2 The Perils of Prediction

“DROP IN BY ROCKET PLANE ON TOTTENVILLE, the sootless garden city where you’ll live in scientific comfort in AD 2000. You’ll eat food from sawdust, shop by picture-phone, [and] cook on a solar range.” This vision from 1950 also predicts vacuum-tube electronics, automation controlled by holes punched in a roll of paper, and houses built of plastic and metal. Family helicopters are common, dirty plates dissolve in hot water and are rinsed down the drain, and influenza and other ailments are no longer complaints.

Everyone wants to know what the future will be like, but as we can see, accuracy is not always possible. In this chapter, we’ll explore the art of prediction and consider ways to see it more clearly. I can’t give clairvoyance, but I do hope to point out some of the constraints on predictions and offer insights to evaluating them. Predicting is a very difficult undertaking: one thorough analysis of past predictions concluded that no more than a quarter of them were accurate. Or to be more specific: Are we trying to achieve more accurate predictions or be sharper in our assessment of those we hear, or both?

Poor Predictions

Prediction is very difficult, especially about the future.

—NIELS BOHR, physicist

Predictions are indeed difficult, but predictions about technology seem especially prone to error. Here are some famous failures.

There is no reason for any individuals to have a computer in their

home. (Ken Olson, founder of Digital Equipment Corp., 1977)

Television won’t matter in your lifetime or mine. (Radio Times

editor Rex Lambert, 1936) The radio craze will die out in time. (Thomas Edison, 1922)

This “telephone” has too many shortcomings to be seriously considered

as a means of communications. The device is inherently of no

value to us. (Western Union internal memo, 1876)

Rail travel at high speed is not possible, because passengers, unable

to breathe, would die of asphyxia. (Dr. Dionysus Lardner, professor

at University College London, 1823)

What can we conclude from this list? Obviously even the experts are too timid when predicting how technologies advance. Take courage, use your imagination, and see a bold new future!

But there is another category of wrong predictions, a larger and more influential category: the overpredictions. These are the dangerous predictions, the ones that stick in our minds and support the myth of exponential change.

Marie Curie predicted that radiation would prolong life (this was in 1904). Ironically, she died from leukemia due to overexposure to radiation.

All trees in the United States will be gone by 1920, cut down for heating and cooking (1890).

Fast electric ships will cross the Atlantic in two days (1900).

Atomic energy would “transform a desert continent, thaw the frozen poles, and make the whole world one smiling Garden of

Eden” (1908).

Thomas Edison predicted, “In 15 years, more electricity will be sold for electric vehicles than for light” (1910).

Animal parts (a chicken breast, for example) will be grown separately, without the need to raise the whole animal (1932).

Buckminster Fuller imagined cities housed under domes (1965).

We’ll have moon bases and passenger rockets to the moon by 1980

and robot soldiers by 1990 (1966).

Electromagnetic fields are so beneficial that classrooms will be deliberately enveloped in them to help students remember better

(1980s).

Let’s try a thought experiment to see how hard it is to make a successful long-range prediction. Ben Franklin, who lived from 1706 to 1790 and was a man of quick and inquisitive mind, once wrote that he wished to wake up in the future. Suppose we could give him his wish. Before we watch him as he marvels at the twenty-first century, however, let’s take advantage of his naive view of our world. We’ll give him a list of impressive developments since his day, some of which have actually come about, and some that haven’t. Would he be able to tell them apart?

The list could include instant worldwide voice communication, electricity to power household lighting and appliances, and flying machines that travel five hundred miles per hour. From the not-hereyet category, we could mix in mental telepathy, the ability to speak with the dead, and houses built of materials other than wood. We could include anesthetics and organ transplants, the ability to noninvasively see inside the body, and medicines that prevent or cure the worst diseases he knows, such as plague, yellow fever, and smallpox; add to that a one-day cure for broken bones, a medicine that removes fat, and cures for arthritis and the common cold. We could then throw in the moon landing, the hydrogen bomb, and advance warning for natural disasters mixed with underground cities, the extermination of mosquitoes and similar pests, and humans bred for specific characteristics just like crops and domestic animals.

