The Fractal Actor on the Fractal Stage: Believability, Agency, Nonlinearity, and Beauty

The Fractal Actor on the Fractal Stage: Believability, Agency, Nonlinearity, and Beauty

by Brian Moore and Frederick Turner

1.  The Believable as Lifelike
Brenda Laurel’s fine book Computers as Theatre is a major step toward a historic synthesis between computer theory and the humanities.  She argues that the most natural and usable model for human work in the computer environment is theater; that we do not need to reinvent the wheel in developing a set of principles for how to make this cybernetic theater work, since there already exists a body of expertise on this subject which is over four thousand years old; that Aristotle, in his Poetics, has the best general summary of this expertise; that our work with a computer should have a satisfying plot, as a drama does, with a beginning, middle, and end; and that we, the users, should be the guest stars of the show.

Laurel’s book admirably traces how the linearity of classical Greek thought can be used as a map for the linearity of a good interactive experience with a computer.  To the extent that a drama can be what William Carlos Williams called a poem–”a machine made of words,” Laurel’s model is the right way to go.  But there is another aspect to poetry and drama–and also, we believe, to the computer/human composite experience.  A good play is not only rational, linear, and partly predictable in its construction, but also, mysteriously, irrational, nonlinear, and unpredictable.  It is precisely in the interplay between our reasonable expectations of what will happen next, and the surprises that actually occur (but which we realize with hindsight are perfectly consistent) that the pleasure–and the believability!–of plot and story really lies.  A good story is both unpredictable and retrodictable.  Only one known class of mathematical or physical objects meets that criterion–that is, nonlinear algorithmic or dynamical systems, with strange attractors that have a fractal form.  What we wish to do in this paper is to explore the nonlinear substance of believable plot, character, and mise-en-scene; and use that substance to flesh out Laurel’s useful framework.

Aristotle, like most of the classical Greeks, was both fascinated and repelled by surds–or irrational numbers, as we usually call them now.  Surds are usually produced by a mathematical algorithm that is nonlinear, such as the calculation of the Fibonnaci series for the golden section ratio, or the square root algorithm for Pi.  The classical Greeks equated rationality, which was for them among the highest of the values, with natural numbers, integer ratios, and linear processes; when they eventually developed algebra, it was in the form of diophantine equations, which have whole-number solutions, rather than infinitesimal calculus, which had to wait over fourteen hundred years after Diophantus, until the arrival of Newton, Leibniz, and Descartes.  Calculus itself, however, is still linear in its answers, though it involves the consideration, and cancelling-out, of infinite regresses.  It was not until this century, and the advent of really fast computers, that truly nonlinear objects and processes, such as the Mandelbrot Set and the Lorenz attractor, could be given systematic study.

But artists and poets have always, in a practical, empirical, and tacit way, known about nonlinearity and its fundamental role in giving inexhaustible life and substance to their creations.  Deconstructionist literary critics have recently stumbled on some of the superficial characteristics of the nonlinear process at work, and misinterpreted them as a general skepticism about the possibility of meaning.  Certainly, linear meaning is hard to fix to a world that is always interacting with itself; but there is a very substantial and vital nonlinear meaning that has always been evolving within nature, and of which we human beings are a peculiar outgrowth.  The experience of that meaning can be roughly compared to the delightful vertigo we feel looking into the infinite depths of a good Julia Set graphic–or the pleasure of looking at clouds, crystal growth, fern fronds, snowflakes, sunflower heads, the bodies of animals, and the turbulence of rapid waters, all of which work according to the same principles.

This quality–we have been using the terms “vital,” “meaning,” “substance,” “life,” “unpredictable-yet-retrodictable”–is, we believe, crucial to “believability.”  Another way to put this idea is to say that agency is a special property of living organisms, which actively alter their environment for reasons of their own; to say that an agent is believable is tantamount to saying that it is lifelike.  Finally, this quality, this combination of complexity, internal integrity, and rich relationship with the environment, has an ancient name: beauty.  For our agents to be believable they must be lifelike and beautiful.

