Zoroaster's Fire

In an earlier chapter, we looked at how symbol-manipulating machines run up against a kind of Babel Barrier when trying to create useful knowledge in a library of written words.

We saw that to surmount that barrier, we'd need a machine that could store and communicate knowledge without using language at all. We'd need an information machine.

An information machine would be completely unlike a computer. Computers rely on the brain of the programmer for information.

Information is the living, creative, adaptive principle that gives rise to life itself.

An information machine would have to be so complex and dynamic that Darwinian processes could work in it.

Zoroaster was an ancient Persian mystic who compared the nature of God to an eternal, uncontainable flame. Like tongues of Zoroastrian fire, an information machine would flicker and shimmer.

It would undergo constant evolution, almost as though it were alive. In contrast to the dynamism and fluidity of Zoroaster's fire, a language-based computer would seem rigid and chiseled in stone, like Moses' Ten Commandments.

What will an information machine look like and how will it work?

Touring knowledge space

Imagine a kind of "space" that stores knowledge. Knowledge isn't stored in this space as symbols that occupy specific places in computer memory or pages in books on shelves.

It contains no language of any sort. Instead, knowledge is stored as a complex web of relationships with other pieces of knowledge.

A piece of knowledge would never exist in isolation, unconnected to other knowledge in the space. In fact, it couldn't exist in isolation. A piece of knowledge exists only in terms of the connections it makes with other pieces of knowledge in the space.

Adding to or modifying any single part of a knowledge space potentially alters the entire space, although more closely related regions will more likely be affected than less closely related ones.

In a sense, knowledge space is "holographic" (cutting a hologram in two doesn't destroy the image projected by the hologram; rather, it just diminishes its resolution).

As they continuously shift their weights to adapt to newly added or modified knowledge, the connections are always in flux and in the process of regeneration.

The smallest discernible piece of knowledge in the space is like a point on the number line. The more closely related two pieces of knowledge are, the closer they are on the line.

Lines of related knowledge likewise lie side-by-side in a plane, also according to the way the pieces of knowledge are related. The planes of knowledge in turn form a 3-dimensional space. More complex relationships create higher dimensions of knowledge space, as high as you like.

A point of knowledge can be related to many other points of knowledge besides the ones in its immediate neighborhood. This makes the knowledge space twist and bend to bring distant points closer together, giving it topological and geometric features.

Venturing through a knowledge space, an explorer could study the knowledge stored in it and discover the relationships connecting it all together.

This would be a bizarre experience for anyone accustomed to working with language and text. It would be like traveling through someone else's mind -- the mind of someone who's read every book in existence.

The only language-based approach that's at all similar to a knowledge space is what's sometimes called "hypertext."

Hypertext is a way of arranging text or pictures so that there's no beginning or end to the presentation. It's like opening a book to any page whatsoever to begin reading without "missing" anything. Hypertext arranges text according to extensive cross-indices.

Think of browsing the Encyclopedia Britannica, as an example. Each article makes reference to other, related articles. It would be unnecessary to start reading at "Aardvark" just because it's the first article in the encyclopedia.

You could start anywhere, and there would be a very large number of paths you could take in reading all the articles. The path you choose depends on your interests and prior knowledge.

A knowledge space would be even more densely connected than hypertext because the indexing would be "complete." Every point in a knowledge space connects to every other through some network of paths.

A knowledge space couldn't be language-based because, as discussed previously, there would always be "gaps" in the indexing.

As an example of how a knowledge space might work, think of a machine that stores every recording of every piece of music known to exist.

Using a digital computer and an index, you could simply list all the pieces in chronological order, in alphabetical order by title, or even by composer with composers arranged by nationality.

You could accomplish that without using a knowledge space. A knowledge space can also relate the music by mood and affect, however. Not only can it relate whole pieces to each other, it can tie tiny parts of them together, as well.

