Amazing Grace Hopper

“This is the first time a woman has ever given a presentation in this room,” she was once told at the Pentagon as she was escorted to a meeting with the Secretary of the Navy. As one of the highest-ranking females ever to serve in the Navy, she was rare. But not as rare as something else she would later be: an octogenarian in the computer industry.

The industry has managed to leave nearly every former wunderkind behind in the sheer pace of its change. When Hopper entered the field in 1944, as she later recalled, “you could fit everyone who had ever heard the word computer into one small room.” They were all wunderkinds then. But Hopper was still one 35 years later, when she finally left the Navy.

In 1944, when she first saw the Mark I, Harvard’s Automatic Sequence Controlled Calculator, she was not merely impressed but startled. Even to someone who had always liked gadgets, the Mark I was preposterous, a 51-foot wall of clatter and light. Hopper wasn’t on hand to gape, though; she was there to communicate with the machine. It had its own one-of-a-kind language—grunts, practically—along with its own ways of absorbing information and analyzing it.

Very few people could hope to communicate with the computer in those earliest days, because few people could as yet adapt themselves to think in its peculiar patterns. The computer was born out of the mathematical study of finite differences—the reflection of the abstract in the real, through the use of formulas established in the purified realm of symbols. “The calculus of finite differences has become the bridge between mathematical analysis and numerical computation,” Grace Hopper wrote early in her career with computing machines. The patterns in the calculus were based on accepted formulas with logical progressions. To operate the Mark I, the builders needed to find people with similarly trained minds. They looked to the mathematics departments of the country’s colleges.

To develop the further potential of the computer, however, would require something more than mathematics, something even more than mere knowledge. It would require a thinking process that could retrain itself with each advance, remaining perfectly in step with the computer. And also one step ahead of it, which is where Grace Hopper stood for 42 years.

The Murray family had a good life in New York City before 1919. The parents were comfortable, handsome, and educated—popular without being social—and the three children, Grace, Mary, and Roger, enjoyed themselves both in the city and during summers at the family’s lakeside home in New Hampshire. Then, starting in 1919, the father, Walter Murray, had to undergo the amputation of first one of his legs and then the other because of a circulatory condition. The fact that the family had a good life even after that necessity was a deep influence on Grace, who was in her teens at the time.

After the operations Walter Murray was told that he had only a year or two to live. He returned to work as soon as he could, determined to make a good example for his children. (As it turned out, he lived to be 75.) Grace’s mother, Mary, took over all the responsibilities of managing the family’s affairs. Both parents placed an emphasis on self-sufficiency, but Walter was especially adamant, telling the children to study hard and to make sure that they could always be independent if necessary.

The children attended private schools that offered well-rounded activities along with a demanding curriculum. Grace played on sports teams, acted in plays, and helped produce yearbooks. Although she was uncommonly bright in math courses and very good in other subjects, she failed Latin as a senior in high school. The trouble was that she liked shortcuts, and when it came to Latin, the easiest shortcut was always simply to speak English. Even so, it is not ironic but downright fitting that she was to become the foremost expert in new languages (computer languages) at the middle of the century. She would merely be looking for shortcuts then, too.

At Vassar College in the class of 1928, Grace Murray majored in mathematics and physics and then won a fellowship that paid for continuing study, which she took at Yale. Her brother would later graduate from Yale; her sister had followed her to Vassar. In 1930 Grace married a man named Vincent Hopper, a Princeton literature major who taught at New York University. She joined the faculty of the mathematics department at Vassar after receiving her master’s degree; the following year she resumed graduate work at Yale, and she received a doctorate there in 1934. In Grace Hopper: Navy Admiral and Computer Pioneer , a biography no less enlightening because it is supposed to be a young adults’ book, Charlene W. Billings notes that of the fewer than ten candidates for doctoral degrees in math at Yale in the early 1930s, four, including Grace Hopper, were women. She was part of a generation in which it was not unusual for young women in America to be taken seriously as intellects—or for them to take themselves seriously as such.