2.  A Recipe for Lifelike Agency
How, then, might we provide the rich nonlinear texture that gives life to the agents we meet in the real world and in fiction?  This question has two aspects: one concerns the internal dynamics of the agent, its inner life and intelligence; and the other concerns its external dynamics, the social and physical world (or virtual world) which it inhabits.  There is an actor and a stage, a citizen and a city.  We will deal first with the internal dynamics of the agent.  What is the recipe for “lifelikeness”?  (If we get it right, we will also have something like a recipe for beauty as well.)

Knowledge.–First, you need a massive database.  In terms of quantity of information, this requirement is not much of a problem for contemporary computers, which can go out on the net and acquire as much information as they can store.  But knowledge is not the same thing as information.  Knowledge is not a passive thing, random access memory stored in the same form in which it was put in.  What we call knowledge is actually two things: skills, or habits of action; and memory, that is, a capacity to recreate or regenerate an earlier experience, using an awareness of the generative principles of what is remembered together with some significant fragment of the original whole that can serve as a seed and test of the result.  We humans seem to reconstitute our memories by a method not unlike the way in which an iterated computer program can generate a complex fractal shape out of a simple mathematical seed or algorithm.  But we also seem to need a background awareness of the whole world as the context and editor of the reconstructed memory; the remembered experience must fit into the right-shaped hole in everything else, and this fit acts as another check of its accuracy.  The way a memory fits in is essentially as part of a story, a program of contemplated or imagined action, which gives point and application to the knowledge.  Thus knowledge, either as a skill of action or as memory, is essentially active and creative.

Thought.–The second ingredient, after this massive database, is a massively complex information processing capacity, that is not clearly distinguished from the database itself: what we call the power of thought.  J. T. Fraser has calculated the number of possible brain states in a human brain–its repertoire–to be of the order of ten raised to the power of ten raised to the power of nine: that is, one followed by a billion zeroes.  To write this number down you would need a hundred books.  It is much larger than the number of particles in the universe.  The processing method itself is not of a single kind, but is a huge collection of different kinds of organization–logical, probabilistic, holographic, symbolic, metaphorical, metonymic, and so on.

Unified Self.–Third, this whole collection must be unified into a single self, which sits at the top of a “chunked” hierarchy of labelled subunits referencing, controlling, and sensitive to smaller and smaller moieties of the whole.  One of the most amazing characteristics of a human agent is its simplicity and unity, achieved only if all its functions can be concentrated and subsumed into a single focus of attention.  This focus derives from the overall motivation of the whole organism; here Brenda Laurel’s excellent analysis of dramatic character is key to the understanding of a lifelike agent.  We, and any believable agent, must have goals, or objectives as they are called in Stanislavsky’s theory of dramatic acting.

Objectives.–Objectives are the fourth ingredient in our recipe for lifelike agency.  Whence do objectives arise?  Essentially, out of a comparison between the experienced reality and a library of stored stories, one of which has already been selected because of its best fit to the actual situation.  We act according to a conscious or unconscious pattern of action, called a story; either by doing the next action the story prescribes, or, when reality does not turn out to fit the story, by changing reality to make it fit, switching stories, or seeking more information (including improved stories).  Action–agency–comes out of the gap between what is and what ought to be; the agent acts until the two correspond, that is, what ought to be is what is.  At present any simple computer program, say a spellchecker, will continue to work on a list of tasks until the list is exhausted–which is what ought to be–and it will then stop.  What will make such a program into a true agent is a function of the richness and nonlinearity of the stories it seeks to fulfil.  We must develop catalogs of stories for the various agents that we wish to bring to life: the secretary, the docent, the database manager, the investment counsellor, the travel agent, the game-antagonist, the consumer adviser, the social secretary, the maintenance man, the attorney, the medical adviser, the marketing consultant, the surveyor,  the translator, the diplomat, the campaign manager, and so on. Each should struggle with his or her own special compromises and plot-obstacles, and the result of this struggle may be truly insightful help.

