Prologue

Machinery of the Mind

by George Johnson

Prologue: Breaking a Thought Into Pieces

The alarm clock rings, you rise automatically and click on the coffee maker as you step into the shower and start thinking about another day. You might as well be on automatic pilot. But the morning fog clears from your brain and, bit by bit, you become human, with decisions, feelings, inspirations both silly and profound - the kind of elusive phenomena that surely can't be mechanized. Or are you deluding yourself?

Suppose that you awake this morning to find, like Franz Kafka's fictional character Gregor Samsa, that you have been transformed into another creature - not, in this case, a giant insect but a very sophisticated digital computer. Or, to make the metamorphosis a shade less absurd, imagine that you have been granted, overnight, unusual powers of introspection, the ability to gaze into your mind and examine the processes inside. Armed with this newfound skill, you perform a thought experiment - A Day in the Life - in which you zoom in close with your epistemological microscope and examine the gears in the machinery of the mind.

You're halfway down the stairs of your apartment building when a thought occurs: you've forgotten your umbrella. Do you climb back up two flights and risk missing the subway or do you chance getting wet? After a split second of what seems like suspended animation, a decision emerges. Acting on a hunch (whatever that is), you dart upstairs, unaware that you've acted out of anything more analyzable than a gut feeling.

But pretend that you work like a computer and your hunch is a costbenefit analysis calculated so rapidly that you aren't aware of the myriad steps in the computation. During that second in which you don't even know you are thinking, your decision-making apparatus performs a memory-bank sweep, retrieving meteorological data gleaned from last night's television news or a glance at the morning paper's weather box, or out the window where the quality of light tells you there might be clouds. You obtain more data from your wired-in calendar clock, which knows that it's a late-September morning and what that might mean. Using the information you've gathered from being alive during a thousand September mornings, you make an educated guess, a hunch-an informal statistical prediction of how, based on your experience, the day is likely to be.

While these machinations are occurring, another process is taking place. You glance at your watch and recall the subway schedule. The neural circuits that are dealing with weather award 9 points to the possibility of rain. The likelihood of missing the subway is given an 8. Up the stairs or out the door? The answer is (subliminally) obvious.

But can it really be so simple? We are each individuals, after all. So introduce more factors. Perhaps there is a dial that controls fear of being late; turn it from to 7. Perhaps this new signal would swing the decision the other way, in favor of forgetting about the umbrella. Or lower the level on another dial that controls one's aversion to rain.

The point is that you can make your imaginary computer as complex as you like, adding dials to control any number of characteristics, making the gradations on the dials as fine as you please, programming the entire system to base its decisions on many times more data. Maybe that's not how the mind works. But it begins to look as though it could be modeled that way, if the simulation were fine enough.

With umbrella in hand, you rush out the door, nodding perfunctorily at a neighbor, whose name you still don't remember. You stop, as usual, at the newsstand to pick up the Times, but today, on an impulse, you buy the Post instead. There are two possible explanations for this seemingly irrational behavior. First, there was that party you attended last week, where someone you were trying to impress accused you of being an elitist when you made a snide remark about the public's fascination with trashy tabloid journalism. Who were you to say a periodical is bad because it has lots of pictures, big headlines, and is aimed at the common man? Actually, you were referring to those supermarket weeklies with stories headlined JKF ALIVE IN SECRET CAVERN, or WOMAN MADE PREGNANT BY UFO, but still the accusation stung. You, an elitist? The word lodged in your brain like a piece of glass. For the next few days, rubbed by other memories and experiences, it seemed to take on a charge, a mounting aura of static electricity that laymen call guilt or shame. Have you ever actually read the Post? Today, just as you habitually reached for the Times, the feeling exceeded a threshold, sparking across the gap of some neurological electrode and jerking your arm, froglike, one foot to the right, where the stack of Posts lay waiting. You may forgive yourself the crude, almost prehistoric, analogy. The spark gap and the accumulating voltage could be simulated by the counter on a digital computer, movingclick, click, click-incrementally higher.

The other explanation for your surprising choice of newspapers is simpler and more direct, though no less mechanical: the Post bears a headline about a disaster in Mexico City, a place you have been vaguely thinking about since yesterday when a song on the radio reminded you of a week spent there with an old friend. Whatever the reason for your purchase, the story about Mexico is what you read as you sit on the subway train and begin moving toward the office.