We quickly see some as old news and the rest as speculation, but could Franklin do the same? I don’t think so. Why can we prevent smallpox but not colds? Why do we build one-hundred-story skyscrapers out of glass and steel but houses out of wood? Who would have guessed that in the twenty-first century, JFK airport outside New York still can’t find a better way to keep birds from runways than falcons?

If we were in Franklin’s position and given a list of future predictions, how would we tell the winners from the losers? For example, Microsoft was just another small company with a big dream until the PC became hot. If you’d seen their business plan among a thousand others in 1975, would you have singled them out for greatness?

One way to explain the poor record of predictions is with Amara’s Law, offered by Roy Amara of the Institute for the Future, which states that we overestimate short-term changes and underestimate longterm changes. The short-term part is pretty easy to explain: when a new technology comes along and begins to catch on, it gets a lot of press. Much of the talk is necessarily speculation or hype, establishing expectations the technology can’t possibly meet.

Underestimating the long-term changes means underestimating how thoroughly today’s technology will eventually insinuate itself into our lives. Looking backward, electricity, cars, the telephone, and other mature technologies are everywhere today, in more places than could have been predicted. It may be more relevant, though, to say that by the time we reach that “long-term” point somewhere in the future, the impact from the technology in question will not be noticed. For example, the pervasiveness of electricity and other mature technologies today would be impressive only to people at the dawn of those technologies—now they’re taken for granted, and we don’t care.

Another aspect to long-term change, as the Franklin example illustrates, is that when a technology is completely new (in the lab or before) and not an extrapolation of a product that exists in the market today, long-term changes that result from the technology are almost impossible to predict. In 1880 predictions about how flight would affect us by 1920 were not off only by a matter of degree (they missed the number of airplanes by a factor of ten, say), they were completely off target. Similarly, today’s impressive new developments such as the Internet and the PC weren’t underestimated forty years ago—they weren’t estimated at all. They weren’t even on the soothsayers’ radar.

Here’s a summary of this updated law assessing how we will predict, or “mis-predict,” technological change. Imagine the year is 2010 and we’re making predictions for 2015 (short term) and 2040 (long term).

1.In 2010, we will overestimate the impact that our new technology will have by 2015.

2.In 2040, we will find that we underestimated how pervasive 2010

3.technology would become, but no one will care: what was new and exciting in 2010 is ignored in 2040, because something else has become the exciting new technology du jour.

4.In 2040, we will see that our 2010 attempts at long-term predictions of technology not yet present in 2010 will be completely wrong.

A thorough discussion of how to make good predictions and spot the bad ones would take a complete book. But because so much of technology hype comes from predictions, some guidelines for evaluating them follow. For more on this subject, I suggest Megamistakes by Steven Schnaars.

Don’t Get Stuck in the Present

One of the problems with predictions of the future
is that by the time it’s clear that they have had little resemblance
to actual events, it’s too late to get your money back.

—RAY KURZWEIL,
The Age of Spiritual Machines (1999)

The first step in evaluating predictions is to discount those that assume exponential change. If exponential change were widespread, we would find that most predictions underestimate, and reality would outpace the prediction. And yet the opposite is true. What’s changing exponentially in many cases is expectation, not technology.

Predictions are often more a picture of the present rather than the future. The 1960s and ‘70s saw predictions of nuclear-powered planes and vacations in space because nuclear power and space exploration were the hot topics. The workweek was shrinking, so predictions about work in the future were also common. Predictions about the population explosion dominated other dire scenarios, but they were wrong because they assumed that the issues, priorities, and concerns of the present would continue unchanged into the future.

We heard about depletion of energy reserves and environmental degradation. Clearly, society had to use less energy. During the oil embargo of 1973-74, an assertion that anxiety about energy use was just a passing fancy would have been seen as ill-informed and even irresponsible—but would have been correct. Once the pressure was off, the issue faded from view. At this writing (2005), oil prices have shot up and we care again.