Internal Feedback.–Fifth, the internal process of the agent must be enormously flexible.  Its input must be able to alter its programming (learning); the program must be able to alter the hardware (habit); different objectives should be able to compete and reach compromises (inner conflict, as in the previous paragraph); the program and the hardware must be able to be fed back into the system as input (introspection and psychological inquiry); and the higher-level functions must be subjected to periods of partial immersion in their own lower-level processing (dream).

External Feedback.–Sixth, the whole system must be fully integrated into a world suited to its capacities, by means of its distribution into a sensing, feeling, and active controllable body.  A large part of intelligence consists in the informative  interplay between an information processor and its world.  That world must itself be highly complex, partly intelligible, and composed of a mixture of predictable and unpredictable elements.  Without such a world the intelligent system could not learn and would be paralysed with boredom.  Thus the architecture of the stage or city within which an agent operates is crucial; it must be as rich and deep as the agent itself.

Emergent Properties.–These ingredients must, lastly, be cooked together in such a way as to generate emergent properties of a reflexive and self-organizing kind.  These include a circular feedback/feedforward loop (consciousness, self-awareness); self-inclusion paradoxes (mystical experience and the higher tenses of time); iterative chaotic non-linear relationships with the world (drama, personal history, humor).  The result should be capable of originality: consciousness is a reality generator.  And originality constitutes freedom, because in changing the world it changes the choices the world offers.  Consciousness is unpredictable (though after the fact its decisions can make perfect sense), autonomous (self-ordering), and creative.

3.  How Can We Get a Machine to Do All This?
How can we build creative imagination, unpredictability, infinite expandability–in other words, freedom–into a machine?  If the natural sciences stood now where they did up to a few decades ago, the project would be impossible by definition.  The universe was made up of deterministic processes interfering with each other in ways which could be calculated and predicted.  Perhaps there was an irreducible element of “noise”–of random vibration–in any system, but even this property was at last reduced into the statistical descriptions of quantum theory.  Chance and necessity–that was all there was.  Philosophers, artists, and believers in the human spirit were reduced to one of two desperate measures to assert the dignity and freedom of humanity.  One was to posit a supernatural world in which freedom could exist but which, unfortunately, required the abandonment of reason.  The other was to equate freedom with sheer chance, sheer randomness; and hence we got the existentialist’s “acte gratuite,” the gratuitous act of Camus’ and Sartre’s heroes.  Even now artists, in an attempt to escape what they imagine to be the deterministic and mechanistic tyranny of reality, resort to aleatory or chance-based processes in their creations.

Roger Penrose, in his book The Emperor’s New Mind, offers an argument intended to negate the possiblity of cooking up artificial intelligence; his argument makes an excellent point, which we will find very useful in cooking up believable agents.  His argument runs as follows.  A computer is a “Turing machine,” that is, a device which carries out the commands of a program or algorithm.  Its workings are in principle predictable, linear, and deterministic, and it cannot generate any kind of order which is not implicit in its input.  The only kind of truth it “knows” is the kind that it can prove.  Human beings, however, without resorting to any mystical assertions of supernatural powers, regularly deal with truths they cannot prove.  For instance, as the great mathematician Kurt Goedel showed, the statement “This statement is not provable” is true, and we can know that it is true; but it is not provable, for if so it would not be true.  All a Turing machine can do is prove things.  A computer, asked to solve a problem of this shape, would never be able to turn itself off–and this inability is the very definition of a Turing machine.  It cannot, so to speak, accept the difference between what is and what ought to be.  The reason why we humans can know that “this statement is not provable” is that we can step outside any system of proof, any group of axioms, and recognize global properties of the system or group as a whole.  In other words, our minds are infinitely expandable; they can generate new perspectives not predictable from the old.  But this process is routine for us; we use it any time we recognize an infinite regress and cut to the chase.  The artistic power, of creative imagination, is thus the first requisite for simple commonsense!

