Fortune's Formula (4 page)

In October 1939 Shannon met a Radcliffe coed, Norma Levor, at an MIT party. Levor remembers Shannon as “a very cute guy” standing in a doorway, strangely aloof. She got his attention by throwing popcorn at him. They spoke and were soon dating. Norma was nineteen years old and beautiful, the daughter of a wealthy, highly assimilated Jewish family in New York. Radcliffe girls were not then allowed to bring boys into their rooms. Norma and Claude’s unlikely trysting spot was the Differential Analyzer room. On January 10, 1940, Claude and Norma were married by a justice of the peace in Boston. They drove to New Hampshire for a honeymoon. When Shannon went to register at the hotel, he was told: “You people wouldn’t be happy here.” Claude had “Christ-like” features, recalled Norma, which must have convinced the innkeeper he was Jewish. They drove elsewhere.

In March, Shannon wrote Bush and belatedly informed him of the marriage. He said they had moved into a house in Cambridge, and his life had been unsettled. The same letter describes a new idea Shannon was working on: a better way of designing lenses. “Do you think it would be worthwhile to attempt to work this out?” Shannon asked Bush. He mentioned that Thornton Fry of Bell Labs had offered him a job. “I am not at all sure that sort of work would appeal to me,” Shannon wrote, “for there is bound to be some restraint in an industrial organization as to type of research pursued.”

AT&T was moving most of its research from Manhattan to an expanded suburban outpost in Murray Hill, New Jersey. Shannon spent the summer working at Bell Labs’ Greenwich Village site. Norma remembers this as the happiest part of their brief marriage. She and Claude frequented the jazz clubs. Their next move was to the Institute for Advanced Study at Princeton. This was the home of Einstein, Gödel, and von Neumann. Shannon began what was to be a year of postdoctoral work under mathematician and physicist Hermann Weyl. He worked on topology.

Nothing came of it. Shannon left abruptly to work with mathematician Warren Weaver of the U.S. Office of Scientific Research and Development. Shannon helped calculate gunfire trajectories for the military. Weaver praised his work; then this too was cut short. Shannon’s marriage was breaking up.

Norma saw a disturbing change in Claude when they moved to Princeton. His shyness deepened into an almost pathological reclusiveness. The institute’s scholars are allowed to set their own hours and to work where they like. Shannon chose to work at home. “He got so he didn’t want to see anyone anymore,” said Norma. She tried to convince Claude to seek psychiatric help. He refused. During one violent argument, Norma ran all the way to Princeton Junction and took the train into Manhattan. She never returned to Claude or to Princeton.

Claude was devastated. Weaver wrote Bush that “for a time it looked as though he might completely crack up nervously and emotionally.”

In the midst of Shannon’s personal crisis, Thornton Fry renewed his offer of a job at Bell Labs. This time Shannon accepted. And once again, Shannon turned his polymorphic genius to something completely different.

I
T WAS CALLED
P
ROJECT
X.
Declassified only in 1976, it was a joint effort of Bell Labs and Britain’s Government Code and Cipher School at Bletchley Park, north of London. It had a scientific pedigree rivaling that of the Manhattan Project, for the British-American team included not only Shannon but also Alan Turing. They were building a system known as SIGSALY. That was not an acronym, just a random string of letters to confuse the Germans, should they learn of it.

SIGSALY was the first digitally scrambled, wireless phone. Each SIGSALY terminal was a room-sized, 55-ton computer with an isolation booth for the user and an air-conditioning system to prevent its banks of vacuum tubes from melting down. It was a way for Allied leaders to talk openly, confident that the enemy could not eavesdrop. The Allies built one SIGSALY at the Pentagon for Roosevelt and another in the basement of Selfridges department store for Churchill. Others were established for Field Marshal Montgomery in North Africa and General MacArthur in Guam.

SIGSALY used the only cryptographic system that is known to be uncrackable, the “onetime pad.” In a onetime pad, the “key” used for scrambling and decoding a message is random. Traditionally, this key consisted of a block of random letters or numbers on a pad of paper. The encoded message therefore is random and contains none of the telltale patterns by which cryptograms can be deciphered. The problem with the onetime pad is that the key must be delivered by courier to everyone using the system, a challenge in wartime.

