Read an Excerpt
Probable Tomorrows
How Science and Technology will Transform our Lives in the Next Twenty Years
By Marvin Cetron, Owen Davies St. Martin's Press
Copyright © 1997 Marvin Cetron and Owen Davies
All rights reserved.
ISBN: 978-0-312-20737-3
CHAPTER 1
GET READY FOR DIGITAL EVERYTHING
"Future computers? Everybody in the business knows what's coming." The speaker was an engineer friend of ours who spends his days tinkering with next-generation prototypes. "In ten years, personal computers will be nothing like the boxes we use today," he said. "They will have the power of a supercomputer, but they will be no larger than a pack of cigarettes. You won't have to carry a keyboard or screen. Instead, just talk to the machine. Tell it what you want to know, and it will tell you the answer.
"Of course," he added, "by then, you could just wire the computer directly into your brain. But manufacturers have such a big investment in conventional technology that they'd go broke if they brought a neural interface to market that soon."
For special purposes, he offered two accessories, each of which would fit into a pocket. Those of us who are artists, designers, and writers all really want to look at our pictures or typescripts, not just talk to a machine about them. So a tiny projector mounted on a pair of glasses will beam a 3-D image directly into our eyes. Wherever we look, it will seem that a high-resolution color screen is hanging in space just where we can see it most clearly. And those who cannot quite free themselves from a keyboard will still be able to "type" and enter commands by pressing buttons. They will just have to carry an extra box, this one much like the remote control for a television. Even with these add-ons, the computer of 2007 will be one tiny package.
Our friend really does know what the computer industry is planning. He has spent the last twenty years of his life helping to develop salable products from the ideas handed down by its research departments. Yet we doubted about half of his forecasts.
For example, scientists have created computer chips that can be linked directly to nerve cells; yet that neural interface is a lot farther off than many optimistic engineers recognize. Before they can usefully wire a computer into our nervous systems, scientists will have to figure out exactly how the brain processes information. That has proved to be a lot more difficult than researchers once believed. And the surgeon who implants our new hardware will face liability problems that no insurance company would cover at any price. Someday, possibly, we will control our computers just by thinking at them—but not by 2007, or even 2017.
In contrast, that keyboard/controller is easy to build. However, we pity the poor marketing department ordered to sell it. Typing a single character on this little box requires pressing several keys at once—your index and middle fingers for "A," perhaps, your thumb and ring finger for "B," and so on. We have seen this concept before. In the mid-1980s, one company offered a single-hand "keyboard" that worked on the same principle. Engineers loved the idea. Customers hated the reality. Learn some weird new finger code? Most touch typists would rather graft a microchip to their brains.
Yet our friend's vision holds the essence of tomorrow's computers. Ten years from now, computers will be at home in our pockets—and inside our televisions and coffeemakers, and almost everywhere else as well. They will pack vastly more computing muscle into tinier, lighter packages. Compared with the dreams of some industry pundits, in fact, the forecast that began this chapter is remarkably conservative. If the most optimistic computer scientists are correct, tomorrow's shirt-pocket computer could hold a billion bytes in its working memory—about two thousand copies of this book—and run at 50 million times the speed of today's fastest personal computers. We have no idea what to do with all that computing power. We doubt that anyone else knows either. Even the lowest estimates of the advances coming in the next decade will give tomorrow's computers a level of utility and convenience we can only imagine today.
Computer engineers love acronyms almost as much as Pentagon bureaucrats do, and one of them has devised the perfect acronym for the future of computing: CADE. It stands for Convergence and Digital Everything.
Convergence is the grand coming-together of technologies that used to be separate. Radio and television and telephones used to work with "analog" signals. That is, they represented sound and images by varying the strength or frequency of their signals. (This is what the "M" of "AM" and "FM" means—"modulation," or varying.) Now they are going digital. Pictures and sound are broken into tiny pieces, and each fragment is represented by a series of computer bits. Just as CD systems give better sound than old-fashioned record players, digital TV gives clearer pictures than the conventional box that now dominates most living rooms. But a digital TV is not merely like a computer; it is a computer, with special software. And when you are not busy crunching spreadsheets or balancing your checkbook, your personal computer will be able to switch programs—in both senses—and display your favorite TV show. Your TV is a computer, and your computer is a TV. That is convergence. It is happening wherever technology meets information.
