A future full of super-intelligent machines is equal parts sci-fi clichi and computer-science holy grail -- grist for both Frankenstein fears and programmer dreams. But it's never been quite clear how that future will arrive. Not long ago, quite a few otherwise respectable scientists believed that artificial intelligence would spring forth fully formed in the lab, like Athena from the brow of Zeus, out of cleverly concocted code.
But there's another way, according to author George Dyson -- the evolutionary way. In his ambitious new book, "Darwin Among The Machines: The Evolution of Global Intelligence" (Addison-Wesley, 286 pages), Dyson suggests that a new kind of intelligence may one day emerge -- undirected and unplanned -- from the incomprehensible complexity of an interconnected, wired world.
The Net is where the intelligent action will be. In the Dysonian schema, fragments of software replicating across the Internet are analogous to strings of DNA replicating in living cells. Evolutionary pressure provides the drive: What works survives; what doesn't gets deleted, either by us or on its own.
So be prepared -- there's a new primordial soup in town, bubbling over with restless code, seething right toward the boiling point of sapience. And this new sapience won't be purely digital; the emergent new order will be defined by a commingling of the silicon and the biological, by a collective merging of human and hardware. That's both a promise and a threat.
Dyson was to the computer manner born. As a child he grew up on the campus of the Princeton-based Institute of Advanced Studies, where his mathematical physicist father, Freeman Dyson, hobnobbed with such luminaries of the computer world as John von Neumann, and where abandoned fragments of the earliest known computers rusted away in nearby barns. Dyson's sister, Esther, is the publisher of Release 1.0, a pricey newsletter that reports from the cutting edge of current digital developments.
Freeman Dyson is perhaps most famous for his theory that all intelligent species will inevitably progress to a point where they enclose their home sun with a vast shell -- a "Dyson sphere" -- in order to maximize living space. On their face, George Dyson's theories about emergent global intelligence might sound equally fantastic. No matter how persuasive the arguments, it's never clear how routers, high-speed telecommunication lines and torrents of data add up to a new intelligent species.
But that doesn't harm the book, because "Darwin Among the Machines" is as much a work of history as of speculation. In a phone interview from his kayak repair workshop in western Washington state, Dyson himself called it "a catalog of beginnings." Dyson explores the roots of computer design, of distributed networks and of the very concepts of artificial intelligence and even evolution.
It's a sound strategy. Just attempting to define terms such as "life" or "intelligence" can confound the most brilliant of minds. Since, as Dyson notes on several occasions, the only thing that can truly describe a complex system is the system itself, it makes sense to focus one's attention on first steps, rather than the goal line.
"You can only sort of hint at things," says Dyson. "How do you explain the explanations? You have to rely a great deal on analogies ... But what it all comes down to is that the more we understand about the way our brains work the more we find that evolution has a lot to do it."
And we also find, when we start at the beginning, that there is nothing new under the sun. From the earliest days of evolutionary theory, Dyson shows, fascinated observers speculated on the possibility that the same evolutionary laws that determined natural development might also shape emergent machine intelligence. As early as 1865, Charles Darwin's contemporary Samuel Butler wrote that "although we grant that hardly any mistake would be more puerile than to individualize and animalize the at present existing machines ... yet we can see no a priori objection to the gradual development of a mechanical life, though that life shall be so different from ours that it is only by a severe discipline that we can think of it as life at all."
From Butler onward, Dyson traces the history of the notion that it's possible to bridge the gap between nature and machine. It's all a matter of changing one's frame of reference.
In Dyson's view, complex living organisms evolve out of the symbiotic
cooperation of simpler organisms -- "symbiogenesis." Symbiogenesis "assumes
that the most probable explanation for improbably complex structures
(living or otherwise) lies in the association of less complicated parts."
"All intelligence is collective," writes Dyson. "This intelligence --
whether that of a billion neurons, a billion microprocessors, or a billion
molecules forming a single cell -- arises not from the unfolding of a
predetermined master plan, but by the accumulation of random bits of wisdom
through the power of small mistakes."
The Net is the ultimate forum for collective intelligence, as well as
for unlimited experimentation. It has long passed the point, as a complex
system, at which it could be definitively mapped or succinctly summed up.
