Norman lifts the Styrofoam out of the box and points to several smaller chips pinned next to the microprocessor. “Your basic microcomputer consists of a microprocessor and a couple of extra memory chips. These two black ones are Harris 256-byte CMOS RAMs, which you can think of as the scratchpads of the system. Because they’re RAMs, the computer can doodle in them while trying to solve equations, and then wipe them clean before tackling the next problem. These chips are new, low-powered, and very hard to find. Apparently not that many people care about saving a watt or two. CMOS, or complementary metal oxide semiconductor, is the name of a logic family that uses metal oxides as insulators and conductors. When you deposit these oxides on different layers of silicon, they form transistors. CMOS is only one of the several logic families. The most common is TTL, or transistor-transistor logic, while Hewlett-Packard uses SOS, which stands for silicon on sapphire.”
Directing my attention to another chip pinned alongside the microprocessor, Norman fingers a purple rectangle that looks like a miniature California van with a sunroof on top. “This is the ROM. It stores the computer’s program in long-term, or read-only, memory. This is a super deluxe model, a Texas Instruments 2532 EPROM with four thousand memory locations that can be re-burned as many times as you want. It took three chips in the old generation of computers to do what this one does.” He explains that the sunroof is actually a quartz window and that by shining ultraviolet light through it one can erase the electrical charges comprising the 1’s and 0’s of the computer’s memory. The chip can then be reprogrammed by establishing a new set of charges.
In the hierarchy of computer memories, the simplest is fixed permanently in read-only memory (ROM). Up one level of computer intelligence is programmable read-only memory (PROM). But the most versatile of computer memories—which mimics what Freud described as the mystic writing pad of the human mind—is erasable programmable read-only memory (EPROM). This can be altered and augmented throughout the life of the chip. “Hackers joke among themselves about a fourth kind of memory known as a write-only memory, or WOM,” says Norman. “You can put information into it, but you can’t get it back out again.”
On looking through the window of the EPROM, I see a gray mass of silicon sitting on a gold platter of conductive lines. “The lines along the edge of the chip decode and control its logic. The homogeneous gray expanse in the middle holds the four thousand memory locations. If you move your head back and forth over the window, you’ll see rainbows. These come from extremely small etchings on the silicon that make up the locations themselves. Unlike the EPROM, if you opened a window into a microprocessor chip, its internal architecture would look far less uniform. You’d recognize discrete areas of silicon for performing different tasks. The complexity of what it does makes the structure of a microprocessor much more complicated than that of a memory chip.
“In fact,” says Norman, “you could build a computer out of a microprocessor and no external RAMs or ROMs whatsoever. Its capabilities would be limited, because the memory chips store not only the information you’re manipulating but also the instructions for operating the computer. But the microprocessor itself holds a half dozen memory locations, and this is sufficient to run the simplest of all computer programs, which says, ‘Jump back to the instruction that says jump back to the instruction.’ The microprocessor sits there jumping back to the same instruction over and over again in a little loop. As worthless as it sounds, we use this program quite a bit when we want the microprocessor to idle between clicks.”
After showing me the EPROM, Norman places a piece of tape over its sunroof. “You have to keep the window covered because the program can be erased by sunlight. Ultraviolet rays break down the electrical contacts in the chip and reorient all the memory locations in the same direction. The program, which is stored in thousands of little charges built up between materials, simply leaks away. In that case we could always reprogram the chip, although that’s sometimes trickier than it sounds. We’ve built a special circuit onto the KIM computer that addresses every memory location in the EPROM, pulses it electrically, and gives it whatever charge we want. But you have to be very careful not to apply the wrong voltage, or the entire chip goes up in smoke. We have a graveyard full of thirty-dollar EPROMs that have gotten fried.”
Pinned into the Styrofoam next to the microprocessor, the two RAMs, and the EPROM is the black cartridge of a Synertek PIA, whose acronymous name stands for peripheral interface adapter. This is the fifth and final chip intended for the bottom half of the computer sandwich. “Your basic computer is made out of a microprocessor and memory chips,” says Norman. “But if you want the computer to communicate with the outside world, you need some kind of interface.”