4.  The Nonlinear Universe and Nonlinear Machines
But the great revolution in scientific thinking through which we have just passed has dissolved many of the old distinctions.  In dealing with thought, memory, motivation, and all those other valuable human capacities we would like our agents to have, we are no longer confined to only two languages, that of the traditional humanities, which is too top-down, and that of physics, which is too bottom-up.  We now have the full and useful vocabulary of biology.  The revolution in the understanding of evolutionary biology and biochemistry, which showed how creative nature could be, retrospectively changed our view of chemistry and physics (we can see this transformation going on in the work of Ilya Prigogine).  Biological systems are not the only ones which involve feedback, non-linearity, and self-transforming flexibility.  There is now a whole new class of mathematics which describes and generates nonlinear, nondeterministic, discontinuous processes.  Whole new branches of physics and chemistry  have arisen that deal with open systems, self-organizing processes, infinitely self-transforming feedback loops, ordered entities whose initial conditions can no longer be extracted even in theory, and whose future behavior, without being in any sense random, is beyond the capacity of any calculator, even one made of the whole universe, to predict.  The universe is itself free, autonomous, creative, self-transcending, self-ordering.  Our conscious awareness and will are not an anomaly in the world but simply the most intense form of nature’s own self-reflective and self-generating process; and the evolution that brought us about is only a slower version of the same creative investigation that revealed the nature of evolution itself.

For living organisms to evolve, three basic factors must be present: some mechanism of variation, that ensures that each new generation will display significant departures in form and function from its ancestors; a process of selection whereby an environment (which may include the presence and interactions of the species itself) can eliminate from the breeding population those organisms not fitted to survive in it; and a conserving mechanism of heredity, to fix in a genetic archive what has been learned in the process itself.  This archive must in turn be subject to variation in the next generation, providing a field within which selection can go to work, the results recorded by the differential preservation of some genes over others, and so on.

The iterative, feedback nature of this process is crucial.  One cycle of the process is not enough.  But the sequence of many cycles generates a self-organizing flow which tends toward what chaos theoreticians call a “strange attractor,” a complex but ordered shape of remarkable integrity, durability, and beauty.  Perhaps our pan-human sense of beauty is a natural recognition, and capacity to create, free self-organizing reflexive feedback systems.  Like the outcome of a good story, an attractor is not predictable from the initial conditions or from the algorithm (for instance, the variation-selection-heredity cycle), but once it has begun to appear out of the apparently random results of the process, it makes perfect sense.  We can tell whether a process is genuinely alive and organic, by the suspense we feel as it works itself out; a suspense that comes from knowing that whatever shape appears cannot be predicted but will be ingeniously adapted to its conditions.  No suspense, no life.

It might well be said that each living species is itself a kind of strange attractor, the natural but unpredictable goal of a given set of genetic inheritances combined with a given ecological niche.  Hence, for instance, anteaters of similar shape and capabilities have evolved independently on three continents, and hence also the remarkable convergence in form and function of the marine dinosaurs, the whales and dolphins, penguins, seals, walruses, and sea-otters.  In a broader sense the four-limbed, backboned, five-digited conformation of the class of vertebrates may be a large general attractor, systematically distorted by the more local pull of species attractors.

The physical world, then, already provides many processes that in principle give us the ingredients of a creative, free, and expandable agent.  A computer is a part of the physical world, and need not be excluded from the world’s natural creativity.  There are now software programs that, within certain narrow limits, can mimic the unpredictable genesis of order that we find in living systems.  Beginning with Conway’s game of Life and the fractal sets of Benoit Mandelbrot and his followers, a whole new class of iterative computer programs has been devised, which reproduce the chaotic self-organization of nonlinear physical systems and the fantastic creativity of biological evolution.  New cybernetic entities called biomorphs are, even now, happily reproducing themselves in competition with their fellow-survivors, within the memory-spaces of private, business and educational computers.  They “live” within Turing machines, but are not Turing machines themselves.