Pieces, and fragments of pieces, cluster together as families in a knowledge space. Each chunk of music belongs to many families. Mood and affect are subjective, and vary from listener to listener, and so families of music chunks in knowledge space are fuzzy around the edges. They vibrate and quiver as new relationships form and dissolve dynamically.

Traveling through the "music space," and listening to the music as you go, you'd hear a dazzling cacophony of sounds. Fragments of pieces by Mozart would lie next to fragments of Keith Jarrett's jazz.

Each slice, or plane, of knowledge space presents music fragments that vary along specific gradients. Those gradients are actually too subtle to be adequately described using words, but as a loose example, one plane might be something like "sweetly melancholic ranging toward spritely melancholic."

The plane above it in knowledge space might be "sweetly wistful ranging toward spritely wistful." In some places, one plane may touch another plane, or even pass through it to touch a plane on the other side.

A true knowledge space doesn't just connect fragments of music according to how they relate to one another. It also connects individual notes according to how they relate to other notes in their pieces -- and in all other pieces of music.

In fact, a knowledge space connects even more than individual notes. It also connects the overtones and micro-tonalities out of which each piece of music is made.

The qualities of knowledge in a space vary continuously from point to point. They vary so slightly that it's barely possible to perceive the differences between neighboring points. To help visualize this, think of color squares used by artists.

Each square is like a page in a book with an assortment colored chips arranged in rows and columns. To make it easy to visualize, let's call the upper-left corner of the page "north," the lower-right corner "south," the lower-left corner "east," and the upper-right corner "west."

As you move toward the northern corner of the page, the chips get lighter in color. The northernmost chip is white. As you go south, they get darker, and the southernmost chip is black. Colors in the east, on the other hand, are more vivid than those in the west. The chips get more intense as you move from east to west.

Each page contains colors with a slightly different hue, ranging continuously across the entire visible spectrum from deepest red to pastiest blue -- without skipping a single distinctly perceivable hue. The pages are like planes in a knowledge space.

By arranging the pages in a stack so that you can look at several of them at the same time, you have a "color space."

Each chip differs from all those around it in the color space by the smallest degree that a trained artist with sensitive eyesight can detect. An untrained person would probably look at adjacent chips and believe them to be the same.

That same untrained person could still perceive the overall trends, however, by looking at chips along lines moving across pages or from page to page. A knowledge space works the same way. Each piece of knowledge is connected so tightly to every other piece that only very discerning persons can detect the difference between neighboring pieces.

Colors in a color space vary in only three ways -- north to south, east to west, and plane to plane. A color space is 3-dimensional. Knowledge in a knowledge space has an indefinite number of dimensions, however.

Semantic engines and the digital network

I believe that one day a universal Library containing everything published in all media will be fed into a massive knowledge space. I like to call this knowledge space -- together with the global computer and communications network -- the "City."

I think it's an appropriate name because it evokes the image of a bustling, crowded place where people meet to exchange ideas, to buy and sell, and to create the things they need and want.

London's one-square-mile financial district is also called the City, incidentally. Although it is really part of Greater London, "the City" refers to the international financial market located there.

In the same way, the City that I'm talking about here will be a forum or communications center, not a place. Just like today's phone system, the City won't be located anywhere in particular, although its equipment will be scattered all around the world, as inconspicuously as possible.

Just as with real cities, the City has both older and newer districts. The older parts of town are mostly the global computer network that I've been calling the Utility, which still uses programmable digital computers.

In general, the Old City is the only part of the City that is still digital. The newer parts of town house and support the universal knowledge space. Machines in the New City aren't von-Neumann-type computers at all.

They are non-programmable, massively parallel, analog computers. Some use light rather than electricity, and some are photo-electronic, a hybrid of both.

The Old City incorporates a technological descendant of today's telephone system. In the early stages of the City's development, voice communications are integrated with text, video, and data transmission.

The foundations have already been laid for an integrated network; it's called ISDN (Integrated Services Digital Network). ISDN provides a protocol, or set of standards, for worldwide telecommunications. Such communications require channels that can carry enormous amounts of data.