SIGSALY encoded voice rather than a written message. Its key was a vinyl LP record of random “white noise.” “Adding” this noise to Roosevelt’s voice produced an indecipherable hiss. The only way to recover Roosevelt’s words was to “subtract” the same key noise from an identical vinyl record. After pressing the exact number of key records needed, the master was destroyed and the LPs distributed by trusted couriers to the SIGSALY terminals. It was vitally important that the SIGSALY phonographs play at precisely the same speed and in sync. Were one phonograph slightly off, the output was abruptly replaced by noise.

Alan Turing cracked the German “Enigma” cipher, allowing the Allies to eavesdrop on the German command’s messages. The point of SIGSALY was to ensure that the Germans couldn’t do the same. Part of Shannon’s job was to prove that the system was indeed impossible for anyone lacking a key to crack. Without that mathematical assurance, the Allied commanders could not have spoken freely. SIGSALY put several other of Shannon’s ideas into practice for the first time, among them some relating to pulse code modulation. AT&T patented and commercialized many of Shannon’s ideas in the postwar years.

Shannon later said that thinking about how to conceal messages with random noise motivated some of the insights of information theory. “A secrecy system is almost identical with a noisy communications system,” he claimed. The two lines of inquiry “were so close together you couldn’t separate them.”

In 1943 Alan Turing visited Bell Labs’ New York offices. Turing and Shannon spoke daily in the lab cafeteria. Shannon informed Turing that he was working on a way of measuring information. He used a unit called the
bit
. Shannon credited that name to another Bell Labs mathematician, John Tukey.

Tukey’s bit was short for “binary digit.” Shannon put a subtly different spin on the idea. The bit, as Shannon defined it, was the amount of information needed to distinguish between two equally likely outcomes.

Turing told Shannon that
he
had come up with an idea for a unit called the
ban
. This was the amount of evidence that made a guess ten times more likely to be true. The British cryptographers used that term, half seriously, in decrypting Enigma ciphers. The “ban” part came from Banbury, the town where the cryptographic team’s scratch paper was manufactured.

It was the bit, not the ban, that changed the world. The defining year for that change was 1948. Shannon remained with Bell Labs after the war. One day he spotted a strange object on the desk of another researcher and asked what it was.

“It’s a solid-state amplifier,” William Shockley told him. It was the first transistor. Shockley told Shannon that the amplifier could do anything a vacuum tube could.

It was small. Shannon learned that the new device worked by having different materials in contact with each other. It could be made as small as desired as long as the different materials touched.

The transistor was the hardware that would make so many applications of Shannon’s theory a reality. This incident would have been in late 1947 or early 1948, before Bell Labs unveiled the transistor on June 30—and just about the time Shannon’s classic paper on information theory appeared.

There is minor scandal associated with that paper. Shannon published “A Mathematical Theory of Communication” in a 1948 issue of the
Bell System Technical Journal
. He was then thirty-two years old. Most of the work had been done years earlier, from about 1939 to 1943. Shannon told few people what he was doing. He habitually worked with his office door closed.

As Bell Labs people gradually learned of this work, they were astonished that Shannon had devised such an important result and then sat on it. In what amounted to a scientific intervention, friends goaded Shannon to publish the theory. Shannon recalled the process of writing the 1948 paper as painful. He insisted that he had developed the theory out of pure curiosity, rather than a desire to advance technology or his career.

The year 1948 was also a turning point in Shannon’s personal life. Shannon would often go into the office of John Pierce to chat. Pierce was working on radar and was known as an avid science fiction fan. Through these visits, Shannon met Pierce’s assistant, Mary Elizabeth Moore. “Betty” Moore had been in the math group’s computing pool, performing calculations on old-fashioned desktop calculating machines. Moore was bright and had a Rosie-the-Riveter knack for making things. She was able to work a drill press and lathe in the lab’s machine shop. She was attractive and one of only three female employees there. (“One was married, and the other was in her fifties,” Betty recalls.) She and Claude had their first date in December 1948. On March 27 the following year they married.

Shannon began teaching at MIT in the spring 1956 semester. This started as a temporary assignment, and at least one friend at Bell Labs (John Riordan) understood the teaching as having an ulterior motive. It was supposed to allow Shannon the free time to begin writing a long-anticipated book on information theory.