In 1980 Newsweek magazine predicted that robots could replace at least half of U.S. factory workers within twenty years. Of course, many factory workers were replaced, but most of the lost jobs were outsourced to other workers, not robots. A 1966 forecast envisioned robot tractors, indoor farms, irrigation with desalinized seawater, and synthetic meats. It missed the real farming issues such as fluctuating prices and a move away from beef for health reasons. In 1967 the future of merchant shipping looked nuclear. Actually, the real issues were increasing international competition and container ships. An 1893 forecast projected a massive expansion of the railroad as well as pneumatic tubes to carry both mail and people, completely missing the huge impact of the car and the airplane.

Thoughtless extrapolations are another danger of an exaggerated fixation on today’s issues. If today there are jet planes, tomorrow there must be supersonic planes, and a plane’s capacity will increase to a thousand passengers. If today has television, tomorrow must have 3D or holographic TV, and telephones will become videophones. If today many diseases are under control, tomorrow we will be able to control them all, and life spans will reach one hundred years. If today we have modest weather prediction, tomorrow we will predict weather a year ahead. And if we’ve just finished the Empire State Building, mile-high buildings are next.

Mark Twain had some comments about careless extrapolation. During his time, a number of engineering projects straightened and shortened the winding Mississippi River. He speculated about this trend.

In the space of 176 years the Lower Mississippi has shortened itself 242 miles. This is an average of a trifle over one mile and a third per year. Therefore, any calm person, who is not blind or idiotic, can see that in the Old Oolitic Silurian Period, just a million years ago next November, the Lower Mississippi River was upward of 1,300,000 miles long, and stuck out over the Gulf of Mexico like a fishing-rod. And by the same token any person can see that 742 years from now the Lower Mississippi will be only a mile and three-quarters long, and Cairo [Illinois] and New Orleans will have joined their streets together, and be plodding comfortably along under a single mayor and a mutual board of aldermen. There is something fascinating about science. One gets such wholesale returns of conjecture out of such a trifling investment of fact.

Tomorrow will look more like today than most predictions would lead us to believe.

Avoid Technology Infatuation

No sensible decision can be made without taking into account
not only the world as it is, but the world as it will be.

—ISAAC ASIMOV, author

Technology is often thrilling, but too often we let our excitement cloud our projections. People still use products for basic reasons, and a new product must work with those reasons. When evaluating a prediction, challenge all assumptions, avoid over-complex logic, and use common sense.

Polaroid instant photography was pretty amazing. So were Polavision instant home movies. But Polavision was a short-lived experiment and Polaroid photography was never more than a niche. As amazing as technology can be, that alone doesn’t make for a successful product.

One 1967 view of the future home saw disposable dishes and inflatable furniture. Don’t worry about spills—just hose down your plastic furniture and let the water run through a drain in the floor. But just because a product is possible, doesn’t mean people will want to buy it.

Scott Paper sold paper dresses beginning in 1966, and five hundred thousand were shipped in six months. You could shorten your dress with scissors, customize it with paint, and discard it after wearing it once. It was a fad whose popularity lasted only a few years, like CB radios a decade later. Neither answered real customer demands.

Predictions about the success of these products were based on such infatuation with the technology itself that they didn’t consider whether the consumer would actually care. Other proposals that never inspired enough demand were underwater hotels, artificial moons for lighting cities at night, dehydrated or irradiated food, moving sidewalks, geodesic domes, cars that drove themselves, and paperless offices.

Even big companies make mistakes. GE and Motorola were invested in CB radio, AT&T and Sears in videotex (a precursor to the Internet), and the New York Times and RCA in fax newspapers (a mini newspaper sent by radio and printed in each home). In the final analysis, consumers just didn’t need what these products provided.

The fax newspaper provides a good case study of the results of infatuation with technology. For a few years in the late 1930s and in another burst of enthusiasm ten years later, dozens of radio stations broadcast them. Some were published four or more times daily to give readers the latest news. David Sarnoff of RCA saw this to be as promising a new technology as television, and one journalism school offered a class in fax newspaper production. Newspapers’ attitudes toward the product were identical to those of many companies toward the early Internet: we’re not sure how to make money in this business, but if we don’t jump in, we might miss something big.