In 1938 Vincent received his doctorate from Columbia, and the next year the Hoppers built a house near the Vassar campus. Vincent commuted weekly to his teaching job in New York City. However, by 1942 they had separated, and they were divorced soon after. As a professor at a women’s college in a small town, Hopper could have easily passed through World War II without great inconvenience. In fact, teaching math at the college level was classified as a vital profession during the war, so men and women holding such positions were not only exempted but barred from entering the military. Hopper wanted to serve, though, and she doggedly talked herself past the facts that she was too thin and doing a job considered indispensable. The person who graduated first in Grace Hopper’s midshipman’s class was Grace Hopper.

Her choice of the Navy may have arisen in part from family pride in her great-grandfather Alexander Wilson Russell, who had been a rear admiral. A more pressing consideration was that the Navy needed mathematicians, even more than did the Army, Army Air Corps, or Marines. Immediately after World War I, Navy ships had begun to depend on long-range radio communication, a field of research that led to radar and later sonar technology. Even standard radio transmissions generated a need for high-speed computing during World War II, because they had to be encrypted to defeat enemy eavesdropping. From the start the British waged a war within the war against Germany, as veritable fortresses of cryptography were built and then breached, time and again. The British may have won the war—indeed both wars—when they developed the first electronic computing machine, called Colossus, to break the German code in 1943. They immediately shared the plans, and about a half-dozen of the machines were in use in America during the last years of the war. However, since they were supersecret, they didn’t influence other American efforts to build a computing machine.

Six American teams were trying to develop a computing machine for military use by 1942. Detailed calculations were as important as gunpowder to some of the Navy’s most promising new ordnance, including long-range missiles. Even water mines had become “smart” weapons, set magnetically to sense a passing ship. The good news was that the Navy’s new ordnance could be devastating; the bad news was that a single calculation took hours to complete, and tens of thousands of calculations were required to support a whole array of new weapons.

Among the teams attempting to create a computing machine, Harvard University had made the strongest start. It might be said that Harvard had also made the earliest start, with a design completed in 1937, but then all the emerging computing machines actually had their beginning in the 1830s, when the Englishman Charles Babbage applied theories of calculus to the operation of the mechanical switches that directed looms in factories. Babbage was a visionary, which is to say that his machine didn’t work, but would someday. Harvard’s new machine would have two things that Babbage’s could not, in addition to the benefit of advances in mathematics. First, it had a refined five-horsepower motor. Second, it had IBM.

International Business Machines made punchcard machines that allowed people to control machinery, to sort and find information, or to make simple calculations (addition, subtraction, division, and multiplication), all by using cards. Its machines advanced on the technology of the nineteenth century by using punch cards in a system of electrically connected switches, which could control not only other machines but also the data itself. Punch cards, once as much the fodder of business life as ticker tape and typing ribbons, were pieces of stiff paper, 3¼ by 7 3/8 inches, into which small rectangular holes could be cut. Where a card allowed electrical current to flow through, it opened certain mechanical switches, which in turn controlled factory machinery, or sorting routes, or tabulating machines. Or other functions entirely; invented near the end of the nineteenth century (on designs developed to operate looms at the beginning of that century), punch-card machines had been applied to myriad business chores by the late 1930s. Even so, at its most clever an IBM tabulator was still pretty dumb and could only add, subtract, divide, or multiply. That, however, would be just about enough.

Howard Aiken, a physics graduate student at Harvard, knew through the work of earlier theorists that numerical analysis could be broken down into five simple steps, no matter how complicated the ultimate calculation: addition, subtraction, multiplication, division, and reference to standard computations (i.e., on charts). Aiken connected IBM punch-card calculators to make a whole that was much, much more than the sum of its parts. In his 1937 proposal he said that the resulting machine, eventually to be called the Mark I, would be “a switchboard on which are mounted various pieces of calculating machine apparatus. Each panel of the switchboard is given over to definite mathematical operations.”