There is absolutely no reason to suppose that iterative evolutionary attractor-seeking processes might not be created within a cybernetic environment, processes which would, moreover, operate at enormous speed, many orders of magnitude greater than that of ordinary biological evolution.  A kind of pseudo-life would appear, perhaps based on a close analogy to the DNA molecule as its genetic archive, perhaps not.

Contemporary theoreticians of brain function–how brain generates and responds to mind–are now using a model of thought and memory which closely resembles the Darwinian process of natural selection, though like cybernetic processing it is blindingly faster.  Many more new brain pathways are laid down, by neural growth and synaptic connections, than we could possibly use.  Some patterns of connection are selected over others by their repeated correspondence with the sensory experience and pleasure-rewarded purposes of the individual, so that mental habits, skills, and memories are formed.  It is known that the pleasure-chemicals of the brain–including those associated with the experience of beauty– are involved in laying down and reinforcing new synaptic pathways.  Other possible patterns are pruned out–selected against.  New experiences, and such continuous spontaneous reshuffling of connectivities as we experience in dreams, would produce a new range of variations upon which the selective process could work and whose survivors would be preserved in improved synaptic connections (dreams, then, playing for mental birth the role of sex in biological birth).  Our enduring, and often pan-human symbols, ideas, myths and images might thus correspond to the strange attractors we have found elsewhere in nature.  Recent experimental evidence seems to confirm this hypothesis–our memory of smells, for instance, is constituted by the attractor of a nonlinear, highly complex, feedback loop among those calibratable neurons known as Hebb cells.

If biological evolution can be copied in a cybernetic space, why not neural evolution?  Perhaps artificial intelligence will come into existence not as a program we design, a set of instructions to a Turing Machine, but as the evolutionary result of a process we initiate; not as a calculation but as a reflexive entity carried and mediated by calculations.

Maybe we can, then, cook up agents with the right properties.  But could they exist in isolation?  How indeed could all the power of the human mind be packed into a one-quart skull?  How could this mess of chemical and electrical connections produce creative imagination, a self, a soul?  And how could the extremely variable, inexact, and almost haphazard process of cellular development produce a person–a fortiori, a believable agent–almost every time a human zygote is allowed to mature?  The answer is that neither humans nor computers exist in a vacuum; their internal complexity is only half the story, and they require a rich web of interactions with an outside world, upon which their intelligence is parasitic.

5.  A Believable Agent Needs a World
Natural agents–persons–do not come full-blown into the world.  They rely on an animal body and sensorium already provided by billions of years of evolution and heredity; they require a world to develop in, a family, a culture, an education.  Indeed, if the latter are present, a specifically human body and nervous system are not absolutely necessary; Koko the gorilla knows some human language, which we have established as the benchmark for personhood.  And if a world, family, culture and education are absent, as they are for human babies that are fed and housed but otherwise neglected, a human being, suffering from what is sometimes called kwashiorkor or marasmus, may never learn as much language, or be as human, as Koko.  The machine of the living brain is not sufficient to make an agent; why should an electronic machine be sufficient either?  If a “natural” agent is permitted a highly complex machine to run on–that is, a sensorium, a body, a family, a culture, a world–then why should we not grant the same conditions to an “artificial” agent?

Perhaps the distinction between natural and artificial agency is itself artificial, and may in many senses be plain wrong.  All living organisms, including even viruses (which exploit the biochemical machinery of other lifeforms) use tools to survive, and could not make it using only their own bodies.  The water in which a bacterium swims is a kind of tool.  But even bodies are a kind of tool.  As far as the genes are concerned, the whole complicated structure of protein chemicals that constitutes the body of an animal or plant is a machine designed to ensure the survival of the genes.  The nervous system is a piece of complex wiring created over eons of evolutionary time to protect and serve the body as a whole; and its emergent properties of consciousness, individuality, and so on are a kind of technology.  Natural agency is  artificial, in this sense; the problem of artificial agency has already been solved, and all we need to do is understand and duplicate the solution.