Somewhere, about the third or fourth paragraph, your attention begins to lag. You pause, fold the paper in half, and lean back, reflecting. A section in your memory called Mexico has been ignited. Like the madeleine in Proust's Remembrance of Things Past, a garish newspaper article and an old Top Forty radio tune have evoked a chain of memories, one linked to another in a procession that recedes into the vistas of your mind. As the train leaves the tunnel momentarily and crosses the river, you gaze listlessly out the window. But in the world behind your eyes you wander the streets of another city, the network of thoughts and images stored somehow in your brain and indexed under Mexico: the hotel with the elevators like brass cages, the crowds of grotesque faces in the Diego Rivera mural in the Palace of Fine Arts. And there is a scene on Paseo de la Reforma, the treelined boulevard that angles from Chapultepec Park toward the Alameda: white-and-gold monuments rising from the center divider, vendors selling lottery tickets, side streets choked with quadruple-parked cars.

And then you remember the traffic circles. What is the word, in Spanish, for traffic circle? Now your internal wanderings become more deliberate as you try to retrieve this elusive piece of information. You can almost feel the word crystallizing into consciousness, almost sense its shape and sound. It has the feel of a woman's name, or, no, a flower. You can barely see its profile-the skyline of letters. It seems there's a t sticking out somewhere, a sharpness rising above smoother, rounder-sounding letters. But when you try to focus closer the word disappears, like those barely existing hazes of stellar light that can be seen only from the corner of an eye. You try to coax the word back, when suddenly, for no good reason, the image of a freeway looms into view. Not in Mexico, oddly, but in New Mexico. It's the interstate highway that runs along the base of the Sangre de Cristo Mountains, from Santa Fe toward Denver. There is an exit that leads to the Pecos wilderness and that night long ago camping in a canyon with fireflies everywhere.

A roadway in your mind has led, unexpectedly, from the streets of Mexico City to another cluster of memories. A detour, it seems, but then you remember. The town at the exit is Glorieta, New Mexico, and the word, in Spanish, for traffic circle is glorieta, which sounds like Gloria, a woman's name. It is flowerlike and round (a traffic circle is round) - and flower in Spanish is flora, glora, glorieta. The word was stored in many different places and crossreferenced according to its shape, its sound, the places where it has been experienced, as well as by what it means.

Well, no one said that your data base had to be simple and clear-cut. It's more a tangle than an array. The mural by Diego Rivera has tentacles extending to the memory of another of his works, on a ceiling in the Museum of Science and Industry, which is filed in a section where you keep vague, fading remembrances of a childhood trip to Chicago.

You could wander all day, back to Mexico or to that canyon in the wilderness, but reality intrudes. You emerge from the network of tunnels (the ones in your mind as well as those the train has been traveling through) and find yourself outside the building where you work. You go inside, ready to spend the day utilizing a more orderly part of your mind. Most of us would agree that there are scores of jobs of such mind-numbing repetition that they easily could be mechanized. But to keep the experiment interesting, let's pretend that you're a doctor. Patient A is complaining of stomach troublea burning pain, strange rumblings in the night, food that seems to sit in the gut as indigestible as concrete. So are we talking ulcer, maybe cancer, or just a nervous stomach? How about cirrhosis of the liver? All have overlap
ping symptoms: nausea, abdominal cramps, a feeling of fullness. To find your way through the web of possibilities, you must gather more information, ask questions whose answers will lead to other questions, and more direct inquiries such as X rays and blood tests.

Does the patient drink, smoke, lead a stressful life, eat a lot of Szechuan food? These activities all have been linked to stomach irritation, which might lead to chronic gastritis and perhaps, in extreme cases, ulceration. Is the patient's skin jaundiced? That is a symptom of liver trouble. Is the spleen enlarged? That can mean cirrhosis, but it can also mean leukemia, Hodgkin's disease, malaria, or mononucleosis, which all have distinguishing and overlapping symptoms of their own.

You ask the patient to be more specific. Are the pains in the upper abdomen? Do they happen between meals and go away after eating? Is there a sour taste in the mouth that abates after taking antacids? Those might be ulcer signs. If the pain is lower and to the right it might be appendicitis, which is often accompanied by a high white-blood-cell count-and that, of course, can mean an infection in any part of the body. Weight loss can accompany cirrhosis or cancer.