As a favor to Harvard and to the Navy, IBM completed the details of Aiken’s design and finished building the Mark I in 1943. Thomas Watson, president of the corporation, had no particular use for the machine and none for its feisty inventor. Donating the Mark I to Harvard’s newly formed Computation Laboratory, the company cooperated with other wartime computer projects but remained generally aloof from the computer field until the early 1950s.

Howard Aiken had his hot rod—not a particularly streamlined model, but lightning fast nonetheless. He also had a commission in the Naval Reserve and pressing commitments to supply the Navy’s Bureau of Ordnance with calculations. To coax them out of the new Mark I, he had two mathematicians helping him in the summer of 1944, both ensigns, Richard Bloch and Robert Campbell. When they heard that a schoolteacher from New York State was arriving to help, they got into a mock fight over who would have to sit next to her.

Aiken, however, was impatient, a commander in more than just rank. “Where the hell have you been?” he said as a greeting when Hopper arrived at the lab on a Monday in July. She started to apologize for taking the weekend off, but he was referring to the months she had spent in midshipman’s school. Before they straightened that out, he gestured to the Mark I, which stood 8 feet tall and 51 feet long, and told her to “compute the coefficients of the arc tangent series by Thursday.” Aiken then left her standing with the machine. Hopper liked working for Aiken, though. He was flinty but fair, and he was consistent: gently helpful the first time a mistake was made and utterly ferocious if the same person made the same mistake again.

Campbell and Bloch worked closely with Hopper, and as a team the three learned to program the Mark I. The two ensigns later admitted that they had been taken aback on meeting her to see that she was not the schoolmarm they’d imagined but vivacious and youthful. Out of uniform she dressed in lively colors. She was known for liking to laugh and make fun, and in fact she is best known today for one of her jokes. A year after beginning work with the Mark I, she was crawling around inside, looking for the source of a malfunction, when she found that a moth was fouling one of the relays. She emerged with it triumphantly and announced to Bloch, “Here it is, the actual bug in the works!” She taped it in the daily logbook and has since been credited with coining the word bug for a malfunction. The expression had been around before, but she was the first person ever to see a computer bug in the flesh.

The Mark I, like every other computer since, slipped empirical data through formulaic screens—programs—with split-second timing. Hopper and her colleagues were assigned to supply a separate program for each type of calculation needed. To write a code that would accomplish a prolonged calculation, they had to think in terms of the overall formula, its construction in simple steps, and the shuttling of empirical data through the units so that it would coincide with those steps.

The actual language they used to communicate the steps, or commands, was made only of numbers. A paper tape 3¼ inches wide punched with holes ran a program through the Mark I. Each line on the tape had space for 24 holes, which for programming purposes were considered in groups of 8 holes each. The first and second group instigated the transfer of two numbers out of calculating units. The third group described what mathematical function (of the five) was to be performed on the two numbers. Depending on the exact placement of the holes, one line could express any of more than 16 million directives. A line of code might read, for example, “24 … 167… 3467.” They were just numbers, but they bridged the gap between simple electrical impulses and sophisticated abstractions of human thought.

The manual for the operation of the Mark I, which Hopper largely wrote, was published as Volume I of the Annals of the Computation Laboratory of Harvard University , and it ran to more than 400 pages. By the time it came out, however, the most advanced computer in this country was the all-electronic ENIAC, designed by J. Presper Eckert and John W. MauchIy in Philadelphia. Harvard was no longer leading the way.