But this conclusion might suggest that the cybernetic search for believable agency is a waste of time, a reinvention of the wheel.  After all, a male and female human being can, in the space of a few delightful minutes, create together an electrochemical organism which, with proper care and programming, will perform all those remarkable operations of reflection, imagination, and laughter to which we have referred.  Why make HAL when you can have a real human baby?  Or, to put it the other way round, will HAL be any use to us as a machine if he has all the moods, rights, failings, quirks, and political orneriness of a human being?  This is a serious challenge, but one which we believe can be met.  The real change, and the exciting prospect, arises out of the redefinition of the problem of believable agency, and the reconception of the arts, which is coming into being.  And this redefinition will make the wheel that we reinvent a very different entity from the one we invented the first time around.

An analogy with biological science may be helpful.  When the DNA code was discovered, and then recombinant DNA techniques opened the way to direct manipulation of the genes, it looked as if we were stepping into utterly new territory.  Some of us were exhilarated, many were terrified.  What monsters might we produce?  But then we gradually realized that human beings have been doing slow DNA manipulation by selective breeding and hybridization for thousands of years already, and that wheat, dogs, sheep, apples, and cattle were all artificial species.  Further, the research of people like Lynn Margulis showed that all bacteria were already quite happily, and in the most polymorphously perverse way, exchanging genetic material in a continuous incestuous group grope; and that through viral insertions into our own genes we too were part of the same millenial orgy.  We  were the chimeras, the Frankenstein’s monsters, the Andromeda Strain.  And so were all our brothers and sisters in the animal and vegetable kingdoms.

But this did not imply that, because nature was doing it already, we should give up recombinant DNA research.  All it meant was that our research was an a natural continuation of nature’s own research and experimentation, and further, that we now had the most expert and ancient guides in our work.  Laboratory DNA splicing is only a faster and more controllable version of the gene shuffling which happens outside the lab, but there are advantages to speed and control.  A new feedback process–that is, conscious knowledge and decision–will have been added to the already rich brew of evolutionary self-organization.

The same applies to believable agency research.  The universe itself is a believable agency research project, and this planet is one of its laboratories.  The animals were the first successes in the project, and they became the laboratory subjects for the next phase.  When some of those laboratory animals discovered Art they discovered how to induce artificial agency in themselves.  Now they are trying to induce it in other forms of matter than the carbon-hydrogen-nitrogen-oxygen machines in which it first arose; perhaps it can be done in silicon doped with metallic impurities.  Perhaps not; but the process is entirely in the spirit of nature and a continuation of it.   And in the process of discovering if it can be done, we may also discover that there may be other forms of agency that inhabit our complex technological, biological, and institutional machinery than the human persons who are christened, vote, get married, and play the central role in funerals.

6.  Alien Agents Among Us
Even within a single human body there can often be several centers of conscious awareness.  Consider the various consciousnesses of a split personality patient.  They are clearly distinct intelligences, and often have different tested I.Q..  (Should each one have the right to vote? If not, do we deny the franchise to the Republican or to the Democratic personality?)  Or consider the introjected and internalized personality of a parent or other family member, or one of the beloved dead.  At their best such persons are for many of us a dear presence with whom we can talk.  At their worst they can be like the elder Mrs. Bates in Psycho.  In any case, they have a great deal of spontaneous and individual intelligence, quite distinct from the mind of the person that harbors them.  Or what about the self that is criticised when we criticise ourselves?  It is not the self that criticises.  You fool, we say to ourselves.  You fool, for calling yourself a fool.

When one converses with another person one must construct an internal model of her consciousness, with its own independent motivations, in order to be able to converse with her at all.  That internal model must have its own internal model of me, with its own internal model of her, and so on.  And if one is a writer of fiction one will often find, if one is any good, that the characters one has created have, as we say, taken on a life of their own, and can engage in a battle of wills with their creator.  A good actor can become consumed by the role he plays.  The role, the fictional character, can surprise its author as thoroughly as any person of flesh and blood.

