The Idea Factory: Bell Labs and the Great Age of American Innovation (18 page)

Those who met Shannon at the time usually came away with an impression similar to Vannevar Bush’s: The young man was diffident, amiable, whip-smart. To those who knew him better, he could also be whimsical and prankish, with an insatiable love for games and gadgets. To those such as Norma who knew him better still he could be cold and remote, vanishing with some frequency into depressive sulks or abstraction. He had been exuberant when they first met—taking her up in the Piper Cub he was piloting, “scaring me shitless with his flying,” and enthusiastically dragging her out to jazz shows in the Village and watching with rapt attention while barely touching his drink. Now he would sometimes sit in a chair for hours, a pack of cigarettes for company, without writing or speaking. In other instances—when he got into a minor accident in the couple’s new Buick, a gift from Norma’s father—he would break down in tears.

Rarely would Shannon talk to anyone about the things he was working through in his mind. In early 1939, he had hinted to Vannevar Bush in a letter that he had begun thinking about communications and the methods by which “intelligence” moves from place to place. In a sense, he’d been thinking about these things for most of his young life. As a boy he’d had a job delivering newspapers, he’d also delivered telegrams for Western Union, and he’d even set up a private telegraph line to a friend’s
house using a wire strung along a fence. Some years later as a student at the University of Michigan, Shannon had read a paper by a Bell Laboratories engineer named Ralph Hartley entitled “Transmission of Information” that made an enormous impression on him. Hartley had proposed ways to measure and think about the rate and flow of information from sender to receiver. Perhaps there were deeper and more fundamental properties, Shannon now wondered, that were common to all the different kinds of media—telephony, radio, television, telegraphy included.
¹³
In his letter to Bush, he hadn’t gone far beyond what Hartley had put forward years earlier, but he hinted that he might, under the right circumstance and given some time, be able to work out some kind of overarching theory about messages and communications.

Nothing seemed to come of this during his summer at Bell Labs, and nothing seemed to come of it just after, as Shannon took a fellowship at the Institute for Advanced Study in Princeton, New Jersey, where Albert Einstein was in residence. “I poured tea for him,” Norma recalls of Einstein, “and he told me I was married to a brilliant, brilliant man.”

Shannon then returned to New York—accepting an offer to officially join the mathematics department at Bell Labs. “I could smell the war coming along,” he would later recall, “and it seemed to me I would be safer working full-time for the war effort, safer against the draft, which I didn’t exactly fancy…. I was trying to play the game, to the best of my ability. But not only that, I thought I’d probably contribute a hell of a lot more.”
¹⁴
Like the other mathematicians and engineers at Bell Labs, Shannon spent all of his daytime hours working on communications technologies for the war. He and Norma had separated just prior to his return to New York, an event he found shattering. She considered him too depressed and difficult. During their year at Princeton, Norma recounts, “I tried to get him to go to an analyst but he wouldn’t. He just got mad. He didn’t want to leave the apartment. He just got grimmer and grimmer. I began to feel that unless he could do something about it, I couldn’t stay with him.”

During the war years, Shannon’s colleagues at the Labs heard little, if anything, from him about communications, intelligence, and messages.
Now a bachelor again and living alone in a spare one-bedroom apartment on West 11th Street in Greenwich Village, Shannon would come home from work and labor over a personal project, writing equations and explanations in pencil, on lined sheets of paper, in a neat, compact script.
¹⁵
There was little or nothing in the icebox.
¹⁶
When he wanted to take a break he would put down his pencil, pick up his clarinet, and play.

T
HE MATH DEPARTMENT
at Bell Laboratories had grown up around a single man, Thornton Fry, the son of a poor Ohio carpenter who was working on a PhD at the University of Wisconsin when Harold Arnold came through town on a recruiting trip in 1916. It was still a few years before the creation of Bell Labs, and Arnold, Western Electric’s research chief, was looking for a young mathematician who could assist the engineers with the complex theory that often accompanied their switching and transmission plans. In a job interview, Arnold asked Fry a number of questions to test his knowledge about the era’s most influential communications engineers. Did the young man know the work of Heaviside, Campbell, or Molina?
¹⁷
Fry shook his head. He didn’t know a single one. Arnold must have nonetheless seen something encouraging in Fry. He gave him a job offer—$36 a week—and Fry immediately accepted.

