Darwin Among the Machines (32 page)

Unfortunately, cryptographically secure transmission of electronic mail and electronic funds can also be used to hatch terrorist plots, launder ill-gotten gains, and evade taxes by leaving local authorities behind. With good intentions but diminishing success, the United States has tried to keep cryptography under government control. The argument over control of cryptography has been going on since codes began. At the end of his treatise on telecommunications and cryptography published in 1641, Petty's colleague John Wilkins considered the abuse of cryptography by criminal conspirators, concluding that “if it be feared that this Discourse may unhappily advantage others, in such unlawfull courses: Tis considerable, that it does not only teach how to deceive, but consequently also how to discover Delusions. And then besides, the chiefe experiments are of such nature, that they cannot be frequently practiced, without just cause of suspicion, when as it is in the Magistrates power to prevent them. However, it will not follow, that every thing must be supprest, which may bee abused. . . . If all those usefull inventions that are lyable to abuse, should therefore be concealed, there is not any Art or Science, which might be lawfully professed.”
³³

Wilkins distinguished between digital coding and pulse-frequency coding, providing an extensive catalog of cryptographic techniques. Noting that “because Words are onely for those that are present both in time and place,” he recognized the power of coded information to penetrate barriers not only of distance but of time. He was the first to observe that high-speed data communications would allow the noon price of something in London to be communicated westward before it was noon somewhere else. “Suppose (I say) this Messenger should set forth from London, in the very point of noon,” noted Wilkins, envisioning a relay of optical signals, “hee would notwithstanding, arrive at Bristow before twelve of the clock that day. That is, a Message may be by these means conveyed so great a distance, in fewer minutes then those which make the difference between the two Meridians of those places.”
³⁴
Time is money. It was
the ability to convey market information, trading orders, and fund transfers even slightly ahead of the competition that led to the proliferation of cryptographically secure telecommunications channels by which electronic money has spread throughout the world.

According to a 1995 estimate by the International Telecommunications Union, 2.3 trillion dollars circulates electronically every twenty-four hours—equivalent to 180,000 tons of gold, or a 1,500-mile stack of hundred-dollar bills. Electronic currency has diffused outward from the central banking networks to penetrate the street corner, the desktop, the telephone system, and a host of card-based payment systems, smart and dumb. Banks are becoming networks, and networks are becoming banks. Having seen corporate mainframes replaced with desktop computers, some analysts believe the powers of large banking institutions will be similarly overturned. But the banks are here to stay. “Commercial banking has been around over 600 years,” consultant Eric Hughes has explained. “Computers are less than 60 years old. Microcomputer software companies are 20 years old and still reinvent the wheel. Assuming a convergence, who do you think will learn the other's business first?”
³⁵

The result will be more money, faster money, and money more tightly coupled to things, network architecture, people, and ideas. The scales are shifting both in distance and in time; the intelligence a large corporation once gathered for its annual report is now available to any small business using a personal computer to manage its day-to-day accounts. “We felt that the distinction between micro- and macro-economics, while appropriate in a non-computer age, was no longer necessary,”
³⁶
remarked economist Gerald Thompson, recalling his final collaboration with Oskar Morgenstern in 1975, two years before Morgenstern's death.

Money is a recursive function, defined, layer upon layer, in terms of itself. The era when you could peel away the layers to reveal a basis in precious metals ended long ago. There's nothing wrong with recursive definitions. (Definition of recursive: see recursive; or, Gregory Bateson's definition of information as “any difference that makes a difference”—the point being that information and meaning are self-referential, not absolute.) But formal systems based on recursive functions, whether in finance or mathematical logic, have certain peculiar properties. Gödel's incompleteness theorems have analogies in the financial universe, where liquidity and value are subject to varying degrees of definability, provability, and truth. Within a given financial system (i.e., a consistent system of values) it is possible to construct financial instruments whose value can be defined and
trusted but cannot be proved without assuming new axioms that extend the system's reach. As Gödel demonstrated for logic and arithmetic, there are two sides to this. No financial system can ever be completely secure and closed. On the other hand, like mathematics or any other sufficiently powerful system of languages, there is no limit to the level of concepts that an economy is able to comprehend.

All free-market economies show signs of intelligence, to varying degrees. Conversely, close inspection of many mechanisms we regard as intelligent reveals fundamentally economic systems underneath. When Oskar Morgenstern was asked to explain the power of game theory to a popular audience in 1949, he used the example of a simplified form of poker for two players, using a three-card deck and a one-card, no-draw hand.
³⁷
To determine all possible strategies for this game by brute-force computation requires two billion arithmetic operations. Simple economic systems are able to arrive at practical solutions to problems that are computationally difficult to solve. That brains in nature operate more as economies than as digital computers should come as no surprise. Indeed, economic principles are the only known way to evolve intelligent systems from primitive components that are not intelligent themselves. As Marvin Minsky explained in his
Society of Mind:
“You can build a mind from many little parts, each mindless by itself. . . . Any brain, machine, or other thing that has a mind must be composed of smaller things that cannot think at all.”
³⁸
Or, as Samuel Butler put it in 1887: “Man is only a great many amoebas, most of them dreadfully narrow-minded, going up and down the country with their goods and chattels.”
³⁹
The archetypal invisible hand of Adam Smith (“He intends only his own gain, and he is in this, as in many other cases, led by an invisible hand to promote an end which was no part of his intention.”)
⁴⁰
appears to be capable of building not only an economy, or a damage-resistant communications network, but a brainlike structure, perhaps a mind. “Probably the closest parallel structure to the Internet is the free market economy,” observed Paul Baran.
⁴¹

