Forty-six years ago, Intel co-founder Gordon Moore published his observation that the number of transistors that could be put on a silicon chip doubled about every two years—meaning that computer capacity would also increase geometrically.
At the time, Moore believed that this growth might continue for “perhaps 10 years.” In fact, he was off by three decades and counting. Over time, advances in chip technology have proceeded almost like clockwork, with capacity sometimes doubling even faster. This spring, Intel announced yet another new chip that keeps Moore’s Law right on track.
Intel says a vertical ridge on its new 22-nanometer Ivy Bridge 3-D Tri-Gate chip creates much more real estate for transistors. “Just as skyscrapers let urban planners optimize available space by building upward, Intel’s 3-D Tri-Gate transistor structure provides a way to manage density,” according to the May 11 company announcement.
But how long can this doubling go on?
At one extreme, pessimists argue that it can’t continue for too much longer—that the circuits are becoming so tiny now that physics will prevent it. Already, the scale is difficult to comprehend: Intel’s May release noted that “more than 6 million 22nm Tri-Gate transistors could fit in the period at the end of this sentence.” The pessimists say that the shock of running into a wall will be a terrible blow to the high-tech industry, which has worked in an environment of exponential growth for generations.
At the other extreme, some futurists contend that Moore’s Law will hold for many more decades. Indeed, they say, we stand at the cusp of a “singularity,” another point like the invention of agriculture or the industrial revolution when suddenly everything about society changes.
Some of these writers and thinkers predict that the time will soon come (writer Ray Kurzweil points to 2025) when computers will be, briefly, as smart as people—shortly after which, they will surpass us, leading to some very strange consequences. In his 2005 book, The Singularity Is Near (Viking), Kurzweil argues that “the pace of change of our human-created technology is accelerating and its powers are expanding at an exponential pace … within several decades, information-based technologies will encompass all human knowledge and proficiency, ultimately including the pattern-recognition powers, problem-solving skills, and emotional and moral intelligence of the human brain itself.” By the end of this century, he believes, “the nonbiological portion of our intelligence will be trillions of trillions of times more powerful than unaided human intelligence.”
Kurzweil says that this shift will lead to revolutions in medicine and other fields as brains faster and better than ours solve problems with ever-greater rapidity. Others see negatives ahead, including a complete collapse in the value of human labor—and maybe even a decision by the new management that our services are just generally no longer required.
This scenario might sound far-fetched, but the singularists can’t be dismissed out of hand. Exhibit A is the astonishing technological progress made in the past 50 years. In his book, Kurzweil talks about being given access to the biggest computer in New England in 1968, an IBM 360 Model 91, which had 1 megabyte of core memory and rented for $1,000 an hour. Today, the most basic Dell desktop comes with 2 gigabytes of random access memory—2,000 times more memory than the 360 had—and it can be purchased for $279, or $43.80 in 1968 dollars.
High-end capabilities keep growing as well. In 1997, Deep Blue, an IBM computer, beat Garry Kasparov in chess. This year, Watson, another IBM super computer, beat two all-time champions of the long-running U.S. trivia game show, “Jeopardy”—perhaps less of a mathematical challenge than the chess game, but a testament to the virtues of a 4-terabyte memory. (A terabyte is 1,000 gigabytes.) “I for one welcome our new computer overlords,” quipped contestant Ken Jennings at the conclusion of the three-day match.
Whether Kurzweil’s singularity is near or far, the consensus seems to be that chip capacity, bandwidth speeds, and storage capacity—the three key elements of IT growth—will all keep increasing at exponential rates for at least the next 10 to 20 years.
Chip capacity – Intel’s latest chips are just 22 nanometers wide, not far off from the point where the rules of classical physics no longer apply. Ten nanometers is the absolute limit for transistor miniaturization—any smaller and you’re getting close to the atomic level, where non-classical effects arise, says Bruno Thedrez, a professor of communications and electronics at ParisTech. However, he adds, stacking chips will provide a third dimension for growth for some years to come.
Bandwidth – Jakob Nielsen, the Internet design guru, has noted that the amount of bandwidth available to high-end Internet users has grown by roughly 50 percent every year since 1983. Nielsen, in an update, recalls that he wrote his original article on this topic in 1998 on an ISDN line that could download 1 megabyte per second, and that this year he could download 31 mbps.
