IT IS ELEVEN feet long, seven feet tall, and weighs three tons. Nathan Myhrvold of Microsoft has provided a million dollars to ensure its completion and fund a replica for his house on Lake Washington. The last of its brass and steel pieces are now being slotted into place in a special gallery at the Science Museum in London. It is one of history's great might-have-beens, an eerie reminder of a technological dead end, the failed catalyst of a revolution that never was. This month, 150 years later than planned, work is finally underway to complete the final section of a mechanical computer designed by the legendary Charles Babbage, the 19th-century mathematician revered today as the grandfather of modern computing. A mechanical Moore's Law?
Part of Babbage's Difference Engine No. 2 was constructed in 1991 by a team at the Science Museum, but about a third of the original design -- the elaborate output printer mechanism -- was omitted to save money. So it is only now, with an infusion of cash from Microsoft's maverick head of research, that Babbage's marvel of Victorian engineering is being realized in its entirety, making it possible to evaluate fully for the first time the likelihood of an intriguing "what if?" scenario. Suppose one of Babbage's numerous attempts to build a mechanical computer, all of which failed, had instead succeeded: would a revolution have followed, built on the back of his ingenious designs?
This possibility was most famously explored by William Gibson and Bruce Sterling in their novel The Difference Engine, a classic of the steampunk genre. The resonances between the modern age and the book's putative Victorian computer revolution are intriguing. Could computers have supercharged the industrial revolution? The answer is trickier than you might expect. Indeed, it is only now, with the parts for the printer built and delivered, that it is possible to put a final price tag on Difference Engine No. 2 -- a figure that gives us the odds on Babbage's great gamble.
CHARLES BABBAGE was not the first person to try to build a mechanical calculating machine. The first known example dates from 1620s, when a German inventor named Wilhelm Schickard built a "Calculating Clock" capable of multiplying two six-digit numbers. Similar devices were built by a number of inventors in the following two centuries (one of whom, a French inventor named Rene Grillet, was so obsessed with keeping his design a secret that nothing is known about it today). Finally in 1820 a Frenchman called Thomas de Colmar built the first arithmometer, a device with a typewriter-like carriage which could multiply two eight-digit numbers in 18 seconds. It was the first commercially viable mechanical calculator, though it was not until the end of the 19th century that it was produced in any quantity.
Yet all these machines were essentially akin to modern pocket calculators -- rather than performing calculations in a pre-ordained sequence, they responded to human input and worked things out on demand. Babbage, however, had rather grander plans. He wanted to build a machine that would calculate and print mathematical tables automatically.
At the time, printed tables of trigonometric functions, logarithmic series, and planetary orbits were widely used as calculation aids by mathematicians, scientists, and navigators. But since every value in each table was calculated by hand -- one set of tables, compiled in France, was the work of 100 mathematicians -- there was a lot of scope for errors. Worse, even if the values were calculated correctly, there was still the chance that errors could be introduced in the typesetting process. One day in 1821, while comparing two supposedly identical sets of astronomical tables and discovering numerous errors, Babbage decided that there had to be a better way. "I wish to God that these calculations had been executed by steam!" he declared.
Babbage started to design a vast calculating engine, far more powerful and ambitious than anything that had been attempted before. It would be capable of calculating successive values of mathematical series using repeated addition, thanks to a cunning mathematical recipe called the Method of Differences. Accordingly, Babbage named his design the Difference Engine.
Having secured funding of £20,000 (about $33,000) from the British government, he contracted an engineer named Joseph Clement to build the machine. But a decade later, in 1833, fewer than half the parts had been completed, the funds had been exhausted, and Babbage and Clement had fallen out. In 1842 the government formally pulled the plug on the project when it refused to provide any more money on the advice of Sir George Airy, the astronomer royal, who denounced Babbage's plans as "worthless."
By this time, however, Babbage was working on an even more ambitious design he called the Analytical Engine. This was a direct precursor of today's computers: in particular, it had separate sections called the "store" and "mill," analogous to the memory and processor of a modern computer. But although he continued to refine the Analytical Engine's design until his death in 1871, Babbage was never able to raise the money to build anything more than a small experimental model representing part of the mill. He did, however, draw up plans for Difference Engine No. 2, an improved version incorporating new techniques he had devised while working on the Analytical Engine. But, in common with Babbage's other designs, it was never built in his lifetime.