The incubation of intelligence within a network requires an exceptionally fluid, arborescent structure and the infiltration of this architecture by a statistical language analogous to the primary statistical language that von Neumann identified as the machine language of the brain. At one level, this language may appear to us to be money, especially the new, polymorphous E-money that circulates without reserve at the speed of light. E-money is, after all, simply a consensual definition of “electrons with meaning,” allowing other levels of meaning to freely evolve. Composed of discrete yet divisible and
liquid units, digital currency resembles the pulse-frequency coding that has proved to be such a rugged and fault-tolerant characteristic of the nervous systems evolved by biology. Frequency-modulated signals that travel through the nerves are associated with chemical messages that are broadcast by diffusion through the fluid that bathes the brain. Money has a twofold nature that encompasses both kinds of behavior: it can be transmitted, like an electrical signal, from one place (or time) to another; or it can be diffused in any number of more chemical, hormonelike ways.

Money has the self-reinforcing tendencies and semantic transparency that allow neural networks to work. The flow of money permeates all components of the network, strengthens frequently used connections, propagates backward, transforms local processing mechanisms, and encourages new connection pathways in response. This architectural plasticity allows neural networks to adapt, remember, and learn to predict events. Freely reversible financial gradients direct when and where new connections are formed and which connections die out. The flow of currency transports, integrates, and accumulates signals; a myriad of financial instruments function as neurotransmitters and bridge synaptic gaps.

“Neural processes are insulated from the extra-cellular fluid by a membrane only approximately 50 angstroms thick,” wrote semiconductor pioneer Carver Mead, explaining how to build integrated circuits modeled after the neural circuits found in our brains. “The capacitance of this nerve membrane serves to integrate charge injected into the dendritic tree by synaptic units. Much of the real-time nature of neural computation is vastly simplified because this integrating capability is used as a way of storing information for short periods—from less than 1 millisecond to more than 1 second. There is an important lesson to be learned here, an insight that would not follow naturally from the standard lore of either computer science or electrical engineering. Like the spatial smoothing performed by resistive networks . . . temporal smoothing is an essential and generally useful form of computation.”
⁴²
Whether conveyed by bullion or binary numbers, accounts accumulate incoming currency over various periods and release outgoing currency at intervals more or less closely related to patterns generated by the currency coming in. In an age when nanoseconds count, it is easy to forget that the components of a neural net must have some temporal delay, however small, to allow the network to compute.

In drawing these analogies, what of the data that now flood the telecommunications net: pictures, sound, video, interactive data
communications, encyclopedias of text? All this traffic means something to somebody, and some of it advances our sciences, our culture, and our arts, but is it the stuff of meaning (or the measure of a utility function) across the system as a whole? Maybe or maybe not. What counts is not so much the data that flow in any given direction, but the money that flows the other way. In the coalescence of the software, banking, and telecommunications industries, we are spawning the precursors of collective digital organisms that will roam the network like social insects, sending packets of digital currency back to their nests. The push toward interactive communications over the Web is aimed not at delivering content to the consumer (this can be done already), but at delivering money, in real time, the other way. Electronic money allows organizations to do things and immediately sense the results.

This was the original premise of purposive systems as expounded by Norbert Wiener and Julian Bigelow in 1943: intelligent behavior evolves as a consequence of the ability to measure and keep account of the effects of a given signal through feedback loops that return a message signifying the magnitude of the result. These principles are common to automatic anti-aircraft guns firing at a moving target, neurons seeking to make the right connections inside a brain, laboratory animals facing a maze, corporations facing a free-market economy, or any other situation where it is possible to place a value on an objective at which to aim.

The goal of life and intelligence, if there is one, is awkward to define. A general aim can be detected in the tendency toward a local decrease in the entropy of that fragment of the universe considered to be intelligent or alive. This is a measurable way of saying that life and intelligence tend to organize themselves. Order, however, is only available in limited quantities, at a certain price. Organization can be increased or created only by absorbing existing sources of order (by eating other creatures as food, joining them in symbiosis, or by photosynthesis exploiting the ordered energy of the sun) or by shedding disorder (by excreting waste, radiating heat, or learning from experience through the attrition of less-meaningful connections in the developing infant brain). In human society, money serves to measure and mediate local markets for decreasing entropy, whether it measures the refinement of an ounce of gold, the energy available in a ton of coal, the price of a share in a multinational organization, or the value of the information accumulated in a book. We invented the science of economics, but economy came first.

In 1965, twenty years after the disbanding of Alan Turing's crew at Bletchley Park, Irving J. Good published his speculations on the
development of an ultraintelligent machine, later described as “a machine that believes that people cannot think.”
⁴³
Central to the development of an indisputable mechanical intelligence is the question of what meaning is and how meaning is evolved. In Good's analysis, meaning and economy are deeply intertwined; where there is meaning, there is an economy of things representing information (or information representing things) by which the meaning of things can be evaluated and from which meaningful information structures can be built. “The production of meaning can be regarded as the last regeneration stage in the hierarchy,” noted Good, “and it performs a function of economy just as all the other stages do. It is possible that this has been frequently overlooked because meaning is associated with the metaphysical nature of consciousness, and one does not readily associate metaphysics with questions of economy. Perhaps there is nothing more important than metaphysics, but, for the construction of an artificial intelligence, it will be necessary to represent meaning in some physical form.”
⁴⁴