Finely resolved, real-time moving color images would never "fit" over today's phone lines. Fiber optical cables can solve this problem, however. This paves the way for "data highways" that help bear the load of the City's torrential communications traffic.

The digital machines in the Old City are still useful for providing precise answers to straightforward problems by doing calculations very quickly and accurately. Old City machines differ from today's computers mainly by the fact that they work in teams -- engineers refer to this as working in parallel.

A well-coordinated team of smaller machines can compute faster than a single, larger machine working alone. Many computers today are already being designed with parallel architectures. Old City machines have more elaborate parallel architectures than today's computers, however.

Although some 20th century machines may remain in service, the Old City employs computers that are overwhelmingly faster and more powerful than today's best.

The New City is a global system of connectionist machines. Those machines may be descendants of today's neurocomputers, although there's clearly an aching gap separating today's little neural-like nets from the system that can house humanity's collective knowledge space.

The machines in the New City are "semantic engines," able to manipulate concepts directly, without relying on any symbol systems at all.

They form the nucleus of the City as a whole, giving the City communications and knowledge-processing powers that we can only dream about today.

Complete connection

Machines in the New City -- the nucleus of the City -- all have several common characteristics. They're probabilistic rather than deterministic.

That means we can only guess what's going on inside them -- we can never know the exact state of the system at any moment. The Nucleus of the "cyber-City" is chaotic and non-linear, like a weather system.

Miniscule shifts in connections within one part it lead to wide-scale changes in the Nucleus as a whole. It's next to impossible to predict the entire state of the system as a whole because it's extremely sensitive to its conditions when measurements are initially taken.

Seemingly insignificant variations in the measurements lead to vastly different predictions.

Scientists who study chaotic systems have sometimes talked about the "butterfly effect." In the example of weather systems, something so slight as the beating of a butterfly's wings could set in a motion a chain of events that eventually lead to a great hurricane in another part of the world.

The Nucleus is in a constant state of disequilibrium, just like the earth's atmosphere.

The Nucleus has "complete connection." No part of it is isolated from the rest of the system. As an example of "complete connection," think of a town where everyone could operate all the appliances and lightswitches in everyone else's house without leaving their own house.

Another example would be a town where everyone was holding a single, simultaneous telephone conversation with every other person town.

The Nucleus is complex, open, and fluid-like enough for Darwinian, competitive selection to work in it.

It retains patterns well enough to accumulate filter structures -- filters of filters of filters, and so on. The New City truly is an informed system, not just a blindly automatic one.

Visual and aural input provide the City with a diversity of forms, just as genetic mutations supply variation in the biological world. In the same way that a child soaks up sensations from his or her environment, the Nucleus also can process raw knowledge about the world.

Much of that knowledge gets fed to it by an extensive network of sensors that have grown out of the older, digital districts of the City, such as the Utilitity.

Without raw sense knowledge, the Nucleus would be as robotic and uninspired as a conventional computer.

Human demand for more interesting and useful designs provides the basis for selection. Human demand for design, in turn, is influenced by the stress of natural selection on human culture.

The collision between the human survival imperative and the rigors of natural selection is what finally determines the shape of cultures. Demand for cultural knowledge shapes what gets selected in the New City's ecology.

The relationship between the City and the human environment is quite analogous to the relationship between DNA and the environment of the organism whose structure is determined by the DNA. The genetic patterns in the DNA molecule "communicate" with the environment of the organism that houses the DNA by using the organism itself as the medium.

In the same way, the City communicates with the human environment through the medium of human culture. In other words, the City is cultural DNA.

Any system that is open to the outside environment has a certain amount of chaos and randomness. No closed system can evolve, however. Chaos supplies the fuel for selection. Without chaos, a system will eventually settle into a useless, perfectly predictable stability.

In thermodynamic terms, that means it can do no useful work. In informational terms, a closed system is sterile and unable to evolve. Human culture is the doorway that opens up the City to the world-at-large.