In even the most routine case, there is an overwhelming amount of information to consider. Men might be more likely to get ulcers than women. Certain personality types are ulcer-prone. And once a diagnosis is made, there is a spectrum of possible treatments. In the case of an ulcer, they range from medication to surgery.

An experienced doctor doesn't consider every single factor, running through a vast mental checklist. Over the years certain symptoms recur so frequently that they are compiled into patterns which, when observed, can elicit a diagnosis almost immediately. They feel like hunches, but so did the decision to go back for the umbrella. After all, medicine is supposed to be logical, at least largely so. The most complex medical decision might be broken into thousands of microdecisions, which, if properly strung together by a computer, would produce the same diagnosis a human doctor would. Each decision need not be black or white. To allow for uncertainty, probability ratings could be included, translated as numbers between 1 and 10, or 1 and 100, depending on how much subtlety is required.

And what about the instincts that seem to contribute to the diagnostic process-the feeling you got that you were dealing with a nervous person, someone who internalized emotional troubles, converting them through the alchemy of neurosis into excess stomach acid? Couldn't that hunch also have resulted from the accumulation of clues: the way the patient fidgeted, asked too many questions, recounted every imaginable symptom in excruciating detail? Without consciously thinking, you weave the clues into a pattern, then compare it against the model of what you know from experience to be the classic ulcer personality.

And yet you wonder. Can it all be so explicable? Is there no such thing
as a creative diagnostician, one who startles colleagues with insights? Well, unless you believe in magic, don't those insights have to be based on information and experience? Isn't the brilliant internist likely to be one who has read more, retained more-whose brain is, perhaps, wired better so that it can recall data more rapidly, weaving it into more possible structures to consider, one against the other, so that when the most logical possibility emerges it seems like a bolt out of the blue?

And so the story goes. When the day is done, you walk outside and feel a raindrop. Unfortunately you've left your umbrella at the office and you don't feel like going back. You get wet, catch a cold, and stay home the next day. Nobody said this was a perfect computer.

Or did the possible pleasures of spending a day sick but not debilitated contribute to your forgetting the umbrella? This is starting to sound like a parody of Freud. All this hyperanalytic nonsense is depressing. You pick up a murder mystery you've been wanting to finish and go back to bed. But the thought experiment isn't over. For now you have enacted your bookreading routine.

Granted, the rules for parsing sentences and correlating words with definitions might be mechanized. But can the experience of reading a novel also be analyzed this way? Just what might be happening on the deeper levels of your mind as your eyes sense words moving by? A subject, a verb, and a predicate combine to form an image in the head, constructed according to the definitions of the words and your experience of what they mean. Using adjectives and adverbs, the writer fine-tunes the images, and arranges them one after another so a story unfolds. Characters are developed by piling image upon image, making complex clusters that interact with other clusters. To build narrative tension, key bits of information are withheld, then supplied in a rush at the climax, filling in holes in the data structure that has been induced in the reader's mind.

The writer's work, like the physician's, can be viewed mechanistically. One takes the X number of basic plots and themes that critics tell us exist and composes variations. And what about style? Within this reductionist paradigm, style can be described through textual analysis, statistically. Writer Z's books are unique because an average of 12 percent of the words are adjectives, 3 percent are adverbs, 21 percent are chosen from a dictionary with a difficulty level of 7. Sentence rhythms can be quantified. Perhaps such seemingly abstruse talents as the use of connotation and metaphor can be explained mechanistically.

"Okay, enough!" you might well interject. It's clear that you can take mental activities and describe them as an interaction of many smaller steps. And if, in this game, we are allowed to break a thought into as many pieces as we'd like, and break those pieces into pieces, and posit an overwhelmingly complex system that coordinates them all - then, yes, you can say that intelligence might be described mechanistically. But what does that mean?
It means, if you accept this argument, that you believe in the possibility of artificial intelligence. That, given good tools and enough ingenuity, we might cut through the veils that obscure our thoughts. Then we can see them not as mysterious, ghostlike essences but as complex structures made from the interaction of millions of pieces, theoretically duplicable on a machine.