And every signpost points the way to increasing volatility -- toward an
environment in which increasingly mobile and autonomous conglomerations of
code migrate from node to node, constantly mutating and reshaping
themselves in response to circumstantial demands.
Just as natural selection resulted in the range of highly adapted
species existing in the world today, so too will evolutionary pressure
ensure that tomorrow's mobile code exhibits capabilities that will astound
and baffle us. Indeed, the packet-switching protocols of the Net itself
are, writes Dyson, a "particularly virulent strain of symbiotic code ...
Successful code is now executed in millions of places at once, just as a
successful genotype is expressed within each of an organism's many cells.
The possibilities of complex, multi-cellular digital organisms are only
beginning to be explored."
The possibilities are not confined to digital limits, either. "The Net
wouldn't function for a minute if there weren't all those people sitting at
their desks," says Dyson.
The Net is as much flesh-and-blood as it is chips and fiber optics.
Humans are indispensable to the new intelligence equation. The emergent
collective intelligence isn't just hardware and software -- it includes us.
And why not? If all of life up to now has been a series of ever more
intricate joint ventures, who is to say this latest synthesis is
unworkable? The principle has been well illustrated; all that's changed is
the speed at which it can all happen.
"The cooperation between human beings and microprocessors is
unprecedented, not in kind, but in suddenness and scale," writes Dyson.
"Now, in the coalescence of electronics and biology, we are forming a
complex collective organism composed of individual intelligences."
Some humanocentric nativists might well be dismayed at the prospect.
They wouldn't be the first to eye the yoke of the machine with suspicion.
Again, Samuel Butler staked out the territory. In an 1863 essay titled
"Darwin Among the Machines," Butler noted that "the machines are gaining
ground upon us; day by day we are becoming more subservient to them; more
men are daily bound down as slaves to tend them; more men are daily
devoting the energies of their whole lives to the development of mechanical
Submission to the machine -- from "Frankenstein" to "The Terminator,"
it's a recurring nightmare. But from Dyson's symbiogenetic perspective,
Butler misses the point. We're not necessarily bound to become slaves, but we
could well be partners. We're all in this together. There is no Net without
humanity, no global intelligence without cooperation.
We could have an important role to play. Then again, we could be written
out of the script.
"For the single-celled organisms, the advent of multi-cellular
organisms, such as the nervous system, was the end of freedom," says Dyson.
"We're at that stage now. Whether it is good or bad completely depends on
how you look at it."
"Sure, we feel threatened," he adds. "We're loyal to our own life form.
Americans feel threatened by foreigners, Earthlings feel threatened by
Martians, and protoplasmic forms of life ought to feel threatened by other
forms of life."
The real danger is not necessarily that machines will grow too smart but
that human intelligence may atrophy. Evolution isn't necessarily on our
side. Just because we are intelligent now doesn't mean that we always will
be. Even as we get subsumed into a greater intelligent cooperative,
individually, we may become dumber.
Evolution, argues Dyson, moves forward by constantly dispensing with
unnecessary baggage. He is fond of pointing out that human babies are born
with more neurons than a fully grown adult. As they grow older, they shed
irrelevant neurons, while reinforcing the connections and neural pathways
that make sense.
From babies, Dyson jumps to kayaks. It seems his true passion is
repairing Inuit watercraft. It's part of a struggle to ensure that the
human race doesn't get dumber. "There's a 10,000-year-old tradition
of kayak design which is in danger of being lost," he says. "I want to
prevent that if I can."
"To build a kayak," he writes, "you assemble a skeleton and then give it
a skin that allows it to float, just as the architectural framework of a
computer is fitted, by evolution or by design, with an envelope of code. To
build a dugout, you grow a tree and then remove everything, one chip at a
time, except the boat. This is how nature creates her intelligences, by
spawning an overwhelming surplus of neurons and then selectively pruning
them to leave a network that, if all goes well, becomes a mind."
The Net is a new spawning ground for software code proto-neurons. There
is no limit to the possibilities it may engender. But we should beware lest
our own intelligence, through willful ignorance or simple inaction, becomes the
part that gets pruned.