Already, the cheapest way to make a long-distance telephone call is to fire up your computer and log onto the Internet. Several companies provide telephone-style headsets for personal computers, together with the software needed to operate them. Using these devices, you can talk with any Internet user who has a similar headset anywhere in the world. At this early stage of development, the sound quality is variable—never up to that of a normal telephone but usually adequate. And the price can't be beat. One of the authors lives in southwestern New Hampshire, where he has an account with the more expensive (but more convenient) of two regional Internet service providers. At daytime rates, a conventional telephone call to Greenfield, New Hampshire—all of four miles from home, but not in the local calling area—costs $.26 for the first minute and $.24 a minute thereafter. Using the Internet, a conversation with someone in Greenfield costs $.005 per minute. So does a conversation with someone in London, Paris, Moscow, or New Delhi. The computer is your telephone, and your telephone is obsolete. Convergence.
This is just the beginning. For years, most of us have had three different sets of wires and cables entering our homes and offices. Electric wires carry power; telephone lines carry conversations and, increasingly, computer data; television cables carry ... umm, entertainment?—video signals, at any rate. Recently, some electric utilities have added a fourth set of lines that allow them to check meter readings without leaving the office; most of us are not even aware these connections exist.
What happens when all the information is digital? Why not carry TV pictures on the telephone wires? Computer data on the TV cable? Or both of them on the electric utility's meter-checking lines? At least one power company already is renting unused capacity on its data lines to carry digital messages for local businesses. One recent proposal would even ship computer data across the power lines themselves. Each of these ideas offers its own benefits, and each requires the solution of some technical and economic problems. We examine them more closely in the next chapter. For now, just think of them as one more example of convergence.
Digital Everything is an even simpler notion. Computers are becoming so small, powerful, and cheap that soon almost any object more complex than pottery will be equipped with its own brain. Lights will adjust themselves to illuminate your book or keep glare off the CV (computer/television) screen. Toasters will learn whether you like your English muffins lightly browned or charred beyond recognition. Intruder alarms will know enough not to call the police just because you left your keys on the dresser, even if you have not upgraded to one of the new voice-recognition security systems. Edward Cornish, president of the World Future Society, recently completed a study of computer-driven trends in society. He cites the example of one Texas man who computerized his sprinkler system and programmed it to come on when errant golfers from the neighboring course trampled his lawn! The Digital Everything revolution has already begun.
Take a more complicated example: Imagine a stove that arrives with all the recipes from Irma Rombauer and Marion R. Becker's classic The Joy of Cooking and Graham Kerr's latest TV program stored conveniently in memory. A ROM-card reader will let you add recipes from new cookbooks as well. Just tell the stove what you want to prepare—like most computers, it will understand verbal instructions—and it will display a list of ingredients on its flat-panel screen. It will announce when the skillet is hot enough to sear a steak, prompt you when the pasta is al dente, give fair warning when the next step requires exact timing. It will sense when a soup is beginning to boil and automatically reduce the heat to a slow simmer. It will schedule all your meal's courses to be done perfectly at serving time. With use, it will remember how you like your food. If your vegetables come out a little too raw for your taste at the factory setting, the stove will learn to cook them a bit longer. Of course, it will learn different preferences for each cook in the family. No doubt it will have many other "intelligent" functions that have not occurred to us.
Fifteen years from now, product designers will still be figuring out startling ways to use the new intelligence of everyday appliances. No one of these innovations will change our lives. But as the artifacts around us gradually learn to accommodate our individual needs, the world will become a friendlier, more convenient place in which to live.
We return to some of the more promising uses of tomorrow's computers later in this chapter. For now, let us look at the technologies that will make them possible.
SILICON MAGIC
When it comes to microprocessors, even our minimum expectations are astonishing. In fifteen years, these chips will be about 15,000 times more potent than the processors that power today's cutting-edge personal computers. Tomorrow's run-of-the-desktop computer will finish in an hour a task that today would keep our most powerful desktop computers running twenty-four hours a day for two years. This seems even more amazing because it requires no revolutionary technological breakthrough. The futuristic superchip of 2010 will be much the same mass of transistors etched into silicon that engineers have been improving since the first commercial microprocessor, the Intel 4004, appeared on the market in 1971.
The difference lies in greater component density and more sophisticated design—the same two factors that make today's advanced microchips so much more powerful than the primitive processors of the 1970s. That first Intel 4004 packed some 2,300 transistors into a space roughly the size of a fingernail. The Intel Pentium Pro, currently the firm's most powerful chip, contains 55 million transistors in a space not all that much larger. There are a lot more circuit elements to do the work, and they are packed more closely together, so the data can move between them faster. This is one reason the 4004 could carry out one program instruction 60,000 times per second, while the P6 can perform 250 million instructions per second. (The fastest experimental processors made today are four times faster yet.)