As Norman explains its function, I get the idea that the PIA operates like the New York Port Authority of the computer. It provides the network and vehicles for moving bits by the million from chip to chip and farther out to the peripheral components. The PIA shuttles bits around the City of Computation on what are called buses. The buses come and go from data ports on the side of the PIA, and they travel through a grid of printed circuits or wires that stretch all the way from the central processing unit to the periphery of the computer. Riding on the buses are signals sent from the outside world into the CPU and back out again. On one side of the PIA are the blinking lights, buzzing solenoids, radio frequencies, key strokes, and phosphorescent words by means of which humans communicate with computers. On the other side is nothing but the silent shuttle of electrons pulsed through silicon gates at a million or more cycles per second.
Norman points into the Kodak box at a cluster of brightly colored wires blossoming from the data ports on the PIA. “Two kinds of buses arrive here,” he says. “There are data buses and address buses. A bit decides to get on one bus or another according to the following schedule. When the microprocessor finishes calculating a roulette prediction, it needs to communicate the news to someone, namely the bettor. It does this by putting sixteen bits, or two bytes, out on the address bus and another byte on the data bus. The bytes on the address bus arrive first and alert a particular pin on the PIA to expect a message. Then the data byte arrives to give it the message. This in turn tells the PIA to send a signal out from one of the data ports, and this final signal buzzes a relay, transmits a radio wave, or does any number of other useful things.
“The data bus is bidirectional, and the procedure works the same way for information coming into the computer. Let’s say the microprocessor has to retrieve something from its memory. The memory, as you recall, holds four thousand bytes on a single chip. The microprocessor wants to grab one of these bytes; so it sends out an address on the address bus, which tells the memory chip which byte it wants. After the memory chip has decoded the address, it sends back the byte requested on the data bus.”
Turning from the microprocessor, memory chips, and PIA, Norman lifts the second piece of Styrofoam out of the Kodak box. Pinned to it are the components destined for the top half of the computer sandwich. If the bottom is devoted primarily to memory, the top of the sandwich will specialize in logic. Five black rectangles, each roughly as long as a fingernail, are attached to the Styrofoam by their golden legs. “These are the logic chips,” Norman announces. “They tell the microprocessor when to turn itself on and off. That happens to be a nontrivial task. Like everything else in the computer, the basic unit of information for the logic chip is a bit. A bit can be carried by one wire, which has either a low level, corresponding to a 0, or a high level, corresponding to a 1. Logical functions in a computer are performed by thousands of these wires built into circuits, and the circuits themselves are organized into various kinds of logical building blocks.”
The “wires” and “circuits” in a logic chip are actually nothing more than microscopic locations—known as logic gates—etched by the thousands into the crystalline structure of a silicon chip. These miniature switching circuits have two states, open or closed, and exactly how these two
possibilities allow binary numbers to solve logical problems is crucial to understanding how computers “think.” The simple steps performed at the level of the logic gates provide the foundation for a computer’s more advanced forms of cognition.
Computers begin their thinking process by shifting relationships between numbers into those between statements. The logic gates accomplish this task by translating combinations of 1’s and 0’s into “true” and “false.” The mathematical logic of digital circuits—the rules by which this translation takes place—was developed by the Englishman George Boole, a nineteenth-century contemporary of Charles Babbage. These rules, known as Boolean algebra, allow for the expression of relationships that are simultaneously both mathematical and logical in nature.
The clearest explanation of Boolean algebra comes from watching the actual process by which a computer controls and operates its logic gates. Each gate consists of an electronic switch with two wires leading into it, and one wire leading out. A wire with voltage on it is said to be “high.” A wire with no voltage on it is “low.” The two wires and two voltages leading into the gate allow for four possible combinations. How the switch interprets these different combinations—and thus allows either a high or a low voltage out the other side of the gate—is controlled by something called a truth table.