Fry moved from Western Electric to the newly incorporated Bell Telephone Labs in 1925. Around 1930, he later recalled, the math group was expanded into a formal department staffed by several mathematics scholars.
¹⁸
At least as it was originally conceived, Fry’s department was not supposed to do research; it was meant to be a consulting organization to the engineers, physicists, and chemists at the Labs who needed help. “At that time engineers of all types were pathetically ignorant of mathematics,” Fry maintained, “so that anybody who could compute or quote a theory—even if he quoted it wrong—was admired by them. A mathematician was something like a nun; he was automatically admirable. He was different than other people.”
¹⁹

As the math department evolved and expanded, it became more than a mere consultancy. Many of its researchers became deeply involved in
the conceptualization of new circuits; others became interested in applying statistical tools to quality control issues—mostly at Western Electric, which churned out phones and equipment at its factories—as well as to congestion problems within the company’s vast switching system. More eccentric projects were sometimes deemed acceptable, too. George Stibitz, a mathematician under Fry whose work paralleled some of the ideas in Claude Shannon’s master’s thesis, decided to use telephone relays—those movable metal switches—in 1937 to build what appeared to be the world’s first digital computer. (Unlike Vannevar Bush’s differential analyzer, it did simple mathematical calculations using Boolean algebra and no gears or rods; the answer to a problem was displayed through light bulbs.) The job of a Bell Labs mathematician under Fry could thus vary tremendously. He could pursue his own interests without managerial interference. Or he (or she, since there were several women in the department now, too) could be asked to consider a question arising in another area, anything from a theoretical physicist’s dilemma to a problem related to a device that the Labs was readying for a Western Electric assembly line. As Henry Pollak, a former director of the math department, recalls, it “was full of people who had a major area where they worked, just like I did, too. But they couldn’t turn a good problem down. If one came by you dropped what you were doing and had fun with it. Our job was to stick our nose into everybody’s business.”
²⁰

By the early 1940s, when Fry asked Shannon to join his department, many Bell mathematicians were focused on the wartime problem known as fire control. Their challenge was to work out the complex mathematics behind the automatic firing of large guns that were trying to protect against enemy attacks—essentially, to create primitive computers that would gather information, largely through radar scans, on the location, speed, and trajectory of an incoming German rocket or plane. Their computers would immediately calculate the future position of the rocket or plane so that it could be intercepted and exploded in midair by shells or bullets. It took years to make this system workable, but it ultimately changed modern warfare. A defining moment came in 1944 during the defense of Great Britain against Hitler’s V-1 rockets, known as “buzz bombs.” Along the
English Channel, one Bell System historian noted, these gun directors intercepted 90 percent of the V-1 rockets aimed at London.
²¹

Shannon contributed some vital ideas to the fire control effort. Eventually, though, he drifted toward several committees that were working on secret methods of communication, which was far more to his liking. Since childhood, he had been as interested in games as in mathematics; in some respects he still saw little difference between the two. “I was a great fan of Edgar Allan Poe’s ‘The Gold Bug’ and stories like that,” he later recalled to an interviewer.
²²
The Poe story, written in 1843, revolves around the solution to a short cryptogram, whereby numbers and symbols—substitution ciphers, as cryptographers call them—are translated to individual letters. These letters in turn revealed secret coordinates near a huge tulip tree (“forty-one degrees and thirteen minutes northeast and by north”) and, ultimately, a pirate’s chest, once belonging to Captain Kidd, full of gold, diamonds, and rubies. Shannon often worked on similar cryptograms growing up, and his fascination with puzzles showed no perceptible ebb as he got older. At one point at Bell Labs, when Shannon read about a $10,000 crossword puzzle contest in the newspaper, he disappeared for a couple of weeks to the Labs library.
²³
Thus to recruit Shannon into the wartime work on codes and codebreaking, in other words, was to pay him to work on his favorite hobby.