Looking ahead, Thedrez says that scientists have pushed 100 gigabytes per second over a single-wavelength optical fiber and that within 20 years there is now no technical reason that fiber capacity couldn’t climb to 50 terabytes, or 50,000 gigabytes.
Wireless bandwidth growth is a bit more problematic, Thedrez says, first because the power required would make your phone too hot to handle, and second, bandwidth is already divided by governments. It’s a lawyer limit, he jokes, the one kind physicists have no tools to fight against.
Storage – Mark Kryder, former chief technology officer at Seagate, the hard-disk company, and now a professor of electrical and computer engineering at Carnegie Mellon University in Pittsburgh, says that the capacity of hard disk drives increases about 40 percent every year. He estimates that by 2020, people will be able to buy a 15-terabyte, 2.5-inch disk for less than $150—and says that he thinks there is a good chance he is being too conservative. To put it into context, that’s a much more powerful and compact disk than anything available on the market today for consumers at any price. By comparison, Best Buy now offers a 10TB external disk for $1,799 that’s about 5.9x7.3 inches.
However, Kryder says he suspects the storage industry’s 40 percent growth rate will break down between 2020 and 2025. At that point, he says, the industry will either move toward a different technology or, if exponential growth is continuing in other parts of the computer, such as the processor, more effort will be made to ensure that storage can keep up. “If one component (processor, memory, storage, for example) becomes much more costly, resources will be shifted to bring its cost down and in line with the others,” he says. “Economics really does control what gets developed.”
But do we really need a 15-terabyte disk drive? “Throughout my 45 years of working in the storage industry, people have always said that they couldn’t imagine what they would do with more storage capacity, but demand has always been there for the exponential increase,” Kryder says. “Today, for example, we look at YouTube videos that have much less than HD resolution. Why don’t YouTube videos come in 3D HD? As processing speeds, communication speeds, and storage capacities increase, they probably will.”
Assuming all this acceleration stays on track, will it matter? Beyond the joys of 3D YouTube video, will such advances bring any real gains to humanity?
The answer seems to be, yes. While Thedrez observes that applications have always expanded to fit the available bandwidth—sometimes in ways rational thinking would not predict—he also says that future growth should help the trend toward cloud computing and pave the way for the coming Internet of Things.
In addition, many complex problems should become easier to solve. Consider the impact on medicine, where computers seem likely to change everything.
Gene sequencing, for example, is becoming wildly cheaper, almost by the month, a trend that should produce big medical breakthroughs much sooner than anyone had a right to hope. The first complete map of a human genome cost $3 billion, and was finished in 2003. Two years ago, a Stanford engineer announced he had developed a way to do it for $50,000. In 2010, that figure dropped to $10,000. Now, 18 startups are racing to be the first to sequence a genome for less than $1,000. It’s happening so fast that the computers can’t really keep up with the data being generated. After all, computers are improving “only” at Moore’s Law speed.
For quadriplegics, life might become much easier soon, thanks to leaps in computer power. Toyota is experimenting with a wheelchair that can steer by mind-control. (Honda, meanwhile, is developing mind-controlled robots.)
And IBM is sending Watson, its Jeopardy-winning computer, to medical school. The company is working now with Nuance Communications and two U.S. medical schools to develop a program that will help doctors diagnose patients. Estimated time of arrival: 18 to 24 months.
“We have now been living with exponential growth for decades and both industry and society expect it to continue,” Kryder says. “So long as Mother Nature allows us to sustain that exponential growth, I believe it will be maintained.”
Of course, hardware isn’t the only way forward. Even when IT engineers begin to hit physical barriers, many advances in programming might still be made, through gains in parallel processing, suggests Andrew Odlyzko, a professor of mathematics at the University of Minnesota and a technology historian.
However, beyond a certain point, even massive gains may start to feel incremental. When it comes to communications, for instance, Odlyzko argues, the phone was much faster at transferring data than snail mail, and DSL much faster than dial-up, but moving from DSL at 6 mbps to high-speed broadband at 8 mbps may not matter all that much.
In the end, perhaps the only certainty—whether computers end up at the top of every class or just remain extraordinarily useful machines—is one that Odlyzko recognized in 1999, when he was head of mathematics and cryptography research at AT&T Labs: computers will stay annoying.
“We were frustrated with computers a decade ago, we are frustrated with them now, and will continue to be frustrated in the future,” he said then. “As long as technology offers enticing new products and services, we will continue to live on the edge of intolerable frustration.”