That's the source of its spontaneity and unpredictability. Without the doorway of human culture, the City would either fall into a useless, repetitive state of equilibrium, or else it would choke on the anti-information that is a by-product of creating useful knowledge.

It would either starve to death or else suffocate in its own waste.

The digital membrane

The Nucleus will be unlike any machine that presently exists. It will be a kind of "eco-computer," function on many of the same principles that work in ecosystems. The Nucleus has no moving parts, but it is "inhabited" by a very large number of machines

Machines in the Nucleus are called mindlets. In a sense, mindlets are like computer software, which may exist simultaneously in any number of computers at once.

However, mindlets don't reside in a particular space in computer memory or on a computer disk. They inhabit a very special environment, called a reticularium (a Latin word meaning place for nets).

A reticularium, or ret, is made of a material through which a practically unlimited number of distinct electric or photonic circuits can be open at the same time.

Mindlets are patterns in the connections among those circuits -- called nets. Nets are usually built out of millions of circuits.

Points of connection between circuits or nets are called nodes. Because thousands, or even millions, of nets may be active at once in a ret, the number of distinct nodes is extremely large.

A ret is like a tangled nest of squirming circuits.

All nodes are not equally well-connected. Some nodes are very densely connected -- they have links with many other nodes far and wide.

Further, some connections are much stronger than others. When a node is rich in strong connections, it's a hub node. Hub nodes are centers of communication in a net.

Weaker, thinner connections are spindle nodes. Every net in the ret is connected with every other net through spindles, at least, and they often even share hubs, as well. Nets communicate with each other through their shared spindles and hubs.

Connections between nodes are continuously appearing and disappearing. As a result, hubs and spindles grow and shrink. The whole ret is far from equilibrium.

Stable centers of connective activity may emerge for awhile, but they eventually fade, yielding to newcomers. Sometimes, massive "mega-hubs" appear. 

They're like hurricanes. They hold so many powerful connections that they skew the connective activities of the entire ret.

At other times, the ret may be quite placid, with no outstanding configurations of connectiveness.

Mindlets are dynamic patterns of connections in the nets. Those patterns are continuously shifting. They shift so chaotically that the only way to keep track of them is probabilistically.

The regions where patterns are most likely to appear can be analyzed as though they were waveforms. Mindlets are probabilitistic waveforms in the patterns of connections. Mindlet waves can propagate through the ret and even interfere with each other, just as any other kind of wave can.

When mindlet waves meet crest-to-trough, they cancel each other. Otherwise, their amplitudes reinforce each other. Although the patterns themselves jump around unpredictably, their mindlet waves move smoothly along continuous paths.

Because the nodes and connections making up the patterns in the mindlet wave are distributed throughout the ret, the "movement" of the wave doesn't necessarily involve displacement or change of location in the ret itself.

Rets are like little jungles. Mindlets interact with each other in life-like ways. The informational principles of collection, selection, connection, and correction are at work in the ret as a whole.

Mindlets busily work to create useful knowledge out of the mess of knowledge continuously flowing into the City.

A churning storm of newly formed nets continuously appears and disappears in the ret. A new net will roam aimlessly around the ret until it encounters another net, which then either accepts it or rejects it.

When an established net rejects the new net, the new net may lose some of its circuits to the established net, which may cannibalize them.

After a few such unsuccessful encounters, the new net can lose so many of its circuits that it will completely disappear. If an established net accepts the new net, though, then the new net becomes integrated into the established one. That's how nets grow more complex.

An important principle in biology is biogenesis -- living things can descend only from other living things.

A similar principle is at work in the City. Nets can only arise from nets. In other words, digital signals cannot meaningfully inform a net. Although nets may only arise from other nets, complex nets can arise from simpler ones through Darwinian selection.

The surface where the Nucleus, which contains the rets, makes contact with the digital machines of the Old City is called the nuclear membrane.