Too often, forecasts about a fledgling industry begin with the assumption that success is inevitable. The only question remaining is: What growth curve best documents that success? And yet, success is not inevitable; it’s not even likely. Most new products fail.

A good forecast shows analogies to past successful products, but it also examines failed products to show why the new product won’t be like them (see figure 2). Successful products follow an S-curve: an S-shaped graph of product sales over time that shows slow growth initially, fast growth as the product becomes mainstream, and slowing growth as the market becomes saturated. A forecast made in advance of sales, however, and assuming that sales will follow an S-curve is very optimistic since an S-curve only applies to successful products. The assumption of fast growth for the videophone or supersonic passenger airplane or any other failed product gave the wrong answer no matter how clever the presentation or how powerful the supercomputer that helped with the analysis.

A long and complex argument with fancy analysis and statistics is another warning sign of erroneous predictions: this usually obscures more than it illuminates. The fundamental market factors are still pretty simple. Who are the potential customers, and how many are there? What product or process will be replaced and why is the new approach better? Where is the new approach worse (for example, do you require customers to change their habits)? What social trends work to the new product’s favor or detriment (changing concern for health or finances, for example)? Does the benefit outweigh the cost? Don’t forget to ask if the forecast came from a possibly biased source like a company or industry group that would benefit if the prediction came true. There are more questions like this, of course, but common sense takes you a long way to avoiding technology infatuation.

New Products Don’t Win on Every Point

“But he has nothing on!” at last cried all the people.
The Emperor writhed, for he knew it was true.

—HANS CHRISTIAN ANDERSEN,
“The Emperor’s New Clothes” (1837)

In the short story “The Man Who Came Early,” by Poul Anderson, a U.S. Army soldier stationed in Reykjavík is mysteriously transported one thousand years back in time.

Struggling in this strange land to find something he is skilled at and with which he can repay his new friends, he describes to the Vikings new boats using triangular sails that allow them to sail into the wind. Surely this is better than the Viking longboat whose square sails are effective only when the wind is at its back and must be rowed when moving in any other direction. But the Vikings point out a number of deficiencies. The new boat’s deep keel would prevent it from going up a shallow river or being beached, often done to find shelter from storms and protection against attack. In addition, there were no docks or piers to provide access to such sea-bound boats. The weighted keel would also be too difficult to make. In short, the marvelous new invention had far too many drawbacks.

A new technology is rarely superior to an old one in every feature. It can improve with time, of course, but out of the gate it’s not the obvious winner. We find many examples of this with modern products. MP3 songs are convenient, but the sound quality is worse than compact discs (it’s worse even than that of a new vinyl record). Laptops must be plugged in or frequently recharged, and they’re heavy and expensive, unlike paper and pencil. Computer monitors have just 5 percent of the resolution (dots per inch) of a high-quality book or magazine. Computer LCD displays cost more and have a narrower viewing angle than monitors (CRTs). Digital video (DVDs and digital TV) has new visual anomalies not present with analog video (such as cable or VCR). Web pages aren’t as high resolution as those in a paper catalog, and a Web site doesn’t allow the equivalent of quickly leafing through a catalog. A David Sipress cartoon illustrates this retrenchment with “The Off-Line Store.” Signs in the window read, “All items are actual size!” and “Take it home as soon as you pay for it!”

Another example of a new technology that is not necessary better is 3D movies. A screenwriter developing a script for such a movie must contrive scenes to show off 3D’s benefits. I remember one 1981 movie with a scene showing nuts poured down a well. The camera shot was from below, and the audience saw the nuts whizzing past them. This was definitely cool, but it had absolutely nothing to do with the story. The downsides to 3D films may be modest (not every theater is able to show them and viewers must wear special glasses), but the benefits still must be big enough to compensate. In its half century of existence, 3D hasn’t provided them.

