installed in a robotic body, and interact with human society and other intelligent robots. He finds it more plausible and fruitful to think of human beings and machines in a symbiotic relationship. Already we depend on machines and machines depend on us, and this mutual dependence will become more complex and far-reaching over time. Both poles of this symbiotic relationship evolve, and together we co-evolve. Dyson wants to break down the distinction between technology and biology, and see machines, together with the human-machine symbiosis, as subject to natural selection. The examples of other collectives which have evolved to surpass their components in survival strength and intelligence suggest that the future of machine intelligence is more collective than isolated, more natural than artificial, and more a vector or society of human intelligence than a rival of it.
Peter Suber, Earlham College
Earlham College, Richmond, Indiana, 47374,
firstname.lastname@example.org. Copyright © 2001, Peter Suber.