WHY DID BABBAGE FAIL? Contrary to popular belief, it is not the case that Victorian engineering was unequal to the task of implementing his designs. Most of the components for Difference Engine No. 1 were melted down for scrap, but measurements of those that survive show them to be manufactured with impressive precision. And although problems could have arisen when trying to fit together components made by different manufacturers, Babbage's decision to use a single supplier meant that there was no technical reason why the Difference Engine could not have been successfully completed.
It would have worked, too, as was proven in 1991, when Doron Swade, the curator of computing at the Science Museum and his team successfully constructed the calculating apparatus of Difference Engine No. 2 using modern manufacturing techniques. As Babbage had anticipated, it was capable of cranking out the values of seventh-order polynomial equations with 31-figure accuracy.
But the Science Museum team had omitted the printer -- a vital part of the design which would have eliminated the possibility of human error when reading results from the calculating engine's 31 output wheels. Babbage had realized that the only way to ensure error-free tables was to automate the printing process as well, so he devised an elaborate printer mechanism capable of printing in two, three, or four columns, in a choice of two fonts, and with adjustable margins. After years of searching for a sponsor, Swade and his team announced in January 1998 that Myhrvold had agreed to fund the printer. Construction should be completed by the end of this year, at which point the printer will be plugged into the calculating engine's existing "printer interface" -- a complex set of gears and drive-shafts. The fact that one of the world's richest men (and, ironically, a man who got rich working in the industry that Babbage did his best to kick-start) had to step in to ensure the Difference Engine's completion tells us something significant: Babbage's machines cost a lot of money.
IN ORDER TO take off, successful inventions generally require the confluence of technology with an overriding social need. In Babbage's case, both were present. The problem was money. Not the lack of money required to build a prototype, but a more fundamental problem: mechanical computers are expensive. Suppose Babbage had been able to raise enough money to complete his original design and stage a successful demonstration. In the best case, he might then have been able to raise the money to build the far more ambitious (and expensive) Analytical Engine, a genuine mechanical computer, rather than a glorified automatic calculator. But even if Babbage had built a working Analytical Engine, it would have been equivalent to an costly early mainframe computer. Mass-market computing devices would not have followed. Why not?
Unfortunately, the economics of mechanical computing would have conspired against a 19th-century computer revolution. The motor of today's computer revolution is the constant improvement in price/performance, encapsulated in Moore's Law. This well-known rule-of-thumb states that, on average, processor power doubles every 18 months, and by extension, that computers halve in price for a given level of performance every 18 months. So one way to evaluate the likelihood that mechanical computers would have taken off is to compare the one-off cost of building a Difference Engine in Babbage's day to the one-off cost of building the same machine today, only with modern manufacturing techniques (which is essentially what the Science Museum team has done).
Swade estimates that Difference Engine No. 2 would have cost around £15,000 to build in Babbage's day. (This is based on the relative complexities of the two Difference Engine designs, and the known cost of building half the parts for Difference Engine No. 1). In today's money, that is equivalent to £1,500,000 -- or about $2.5 million. The cost of building the same machine today turns out to be $850,000, of which $500,000 is the cost of the calculating engine, and the rest is the cost of the printer. In other words, the cost of a Difference Engine has fallen by a mere 66% over 150 years -- a far cry from Moore's Law-style price drops of 50% every 18 months. It's analogous to the old saw about automobiles -- if their development accelerated according to Moore's law, they would all be able to do zero to 200 mph in five seconds and cost a quarter. The constant improvement implied by Moore's law is a direct consequence of the way semiconductors are made: the tiny components etched onto silicon wafers are the physical manifestations of intricate patterns of projected light. Chip-making has, in some senses, broken free from the constraints of the physical world, and mechanical manufacturing techniques cannot hope to compete. Although mechanical engineering techniques have improved since Babbage's day -- Myhrvold points out, for example, that the cost of clocks and watches fell dramatically during the 19th century -- it is hard to imagine how mechanical devices could show constant improvements in speed and reductions in size and cost that even come close to those made possible by integrated circuits in the last three decades.
But the lesson here is broader. Even if Babbage had succeeded, there still would have been no mechanical Moore's Law to drive a Victorian computer industry or the financial resources to support it. At best, there might have been a handful of computers for government and military use, but little else. Babbage's failure underlines a repeated theme of the history of technology, one often obscured in the rush of upgrades and improvements: for great inventions to succeed, they must be born into an age that will act as an accomplice.
Tom Standage is a science correspondent at The Economist, and author of The Victorian Internet: The Remarkable Story of the Telegraph and the Nineteenth Century's Online Pioneers (Walker).
©1998 FEED Magazine