The other reason is a matter of tactics. A computer program consists of an enormous list of minute steps: Bring one number from external memory into the processor; bring in another number; move one of them from its temporary storage register into the number-crunching area; add the other number to it; move the answer into another storage register; look up the programmed instruction for the next step; and so tediously on. Early computer chips processed these instructions one at a time until the program was complete. It was a lot like trying to move the entire population of New York City through a single subway turnstile.
In recent years, computer engineers have worked out ways around that bottleneck. Modern microprocessors are "pipelined" and "superscalar." Pipelined processors move several instructions through the system at once, opening more turnstiles. A single stage of pipelining halves the time it takes to work through a program. Current processors are nearing ten stages of pipelining, and their descendants will have many more. Superscalar processors can perform several instructions at once, in effect stuffing several people through each turnstile at the same time. Again, this multiplies the machine's effective processing speed. Current processors carry out three to six instructions at one time. In fifteen years, the number is likely to be several dozen. These incremental advances alone will make tomorrow's computers several hundred times more powerful.
We are less optimistic about another strategy from which researchers have long expected much greater advances in computing speed. This is parallel processing. This technique aims to break a problem into many smaller tasks, perform each one simultaneously on its own processor, and then recombine the results of the individual computations into a single answer. In theory, this should be the ultimate upgrade.
At this point, we have worn out our subway analogy. Instead, imagine adding two very large numbers. We add the ones column, carry a number, add the tens column, and so on. A parallel processor might tackle this simple problem by adding all the columns at the same time, each on its own subcomputer; still more processors keep track of the carries. In the end, the machine combines the results of all these separate calculations into the final sum. Nothing could be faster, assuming that it can be done at all.
It turns out that making such a computer is not terribly difficult. Engineers have built parallel-processing computers that contain hundreds of identical microprocessors; a few small companies even specialize in making this kind of machine. Computers with many thousands of processors already are on the drawing board. For certain specialized kinds of calculation, nothing else approaches the speed of parallel processors.
However, distributing each program among many processors has proved to be almost as hard as passing one of our New Yorkers through several turnstiles at once. It is difficult enough to separate most computing problems into easy-to-process fragments, and that is only the first hurdle that programmers face. Distributing those many parts to individual processors, keeping the subcomputers in step with each other, and then reassembling their answers into one grand result has proved all but impossible, save for those few specialized chores that lend themselves to subdivision. These problems will not be solved until someone achieves the kind of conceptual breakthrough whose appearance no one can predict. We suspect that programmers will still be struggling with them fifteen years from now.
While they do so, engineers will overcome some obstacles that make it difficult to continue cramming ever more transistors into silicon little larger than a postage stamp. It will not be easy, because the circuit elements of a microprocessor are so tiny already. The smallest structures on today's memory chips are only 0.35 micron across, or 35 hundred-millionths of a meter, just over one-three-hundredth the diameter of a human hair. By 2000 or so, they will be 0.1 micron across, or roughly the width of a coil of DNA.
Microchips are made by photolithography. Though the process has become extraordinarily complex, in practice it is simple, much like printing an ordinary photograph. The printing machine, known as a stepper, shines light through a mask onto a wafer of silicon, which is coated with a material called a photoresist. The mask is a simple black-and-white picture of the circuit being created. (Today's masks are made of chromium deposited on quartz.) Wherever light penetrates the negative, the photoresist hardens; where the negative blocks the light, the resist is unchanged. Unhardened resist is then washed away, leaving bare silicon. These exposed areas are etched with acid to create the basic circuit, coated with metal to form conductive areas, and treated with a variety of other materials that alter the silicon's electrical characteristics. A single chip can contain as many as twenty layers of circuitry, each one laid down by this exacting process.
Those circuits have become so small that photolithography is hard-pressed to make them. The waves of light itself are as large as the components, so they cannot be focused sharply enough to produce a usable image on the silicon. The traceries of photoresist that adhere to the nascent chip are too bulky as well, and when made smaller they tend to flake off the silicon. Over the years, chip makers have moved from using visible light to ever-shorter wavelengths of ultraviolet, and they have developed new resists able to capture ever finer circuit details. These refinements have been a major factor in the development of today's densely packed microchips, but they have nearly reached the end of the line. In the next generation of processor chips, or at the latest the generation after that, the components will be so small that radical changes will be required to make them. For state-of-the-art chips, photolithography soon will be obsolete.
(Continues...)
Excerpted from Probable Tomorrows by Marvin Cetron, Owen Davies. Copyright © 1997 Marvin Cetron and Owen Davies. Excerpted by permission of St. Martin's Press.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.