This is also the marketing challenge of the videophone. It’s definitely cool, but it just isn’t that useful. Proponents imagine Grandma marveling at her new grandchild from across the country or a similar situation where video adds a lot to a telephone call. But these instances are rare. The minor benefits don’t outweigh the hassles of installing and using it. Perhaps only by making the videophone a prominent computer feature (and free) is it likely to get used.

Of course, MP3 players, calculators, laptops, digital video, and other products have all been successful. The point is that new technologies, even the successful ones, are superior to existing technologies on some features but worse on others. And the more features that compare poorly, the less likely a product will succeed. If you’re surprised at the slow acceptance of a new product, make an objective comparison of how the product competes on every point to find the logic behind the market’s indifference.

Consumers are surprisingly logical when evaluating products. Maybe a new audio player is both cheaper and higher quality than current players. Sounds like a winner, right? But consumers will want to know whether it supports their existing music library of CDs or MP3 files. If not, how much hassle and expense is it to convert to the new format? Are there players for the bookshelf, for the car, and for jogging? Can the new format be played on a PC? Will all music companies provide music in the new format? Must existing players be discarded and replaced with this new one? Consumers need compelling reasons to make such a switch.

Some industry watchers claim that this process is not always logical. They give as an example the Betamax video format. It had better video quality, and yet VHS became the dominant format. Did the market choose an inferior product in this case? No—Beta actually did not have noticeably better quality (Consumer Reports at the time showed it a toss-up) and, more importantly, VHS beat Beta on recording time. Another example: Did the market act illogically when it kept the ubiquitous QWERTY keyboard layout rather than the more logical Dvorak? No—QWERTY was entrenched when Dvorak came along, and Dvorak didn’t offer a big enough improvement to outweigh the hassle of the change. If the new product is better (in a comparison of all relevant features), it’ll sell. If it doesn’t sell, it’s not better. We must never forget to look at the big picture when predicting how a new product will do in the market. As Russell Ackoff said in his book The Art of Problem Solving (1978), “Irrationality is usually in the mind of the beholder, not in the mind of the beheld.”

Finding the Next Big Thing

Asking the right questions is superior
to finding elaborate answers to the wrong questions.

—STEVEN SCHNAARS, Megamistakes (1989)

Where will the Next Big Thing come from? It’s rarely from the company about to get hammered by it. It’s hard to predict the future when we don’t even know where to look.

The digital watch didn’t come from the established watch companies. The calculator didn’t come from the slide rule or adding machine companies. Video games didn’t come from Parker Brothers or Mattel. Semiconductors didn’t come from the vacuum-tube makers. The ballpoint pen didn’t come from the fountain-pen industry. The Internet browser didn’t come from Microsoft. Looking further back, cars didn’t come from wagon makers, refrigerators from ice companies, or light bulbs from candle makers.

More recently, traditional phone companies had to buy their way into the cellular telephone business. The top players in paper directories, like Thomas Register and Yellow Pages, are not the top players in Web search engines. The leader in robot vacuum cleaners is iRobot, not Electrolux or Hoover.

This shouldn’t be too surprising. The trick with a robot vacuum cleaner is the robot, not the vacuum cleaner. Skill in dreaming up board games doesn’t carry over into skill in designing video games. Vacuum-tube makers were not particularly well placed to see the need for (or path to) the integrated circuit. The market leaders are probably where evolutionary products will come from, but this is less likely for more revolutionary products. And when it comes to a completely new kind of product—something that gets the job done in a completely new way—it usually comes from outside the industry. The legendary garage shop as a source of innovative products makes the job of prediction much tougher.

There seems to be a rule of constant value with predictions: longerterm predictions have more potential value and yet are less likely to come true; short-term predictions are more certain, but they don’t tell us much new information. Long-term predictions can be a good starting point for a debate. We can then take steps to steer toward or away from that vision of the future as appropriate. But be skeptical in proportion to how far in the future the prediction tries to reach. Keep in mind that bubbly predictions form much of the foundation of today’s hype.

When you get the urge to predict the future,
better lie down until the feeling goes away.

—Forbes magazine (July 10, 1978)