The membrane is the most critical part of the entire City because it's the point at which the Babel Barrier is vanquished. On the outside of the membrane -- in the Old City -- all communications are digital.

Within the membrane -- in the New City rets -- communication is analog and non-deterministic. The membrane is the medium through which the two worlds communicate.

The membrane is sensitive to patterns in the stream of digital data. Digital machines that move freely through the global Utility can also communicate with the membrane. The membrane senses patterns and fluctuations in the data and excites then the mindlets within the Nucleus.

The mindlets then process those patterns by arranging themselves into structures that reflect the patterns passed by the membrane. Those structures are continuously shifting and recreating themselves because there is rarely only one structure to fit a membrane pattern.

Many competing structures may emerge and coexist for a long time before one prevails over the others.

When a digital machine processes language or data, it passes its results to a higher level machine, which may process them further, and so on.

The buck stops at the Nucleus, however. There's no homunculus or "little man" sitting inside who can finally find meaning in the data. The meaning resides in the dynamic interplay of the mindlets as they vie for position to fit a pattern on the membrane.

Mindlets that cannot fit any patterns will eventually be purged. Their connections will be unreserved and turned over to other, more fertile mindlets. The machines that unreserve their connections are themselves a specialized kind of mindlet.

Every machine in the Nucleus is a mindlet, even though they differ widely in form and function.

Paying for creativity

Just as it does with matter and energy, nature keeps a strict account of the flow of information. Nothing comes for free. A digital computer cannot create information at all -- its information comes from the brain of the programmer. 

The pattern of tread in the tires of a car repeats itself endlessly as the car rolls along a sandy beach. The form itself, however, emerged in the minds of the engineers who designed the tire.

Digital computers can rapidly and endlessly repeat the decisions made previously by the programmers, but no new form can emerge because a digital computer cannot make a selection.

The Nucleus is a true information machine because it does make selections. Those selections come at a cost.

Think of a map that can shift scale at the touch of a lever. You can take a close-up view, showing fine details, or you can take a wide-angle view, giving you the big picture. The size of the screen displaying the map is always the same size, though.

It has a limit for how much pattern it can display at a time. Going in closer means you sacrifice the bigger picture, and drawing back means giving up details.

Each detail in the map has to be drawn in the computer's memory -- which takes time. That time limit is what will finally stop the screen from being able to move up or down the scale. It's interesting that the universe itself seems to have an unlimited scale.

You can look closer or further without ever running out of map. That's the ultimate distinction between any knowledge space and the universe itself.

A knowledge space -- presumably also the one inside a human's head -- eventually runs out of map.

The Nucleus uses Darwinian natural selection to close this gap. Although the amount of data that can be loaded into a computer will always be quite limited, the Nucleus is capable of arranging itself into a conceivably endless number of structures to accommodate that data.

The cost is paid for by the fact that the rets can hold such a vast number of connections. A digital computer that had to make calculations for each of the configurations in the nets would never be able to run fast enough to be of any use.

The mindlets would seem to move as slowly as the stars in the Milky Way. The digital computer would be able to make each selection precisely, not probabilistically.

On the other hand, mindlets are probabilistic, and never precise. You can never be sure that you have the best fit with the data being impressed on the membrane. It's only the fit that has worked best so far.

As a result, the City will be full of high speed digital machines that do precise jobs precisely, but are subject to the Babel Barrier.

Most of those machines will be able to move freely around the Utility network, something like the way a computer "virus" or "worm" roves through computer networks today. They will communicate with other software entities, trading resources and cooperating in order to do their jobs.

Some software machines will debug other programs, for example. They will also buy and sell from hardware entities, such as memory warehouses, databases, or number-crunching factories.

Some regions in the City will be like giant software junkyards full of pieces of computer code that can be purchased and screwed into larger programs.

Continuity and knowledge space

Language forces knowledge to conform to fixed categories of thought. How we choose to classify things depends much more on personal and cultural perspectives than what most people earlier believed.