And here we enter the world of all those agencies that are carried and mediated collectively and institutionally, whether they were once living persons, or were created by artists, shamans, politicians, cartoonists, or advertisers: Uncle Sam, Falstaff, Florence Nightingale, Wild Bill Hickok, Liberty, Santa Claus, Jesus, Athena, Buddha, the Jolly Green Giant, the battleship Big Mo; the Gods, Goddesses and Heroes; the patron saints; the presences that haunt rivers and groves, gallows and childbirths; the spirits of nations; and the commercial corporate personalities to which Locke and Hobbes granted semi-legal reality, and which we tax, criticise, sue, and reward.

These various entities form a continuous series, in which it is rather hard to draw a strict line of demarcation separating genuine independent agents from “mere” fictions.  If they are intelligent, they are not the same intelligences as the human beings that support them.  We would maintain that they are perfectly adequate examples of existing artificial intelligence. They live, like parasites or symbiotes or viruses, upon the host circuitry of individual people or groups of people, using much of their working hardware and even input, but transforming it in ways unavailable to the unaided host.  But in this they do not differ from the human beings themselves, which live as symbiotes upon their own bodies, their culture, their language, their world.  Are these artificial persons self-aware?  Do they possess subjectivity?  Some of them clearly do, like the split personalities of the mental patient; and many actresses will tell you that they have experienced, say, Cleopatra’s own motivations and selfhood from the inside.  With the Easter Bunny, Britannia, and the Pillsbury Doughboy the question is more problematic but not utterly ridiculous.

These secondary intelligences even have a dormant existence in books, musical scores, paintings stored in attics, and so on.  They are activated when they are played on a human brain, using the brain tissue as their hardware.

In the last few years a number of computer researchers involved in artificial intelligence and information architecture have come to study the arts and humanities with one of us (Turner).  These researchers felt that the cybernetic disciplines themselves did not have a rich enough body of theory and sufficient practical experience with the nature of Mind to be able to solve the new and interesting problems that lay beyond expert systems, symbolic logic, and holographic image-recognition.  Artificial neural networks, the new hardware approach to artificial intelligence in which electronic systems were designed to mimic the human brain, had bogged down in an atmosphere of caution and pessimism, and a disinclination to set the sights higher than the solution of immediate practical problems.  These students came to Turner looking for a wider conceptual structure.  In the process they became his teachers.

He sent one of them to look at the impressionists in the Kimbell Museum of Art, and the student came back with a visionary gleam in his eye.  “It’s AI!” he said.  He had seen a Cezanne and found himself hooked into another worldview, another way of thinking, a new subjectivity.  It was not just AI but telepathy.  The light on that jumble of rooves across the Seine, the strange lovely drear feeling of that afternoon, and then the oddly pastoral thoughts, the sense of the city as a sort of countryside, transcended all those years, the language barrier, the valley of the shadow of death.  Mike had become Cezanne.  Turner has felt that strangeness himself, translating the poetry of the great Hungarian poet Miklos Radnoti;  he did not know sometimes whether he was Fred or Miklos, he genuinely felt Radnoti’s subjectivity, the spontaneity of his wit and feeling.  And is not a violin soloist transformed into the Mozart she plays, perhaps into something that is more Mozart than Mozart was himself most of the time?

We already have artificial intelligence. The arts are traditional forms of artificial intelligence going back some tens of thousands of years.  This is not a metaphor, and it arguably constitutes the most accurate definition of art itself.  Milton said that a book is the precious lifeblood of a master spirit.    The quality of “lifelikeness” we are seeking in our cybernetic agents is the same thing as the classical quality of Beauty.  Poems, paintings, novels, sculpture, music are artificial intelligence programs to be run on meat computers.  Until now those intelligences relied on human hosts to do the work of bringing them into existence, maintaining them, and serving them by action and perception.  They could only be stored in books, paintings, scripts and scores.  Now we may be on the verge of giving them bodies of their own, even if they are as yet crude pieces of electronics.  But we must also give them a world to live in, a stage on which to act, a city in which to play and do business.