At the end of the war, Grace Hopper was confronted with an important choice. She would have to decide whether to stay at Harvard, but she knew she wanted to remain in the computer field. Placed on inactive status as a naval officer, though remaining in the reserve, she was drawn away from Boston in 1949 to a position as a senior mathematician with the Eckert-Mauchly Corporation, the only company in the country entirely dedicated to the future of computers. That suited her; to her dying day in 1992, she believed that computers had still only begun to reach their potential. But in the late 1940s the accepted wisdom in business circles was that they had done all they could—as implements of military engineering. In the first place, costing as they did close to a million dollars apiece, they were too expensive to use in business. In the second place, they were too difficult to program for specific purposes. Hopper would address that at Eckert-Mauchly.

Any good engineer is innately lazy, as a frame of mind, not as a work ethic. While working out endless ribbons of numbers for Mark I programs, Hopper had often daydreamed of sitting back somehow and making the machine program itself. It liked drudgery, after all, and was good at churning out accurate numbers. The Mark I couldn’t possibly program itself, but at Eckert-Mauchly she would have a machine that could. Its UNIVAC was a true computer, not a computing machine like the Mark I and most of its other predecessors. It could store data, on magnetic tape or electronically, and it could learn. “UNIVAC at present has a well grounded mathematical education fully equivalent to that of a college sophomore,” Hopper said in 1952. “And it does not forget and does not make mistakes,” she added.

As a first step toward teaching the UNIVAC to help create its own programs, she stored pieces of reusable coding in the computer and identified each by a number; she could then direct the machine to locate a whole piece, as needed. That was a shorthand in programming, but it did not accomplish the goal of making the machine program itself. To do that, Hopper created a new type of master program that taught the computer to compile working programs from subroutines, in such a way that each subroutine could be translated into usable form and linked in a proper sequence. She was inventing the compiler, a part of every computer today. The theory behind the breakthrough came from her days at Vassar—but not from any classroom.

On the basketball team the best way to move the ball was to dribble as far as possible (and under the rules then in force for women, players were limited to one dribble), pass the ball forward, run past the player with the ball, and take a forward pass in turn. The concept leaped forward itself as the first computer compiler in 1952. “I tucked a little section down at the end of the memory which I called the ‘neutral corner,’” she told Billings in 1982. “At the time I wanted to jump forward from the routine I was working on, I jumped to a spot in the ‘neutral corner.’ I then set up a flag for an operation which said, ’I’ve got a message for you.’ This meant that each routine, as I processed it, had to look and see if it had a flag; if it did, it put a second jump from the neutral corner to the beginning of the routine, and it was possible to make a single-pass compiler—and the concept did come from playing basketball!” The compiler helped the UNIVAC break through as the first computer embraced by business. So did sales support from Remington Rand, which had purchased the top-heavy Eckert-Mauchly Corporation in 1950.

For a time in the 1950s, UNIVAC was the household word for “computer,” the way Kleenex was for “tissue.” Forty-six of the first generation of UNIVAC computers were delivered to companies in aerospace, insurance, publishing, and finance between 1951 and 1955. The wide variety of those first applications—from charting subscriptions to tracking missiles in flight—started the computer revolution. The Mark I and other military-use computers could have gone back on a shelf. The UNIVAC could not. It was soon everywhere, and Hopper’s job was to make it fit wherever it went. “It is the current aim to replace, as far as possible, the human brain by an electronic digital computer,” she said in 1952.

By 1954 a dozen major companies were manufacturing computers, and by the mid-1950s computers were mimicking humans in at least one respect. They were all speaking different languages and standing farther and farther apart because of it. At Remington Rand, Hopper was jumping ahead to what she considered the inevitable: Computers had to understand English. If they were to travel far and wide—and she was already saying publicly that they would someday be the size of shoeboxes and go everywhere—their programmers could not always be expected to have advanced training in math. Even while predicting that software (a word not yet in use) would soon be more important than the computer itself, Hopper started work on a programming language that would be user-friendly (an expression definitely not yet in use).

The schoolgirl who had had to repeat a year of Latin had grown up to join a hell-bent chase in the mid-1950s to make a new language. She used English as the basis of the language, and it even had a mellifluous name, unlike the acronyms that already littered the industry. It was called Flow-matic.