Many philosophers, on both sides of the old empiricist vs. rationalist divide, have taught that there exist objective categories of knowledge.

As a result of work done by psychologists and linguists, it now seems more reasonable to think of categories as families of relationships made on an ad hoc, functional basis, as Wittgenstein taught.

Language, and writing in particular, tempt us into thinking of categories as existing objectively, as being "out there."

Actually, though, they exist only in our minds as an artifact or by-product of the way our minds process experience. Categories are the "filing system" of the mind, and don't really exist at all.

When people use language, they create prototypes with which other things in their experience are compared and contrasted.

A classic example, given by a linguist named Charles Fillmore, is the word "bachelor." The supposedly objective definition of this word is "unmarried man." However, some unmarried men are more "bachelorish" than others.

Most speakers of English probably wouldn't think of the Pope as being a bachelor, even though he really is one according to the strict definition of the word.

Cross-cultural studies of how people categorize things confirm this prototyping principle.

Relationships between objects in a knowledge space aren't discrete or exact. By varying the strength and density of the connections between objects of knowledge, the problem of rigid categorization would disappear.

In a knowledge space, distinctions are fuzzy, with a continuous spectrum of connections.

This isn't really the same as "fuzzy logic," incidently, which assigns probabilities as values instead of using Boolean values of true or false -- knowledge space probably isn't algebraic in any sense.

What would the mathematics of a knowledge space actually be? Just as there are infinitely many numbers between any two integers on the real number line, there are in principle an unlimited number of categories between any two experiences in a knowledge space.

A knowledge space conceivably allows ideas and experiences to be connected in an unlimited number of ways.

The mindlets in the nucleus of the City can store and handle knowledge in the form of a space.

Video as language

Although knowledge is stored nonlinguistically in a knowledge space, it has to get translated into some form that people can understand. For this, we'd need some kind of "language gate."

Because the connections possible in knowledge space are so vast, the number of "strings" of words that the gate would have to produce would also have to be immeasurably large.

In fact, the number of connections would always exceed the number of language strings producible by the gate. This is the Babel Barrier again.

One interesting effect of the City will be the evolution of video communications. Rather than create text to express the knowledge connections in space, a gate will create a flow of sounds and images.

The gate will show how something looks or sounds, instead of describing it in words. Using sophisticated moving charts, graphs, schematics, and so on, the gate can also illustrate abstract ideas. This might be something like the doodles or sketches a person draws on a napkin over lunch to express some new idea.

As people become more accustomed to multimedia presentations, certain shorthand iconic conventions will begin emerging. A particular image, or part of an image, might come to represent a certain idea.

This will eventually evolve into a rich multimedia language. Unlike spoken languages, though, words in a multimedia language will be "fuzzy," with a spectrum of subtly varying but closely related meanings. I like to call the multimedia language "Vid," short for "video language."

By using Vid, the language gate can communicate vast amounts of knowledge quickly to a learner. As the saying goes, "a picture is worth a thousand words."

The evolution of Vid will be like the early development of Chinese writing. However, Chinese writing is different because it lost its "iconicity" early on. In other words, people quickly forgot what the earliest pictographs represented, and instead began processing the characters as more or less arbitrary compositions of strokes.

The component of the character called the "radical" sometimes bears a connection to the character's meaning, but this is certainly not always the case. The Chinese characters lost their iconicity because of the sensual poverty imposed by the limitations of the medium.

Characters can't move or change color, for example. They often have to be written laboriously by hand. Esthetics also constrain the character's complexity. Electronic video languages won't suffer the same limitations, however.

In many ways, Vid already exists in a very primitive form. Many video conventions have become widely recognized. This has been a gradual, almost invisible evolution.

When looking at the earliest examples of the motion picture medium, this evolution becomes more evident. The first motion pictures were like stages-on-screen. The filmmakers planted the camera right in front of center stage.