“I am having a very enjoyable time here at M.I.T.,” Shannon wrote his Bell Labs boss, Hendrik Bode. “The seminar is going very well but involves a good deal of work. I had at first hoped to have a rather cozy little group of about eight or ten advanced students, but the first day, forty people showed up, including faculty members from M.I.T., some from Harvard…”

After just a few months at MIT Shannon wrote Bode to resign from his post at Bell Labs. He was taking a professorship at MIT. He found that he and Betty liked the intellectual and cultural life of Cambridge, so worldly next to the New Jersey suburbs. “Foreign visitors often spend a day at Bell Laboratories but spend six months at M.I.T.,” Shannon explained to Bode. “This gives opportunities for real interchange of ideas. When all the advantages and disadvantages are added up, it seems to me that Bell Labs and academic life are roughly on a par, but having spent fifteen years at Bell Labs I felt myself getting a little stale and unproductive and a change of scene and of colleagues is very stimulating.”

Shannon had approached MIT about a permanent job, not the other way around. Money was not the issue. Bell Labs offered a “flattering” raise. Shannon turned it down (he retained an affiliation with Bell Labs through 1972). His initial salary at MIT was $17,000 a year.

Shannon enjoyed the stimulation of MIT in limited doses. He did his best work alone. He had perhaps underestimated the volume of distraction confronting a living legend at a large urban university. Shannon “started disappearing from the scene,” recalled Robert Fano. “He kind of faded away, Claude.”

Shannon took few Ph.D. students. They often had to meet him at his home in order to get advice. One student, William Sutherland, remembers walking in on Shannon’s oboe practice more than once. “He slept when he felt like sleeping,” said Betty, and would spend hours at the kitchen table thinking.

Shannon’s career as publishing scientist was just about over. He never completed the book he spoke of. Shannon’s papers at the Library of Congress include nothing more than a few handwritten notes that may have related to this project.

Artificial intelligence pioneer Marvin Minsky speculated that Shannon stopped working on information because he felt he had proven almost everything worth proving. The self-contained perfection of Shannon’s early work was unsurpassable. Fano mentioned an uncanny phenomenon. With rare exceptions, it seemed that whenever an information theorist mentioned a current problem to Shannon, (a) Shannon was aware of the problem, and (b) Shannon had already solved it, but hadn’t gotten around to publishing it.

“I just developed different interests,” Shannon said of his near-abandonment of the field he created. “As life goes on, you change your direction.”

One of these interests was artificial intelligence. Shannon organized the first major academic conference on the subject, held at Dartmouth in 1956. Shannon’s stature contributed to making the field respectable. Some of the devices Shannon built, including an early chess-playing computer and the so-called outguessing machine, figure prominently in the early history of machine learning. Shannon was an articulate advocate, visionary enough to see what fantastic things were possible and practical enough to appreciate that they were not going to happen in his lifetime. He had a talent for parrying the inevitable ham-handed questions.

Q. Will robots be complex enough to be friends of people, do you think?


A. I think so. But it’s quite a distance away.


Q. Can you imagine a robot President of the United States?


A. Could be. I think by then you wouldn’t speak of the United States any more. It would be a totally different organization.

Letters, papers, and phone calls, many from world-renowned scientists, poured into Shannon’s office. They wanted Shannon to review a paper or contribute one; give a talk, an opinion, a recommendation. Shannon turned down an increasing share of these requests. As Shannon’s name became known to a broad public, he began receiving letters from schoolchildren building science projects and crackpots building paranoid complexes about scientists, computers, or the phone company (“Dear Sir,” begins one letter, “Your mechanical robot Bel, the idol [Daniel 14] in the Bible, is a mechanical monstrosity…You are making a traitor out of the President of the U.S. and the F.B.I. by letting your robot deceive you. I have threatened to sue the N.Y. Telephone Co. of N.Y. City, and I will, if you don’t wake up”).

From time to time the CIA and other agencies turned to Shannon when challenging cryptographic problems arose, only to be informed politely of Shannon’s retirement. “We really are not approaching you accidentally,” read a 1983 letter from the CIA’s Philip H. McCallum. “We need an excellent original thinker, and at the risk of kowtowing, find that you are still the best for what we have in mind…Although we understand that you do not need the money, we would still pay you a fee.”

Shannon did not like to answer a letter until he had composed the perfect reply. Since it took a while to create a perfect reply, Shannon dealt with correspondence by shuffling it from folder to folder. On these folders he would write labels like “Letters I’ve procrastinated on answering for too long.” These letters are now neatly boxed with Shannon’s papers at the Library of Congress, many still awaiting answers.

Shannon was yet in his forties when he took what amounted to an early, unofficial retirement. Thereafter Shannon was MIT’s Bartleby, whose characteristic reply was “I would prefer not to”—clerk of his own private dead-letter office.