Shannon summarized his war work on secret communications in a 114-page opus, “A Mathematical Theory of Cryptography,” which he finished in 1945. The paper was immediately deemed classified and too sensitive for publication, but those who read it found a long treatise exploring the histories and methodologies of various secrecy systems. Moreover, he had offered a persuasive analysis of which methods might be unbreakable (what he called “ideal”) and which cryptographic systems might be most practical if an unbreakable system were deemed too complex or unwieldy. His mathematical proofs presented the few people cleared to read it with a number of useful insights and an essential observation that language, especially the English language, was filled with redundancy and predictability. Indeed, he later calculated that English was about 75 to 80 percent redundant. This had ramifications for cryptography:
The less redundancy you have in a message, the harder it is to crack its code. And this also, by extension, had implications for how you might send a message more efficiently.
²⁴
Shannon would often demonstrate that the sender of a message could strafe its vowels and still make it intelligible. To illustrate Shannon’s point, David Kahn, a historian of cryptography who wrote extensively on Shannon, used the following example:

F C T S S T R N G R T H N F C T N

To transmit the message
fact is stranger than fiction
one could send fewer letters. You could, in other words,
compress
it without subtracting any of its content. Shannon suggested, moreover, that it wasn’t only individual letters or symbols that were sometimes redundant. Sometimes you could take entire words out of a sentence without altering its meaning.

The completion of the cryptography paper coincided with the end of the war. But Shannon’s personal project—the one he had been laboring on at home in the evenings—was largely worked out a year or two before that. Its subject was the general nature of communications. “There is this close connection,” Shannon later said of the link between sending an encoded message and an uncoded one. “They are very similar things, in one case trying to conceal information, and in the other case trying to transmit it.” In the secrecy paper, he referred briefly to something he called “information theory.” This was a bit of a coded message in itself, for he offered no indication of what this theory might say.

A
LL WRITTEN AND SPOKEN EXCHANGES
, to some degree, depend on code—the symbolic letters on the page, or the sounds of consonants and vowels that are transmitted (encoded) by our voices and received (decoded) by our ears and minds. With each passing decade, modern technology has tended to push everyday written and spoken exchanges ever deeper into the realm of ciphers, symbols, and electronically enhanced puzzles of representation. Spoken language has yielded to written language, printed on a press; written language, in time, has yielded to transmitted language,
sent over the air by radio waves or through a metal cable strung on poles. First came telegraph messages—which contained dots and dashes (or what might have just as well been the 1s and 0s of Boolean algebra) that were translated back into English upon reception. Then came phone calls, which were transformed during transmission—changing voices into electrical waves that represented sound pressure and then interleaving those waves in a cable or microwave transmission. At the receiving end, the interleaved messages were pulled apart—decoded, in a sense—by quartz filters and then relayed to the proper recipients.

In the mid-1940s Bell Labs began thinking about how to implement a new and more efficient method for carrying phone calls. PCM—short for pulse code modulation—was a theory that was not invented at the Labs but was perfected there, in part with Shannon’s help and that of his good friend Barney Oliver, an extraordinarily able Bell Labs engineer who would later go on to run the research labs at Hewlett-Packard. Oliver would eventually become one of the driving forces behind the invention of the personal calculator.
²⁵
Shannon and Oliver had become familiar with PCM during World War II, when Labs engineers helped create secret communication channels between the United States and Britain by using the technology. Phone signals moved via electrical waves. But PCM took these waves (or “waveforms,” as Bell engineers called them) and “sampled” them at various points as they moved up and down. The samples—8,000 per second—could then be translated into on/off pulses, or the equivalent of 1s and 0s. With PCM, instead of sending waves along phone channels, one could send information that described the numerical coordinates of the waves. In effect what was being sent was a code. Sophisticated machines at a receiving station could then translate these pulses describing the numerical coordinates back into electrical waves, which would in turn (at a telephone) become voices again without any significant loss of fidelity. The reasons for PCM, if not its methods, were straightforward. It was believed that transmission quality could be better preserved, especially over long distances that required sending signals through many repeater stations, by using a digital code rather than an analog wave.
²⁶
Indeed, PCM suggested that telephone
engineers could create a potentially indestructible format that could be periodically (and perfectly) regenerated as it moved over vast distances.