When the pioneer filmmaker D.W. Griffiths first used the closeup shot, the people working with him were afraid that the audience would run out of the cinema screaming because they would believe the actors' legs had been cut off! Needless to say, the closeup shot has since then become a solid part of our video vocabulary.

Video literacy has been evolving faster since producers have started using more computer graphics and computer image processing. Network TV news contains many examples of new video "words," for example.

Most likely, a unique dialect of Vid will be specially created for each learner. Starting when the learner is just a child, the gate and the learner will together create a unique video language that reflects the learner's personal experience and style of thinking.

Each person's Vid will be as unique as a thumbprint. Twins are known to invent their own private "languages" in just this way.

Vid will be a language unlike any spoken one. There's no way to translate a spoken language into Vid, or to translate Vid into a spoken language. Spoken languages are discrete.

Vid isn't discrete because it flows out of the City. As in the thought experiment discussed earlier, the best you could do would be to make a verbal description of the Vid message in the same way as describing the experience of listening to a Chopin Nocturne to a deaf person.

You can't actually "speak Vid" because you'd have to draw all the moving color images by hand and produce all the sound effects yourself.

Since Vid is actually a collection of millions of personal dialects, you won't be able to read messages meant for someone else.

Vid won't work well for communication between people. We'll have to keep relying on spoken languages for that.

Personal knowledge robots

During the course of a lifetime, a learner will make a long trajectory through the City's knowledge space. The path will curve and twist according to the learner's interests and curiosity.

This path will form the learner's specific knowledge profile. A language gate will take the learner's knowledge profile into account when communicating new knowledge to him or her.

That way, it could make reference to what the learner probably already knows, in order to better "connect" the new knowledge. This will be like having your own personal tutor, who'd been with you since childhood, but who had read every book known to exist.

Alexander the Great had Aristotle. Why not the best for everyone?

I like to call these personal tutors "Facers" because they're personal interface devices. Facers will be like electronic -- or photonic -- servants, which represent you in the City.

Although Facers are mainly digital machines, residing in the Old City and the Utility, they can also communicate directly with the membrane. That enables them to use the Nucleus to distinguish and relate concepts, and to navigate the subtle passages of knowledge space.

Facers can communicate with other machines -- both hardware and software -- as easily as they can communicate with humans. Each Facer can have its own voice, life-like human face, and even its own flavor of personality.

Since the Facer resides in the City itself, and not in any particular machine, it is instantly available anywhere in the world -- wherever there's a TV-book or some other multimedia device that can plug into the Utility.

A crucial part of human language comprehension is the ability to understand metaphor. We constantly use metaphor to create new kinds of language.

Metaphor is the foundation for human language. Through metaphor you try to build a bridge between your experience and the experience of your listener.

Facers rely completely on the Nucleus for their ability to follow and create metaphor. Metaphor is also crucial to recognizing humor. It's possible that Facers will be able to laugh at jokes.

The City is teeming with all sorts of knowledge robots, many of which are hawking their services, and jostling to gain an edge in collecting, selecting, connecting or correcting some special kind of knowledge.

This information marketplace is as crowded and noisy as a Middle Eastern bazaar. Transactions take place much too quickly for a human being ever to keep up, so you have to send in your Facer to do your shopping for you.

Just like a farmer's draft horse in the days before tractors, your Facer will be your most prized possession. You'll depend on it for your livelihood.

It'll be like a teacher, secretary, attorney, accountant, medical advisor, librarian, purveyor of entertainment, private diplomat -- a whole society of intimate servants all in one ghostly machine that haunts the invisible City made of pure information.

Some people will build cleverer and more assertive Facers, and so they'll manage to acquire more useful knowledge.

They'll turn that knowledge to their advantage, and use their new wealth and influence to acquire even better Facers.

The City will open up vast new frontiers of opportunity for knowledge entrepreneurs, however, so maybe society won't stagnate with the same stale people always at the top.

Michael Webb, 1992

previous chapter ]
next chapter ]

home ]