A truth table might instruct a gate to open if and only if both wires leading into it are high. Or the truth table could tell a gate that one high wire is sufficient for opening. A switch operating according to the first kind of truth table is called an AND gate, while the second switch is known as an OR gate. The formal logic behind the working of an AND gate is expressed by the statement, “If, and only if, A is true and B is true, then their combination is also true.” When the truth tables of AND and OR gates are inverted into negative propositions, they produce NAND and NOR gates, and these four different kinds of gates are the logical building blocks out of which all digital computers are constructed. The truth table for an AND gate is reproduced here. A and b represent the two lines leading into the gate. C is the line leading out of it. An upward-pointing arrow represents a high voltage, or a 1, or a “true.” An arrow pointing down equals a low voltage, or a 0, or a “false.”
Figure 2. The truth table for an AND gate.
“You can specify what happens to the output line in a computer for every possible input,” Norman informs me. “From these elementary logical devices, which are fairly easy to build electronically, you can make far more complicated systems. The microprocessor itself is composed entirely of these logical building blocks. So too is the memory. The microprocessor holds thousands of blocks, and each block in turn contains several transistors capable of being oriented as either ‘on’ or ‘off.’
“Of the five logic chips used in the Project’s computer, four are made entirely of NAND gates, which are AND gates inverted into their negative. The truth table for a NAND gate will have a 1 wherever the AND gate truth table has a 0. Before you go to the store to buy some chips, you first have to figure out what logical functions you want them to perform. Do you want AND gates or NAND gates? Once you decide on a function, you look in a data book to find out which chips are available for performing that function. Then you drive over to the Valley, step up to a counter, and say, ‘Could I please have a 7400 or a CD 4001,’ and they’ll know exactly what you’re talking about.”
Late in the afternoon Norman declares, “That’s it for the chips. The rest of what you see here are a few resistors and capacitors, an amplifier, a filter, and a couple of transmitters and receivers for sending signals from shoe to shoe. The only other really important part of the computer is this little aluminum canister. It holds the clock that governs the timing of events in the microprocessor. You need some sort of clock to step the computer through its sequence of activities. This one is made out of a quartz crystal that oscillates at one megahertz, which means that the clock puts out a little square wave that changes from 1 to 0 a million times a second. During every cycle of the clock, the microprocessor does something. It fetches an instruction, executes a command, or puts a byte out on the data bus.
“This gives you some idea of the speed with which all this is happening. It’s going on very fast—much faster than human temporal perception—and this is one of the reasons why we use a microprocessor to predict roulette. Human beings are slow compared to the speed of electronics. We’re doing quite well to get our reflexes down to a tenth of a second. That’s a hundred thousand times slower than a microprocessor. Of course,” says Norman, leaning back in his chair and grinning, “we can do other things that take a microprocessor a long time to do, like carry on a conversation.”
Missing from this collection of chips and components pinned to the Styrofoam are a number of elements crucial to the working of the computer, such as batteries for running it and solenoids for outputting its predictions. But the brains of the operation—logic gates by the thousands oriented into a computer program—lie in front of me. “Basically, that’s all there is to a microcomputer,” Norman concludes, putting the top back on the Kodak box. “See how simple it is? One of the most amazing things about the computer industry is that you can use microprocessors and microcomputers—you can control them and tell them what to do and build them into circuits—without having the faintest idea what a transistor is, or what’s actually going on electronically on a chip. Once you specify its characteristics, you don’t have to know a single thing about how a chip is constructed. Even to first approximations, you can be totally ignorant about all this stuff, and you can be ignorant, as well, about the old-fashioned electronics. All you have to do is manipulate chips like little black boxes. It’s magic, but all you care about is whether it works.”
I climb the front steps of 707 Riverside one evening as a guest invited for dinner and roulette. I push open the front door, unlocked as usual, to find Norman sitting on the floor in the hallway. All six feet two inches of him is hunched over a white box that turns out to be, on closer inspection, a portable computer complete with keyboard, one-line display, and modem linking it by telephone to the mainframe up on campus. “I’m on line,” he says, looking up to give me his crooked smile. “You know how it is. When the muse calls, you have to be there.”