Flow-matic was a natural outgrowth of Hopper’s work with compilers. She had already been using combinations of letters, such as tsO and bOi , to tell the computer to perform specific operations. She soon realized that the combinations might as well be A-D-D or R-E-P-E-A-T . As a piece of engineering, the language was much more complex than that, but the result was simple: With Flow-matic, communication between the programmer and the computer became more direct.

The programmer still had to break down operations into tasks that the computer understood, but he or she no longer had to break down those tasks into either their mathematical components (the building blocks of human computation) or electric impulses (the building blocks of machine computation). That was the difference between where Grace Hopper had started in 1944, with the Mark I, and where she stood in 1957, with the full introduction of Flow-matic.

When Grace Hopper first demonstrated a rough version of Flow-matic in 1953, Remington Rand was the leading computer company in the world. The executive committee of the corporation watched the demonstration, which was successful, and then informed her that an English-based programming language was an impossibility. The decision represented a terrific frustration for her because while Remington Rand was stifling Flow-matic, other companies were galloping ahead with the development of their own new programming languages.

Previously, Hopper’s success had been based on a fairly straight-line combination of dedicated work plus quite a lot of brilliance plus trust in the boss. However, Remington Rand let her down from 1953 to 1956 by discouraging Flow-matic, and it also let its shareholders down. A three-year jump on an English-based programming language might well have put Remington Rand where IBM was, when IBM was the business computing industry. Flow-matic was well received when it was finally introduced in 1957, but by then it was just another language in the cacophony.

In 1959 the market for business computers was burgeoning, but it was obviously headed for trouble without quick action. The competing companies reluctantly faced the fact that one programming language was going to have to prevail. The compromise was COBOL, an English-based business-programming language introduced as an industry standard in 1960. Grace Hopper didn’t write it, but her name has been very prominently associated with it ever since; it was fundamentally based on Flow-matic.

While Hopper was continuing her career at Remington Rand in the 1960s, the Navy was foundering in the very computer age that it had helped initiate 20 years before. Hopper still loved the Navy. She wore her uniform on the anniversaries of D-day and Pearl Harbor and gave generously to her favorite charity, the Navy Relief Society. In 1966, however, at the age of 60, she was asked to retire from the Naval Reserve. She said it was the saddest dav of her life.

Then, less than a year later, the Navy realized that it didn’t need less of Hopper, it needed more. She was not only recommissioned but placed on active duty. She was too old, but the Navy made an allowance. She took a leave of absence from Remington Rand and moved to Washington to work at the Pentagon. Her job was to standardize programming for the service’s various operations; to make their computers as compatible as possible; and, hardest of all, to persuade those in the ranks to accept computers enthusiastically in their work. When the job developed into a second (or third or fourth) career by 1971, Hopper retired from what had become Sperry Rand (and is now Unisys) on excellent terms.

As a place of employment the Navy could be at least as treacherous as any corporation, but it treated Hopper just the way she wanted to be treated. She was given both respect, as shown in her regular promotions, and the chance to do new things. While her understanding of computers never stopped advancing, she didn’t consider that her duties as an officer began or ended with them. As she rose to higher ranks, she worked hard at honing leadership traits, both her own and those of junior officers serving with her. She insisted that the members of her staff learn to speak clearly, and partly in jest she instituted a 25-cent fine for every “you know” that she heard slip into a sentence. Another expression, however, riled her to real consternation and sometimes to fury. It was: “But we’ve always done it that way.”

“A ship in port is safe, but that is not what ships are for” was the sentiment Grace Hopper had been raised on. She made it her personal motto. She retired at the age of 79, but only from the Navy. She then joined the Digital Equipment Corporation as a senior consultant and was still on the job when she died at the age of 85. Grace Murray Hopper was buried at Arlington National Cemetery on January 7, 1992, with full Navy honors.