I leave him writing chaos functions into the machine and walk into the living room where Lorna, in black tights from her Jazzercise class, with her brown hair cut in the original Jane Fonda look (parted in the middle and falling around her face in terraces), is watching I Walked with a Zombie on TV. “It’s sort of thirties French expressionism with incredible lighting,” she says. “There are some great scenes.” From her cockpit in front of the TV, Lorna runs most of the daily affairs at Riverside. If you want to know whether the kitchen is stocked with cumin, ask Lorna. If you want to find last month’s phone bill, ask Lorna. She is newly re-enrolled at the university to finish her B.A., and coursework is cutting into her daytime viewing, but if you want to know the plot to any movie ever projected onto a screen, ask Lorna.
The smell of dinner cooking lures me through the dining room into the kitchen. Tonight at the stove I find Jim Crutchfield, which foreshadows Chinese cuisine of note. He is busy marshaling his ingredients on a long table in the middle of the room. I see piles of shaved vegetables, snow peas, bamboo shoots, curls of ginger, and slivers of fish marinating in rice wine. Assistant chef tonight is Grazia Peduzzi, a Milanese leftist now teaching Italian at the university. Dark skinned, with an aquiline nose and flashing smile, she breaks into gales of laughter as Crutchfield shows her how to milk the liquid out of tofu wrapped in cheesecloth. Flushed from an afternoon run on the beach, Doyne and Letty walk into the kitchen wearing cut-off shorts and jogging shoes. The last person to arrive, weighed down under a video camera and tape deck, is Ingrid. “I’m making a movie about women in science working together in groups,” she says. “But it’s taking forever. The funny thing is, I’m doing it with three other women and we can’t agree on anything.”
&n
bsp; There is copious food for eight as the serving dishes fly around the dining room table. Doyne and Norman are legendary eaters from their days back in Silver City when they could pack away a couple of birds apiece at the Holiday Inn all-you-can-eat chicken night. On seeing that chopsticks do nothing to slow them down, Grazia breaks into another round of laughter. Crutchfield eats with the abstracted look of a chef tasting his sauces. Ingrid picks up a copy of Good Times and announces the movies playing in town. Lorna gives a running commentary on their actors, directors, and plots. She comes down hard on anything “sexist or sappy,” but everyone finally agrees on seeing a double feature at the local rerun house of The Damned and Death in Venice. “That should be a heavy dose of freak-out,” Ingrid says.
As we pass through second and third helpings, the conversation veers into talk about computers and evolution. “In fifty to a hundred years we’ll be accustomed to seeing self-reproducing machines,” Doyne declares.
“But at the moment we don’t even have self-repairing machines,” Ingrid says.
“It’s not far from self-repairing to self-reproducing. Until they can reproduce themselves, computers aren’t even as smart as amoebae. But from the first computer built in the 1940s to now is such a short period of time that their evolutionary prospects are amazing. From amoebae to slime mold to frogs to Homo sapiens, the next step up the evolutionary ladder might be machines.”
“We could be the first species to design our own successor,” Letty says. “The new species might come from one of two directions: genetic engineering or self-reproducing machines. But we don’t even have the language yet to talk about these possibilities. Our words for machines and species and evolution are all too confined to the old meanings.”
After dinner, as everyone else leaves for the movies, Doyne and I walk out the back door and down the stairs to the garden. We stand for a moment in the yard. The night air is sweet with the smell of magnolia blossoms and sour with the odor of anchovies washed ashore during their yearly run into the harbor. The earth in the garden is freshly turned. Over the back fence the horizon to the west is lit with opalescent corals and blues, as if ocean and sky are on the verge of exchanging places. Except for the swish of waves on the shore and the barking of sea lions swimming under the town wharf, the night is perfectly still.
The Eudaemonic Pie Page 25