MOVING TO BOSTON HAD POSED SOME PROBLEMS FOR DUDLEY Buck. He secured an apartment at 277 Beacon Street, just a short walk across the Charles River to the MIT campus. Yet his pay as a research assistant was barely enough to get by, due to a new strain on his household finances. Buck had become a father, of sorts.
During his two years in Washington, DC, Buck had taken charge of a scout troop attached to the Industrial Home School, a place for orphaned or homeless children and juvenile delinquents. Having had an unconventional upbringing himself under Grandma Delia’s care, he could relate to their circumstances. Many of the boys had lost their father during the war.
Buck had pulled some strings to allow the boys to visit the US Naval Observatory. Scout Troop 31 even got a tour of FBI headquarters and posed for a photograph with director J. Edgar Hoover, alongside their beaming scoutmaster. He took them on hikes and camping trips, squeezing as many as fourteen boys into his 1937 Packard coupe—seven in the front, and seven in the rumble seat in the back. Often, he smuggled some science classes into their scouting curriculum too—once setting fire to a piece of parachute round the campfire while trying to demonstrate the power of hot air.
At the heart of the scout troop were three brothers, Glenn, Herb, and Bill Campbell, who lived at the school. The Campbell boys’ father had died four years earlier, in 1944, after crashing a crop-dusting airplane into a barn during the night. Their mother was in and out of St. Elizabeth Hospital with mental health conditions for which there was no treatment at the time. When not at St. Elizabeth she was a waitress at the Metropolitan Hotel. She also had two girls, but they had both been adopted out.
Glenn, the youngest of the brothers, latched onto Buck. The boy had been abused by a foster parent who regularly beat him with a studded leather belt or, if that wasn’t at hand, a coat hanger. Buck became a father figure to him.
Buck would occasionally let Glenn ride in his Navy truck as he went on low-level assignments around Washington. Glenn hung on every word Buck said. He even helped out Buck with a side business he had set up trading car parts. Glenn also developed a soft spot for Virginia, Buck’s sister. They started to become something of a little family, spending time together on weekends.
After hearing of his scoutmaster’s departure from Washington, Glenn, who was then thirteen years old, cried incessantly. Buck, for his part, fretted for the boy’s future. Although he was bright, Glenn Campbell had a fondness for getting into trouble. Buck proposed a solution: he would attempt to foster Glenn, and take him to Boston.
Obviously, this was problematic. It was hardly conventional for a single, twenty-three-year-old postgraduate student to foster a teenage boy. When it was first proposed, the Child Welfare Division in Washington would not hear of the idea. Foster children needed a two-parent home, the authorities said in a sternly worded letter.
Buck had a solution, however. His sister Virginia, who was about to join the US Foreign Service, also worried for Glenn. She agreed to become Glenn’s foster mother, at least initially. The Foreign Service delayed Virginia’s entry into her post for a few months to accommodate the plan.
To win over the child welfare authorities, Buck also sought some high-powered recommendation letters. Joe Eachus, the wartime codebreaking hero who had worked with him at Seesaw, wrote to the Child Welfare Division, providing a glowing character reference. A second reference came from Joe Keller, an attorney with the Washington law firm Dow, Lohnes & Albertson, who had gotten to know Buck and who had been his landlord for a short while.
Soon the fostering process was in motion. The paperwork could not be arranged before Buck was due to start work, so he went on to Boston alone. Two weeks later, a brief Western Union telegram was delivered to him at MIT: “Plan has been accepted for Glenn to go to Boston Love = Virginia.”
Buck drove back to Washington the following weekend, piled Virginia, Glenn, and all their worldly belongings into the back of the Packard coupe, and took them north.
Buck initially referred to Glenn as his foster son. After just a few days Glenn asked him to ditch the “foster” part and to just call him his son. Glenn was starting to show interest in music so Buck bought him a bright red piano with white trim, and hired a moving firm to take it in through an upstairs window of the apartment.
Virginia, meanwhile, found a job as secretary to Edwin Land, the inventor of the Polaroid camera; the Polaroid headquarters were close to the MIT campus in Cambridge. The unconventional little family seemed to work. Yet paying the rent for an apartment in central Boston proved tricky. Buck also worried that the Back Bay district—now an extremely affluent, gentrified part of the city—was a place where Glenn could fall in with the wrong crowd.
He decided to use his G.I. Bill—cash benefits received by all veterans at the time—to buy a house in the suburbs. Methodical as ever, Buck took out a map of the Boston area and a protractor. He drew circles, marking various distances from the MIT campus. He and Virginia then dedicated their spare time to house hunting. It soon became clear they would have to go farther out than they had planned.
In November 1950, the family moved into 9 Birchwood Road, Wilmington, Massachusetts—a newly finished two-bedroom, one-bathroom ranch-style house being sold by a developer who needed a quick sale. Buck negotiated a price of $7,900 after securing a small mortgage from the Medford Co-operative Bank. There wasn’t much left with which to furnish the place. The garden was a pile of mud with a couple of scruffy pine trees. The plaster on the walls was still drying, creating a convenient excuse for Dudley to avoid spending money on paint.
He bought two iron cots at the Army Navy surplus store in Cambridge—one for him, one for Glenn. A box-spring bed and mattress were bought for Virginia’s room. They rigged some curtains from old bedsheets for the bedrooms and bathroom, leaving the other rooms bare.
Virginia, who had put her own career on hold, was unimpressed. “Sis, just think of all the different routes we can take to Cambridge,” Buck told her in a vain attempt to enthuse her about the move. “Won’t that be interesting—and valuable quiet time.”
There was no cash left to hire someone to move Glenn’s piano back out of the window of the Beacon Street apartment. So Buck got out his tools and disassembled it, piece by piece.
“Dudley was very frugal,” recalls Glenn. “So Dudley took that piano completely apart and moved it to Wilmington in that ’37 Packard. Somehow he got it in that rumble seat. It was amazing that he got it back together. There are a lot of strings on a piano. Most of those middle notes had three strings per hammer. It never did get back to tune. It would get close, but never exactly back into tune.”
Life in Wilmington soon settled down. The Child Welfare Division in Washington was so impressed by the Buck siblings’ foster parenting that, by the following spring, they asked whether there was a possibility they could also take in Glenn’s sister Gretchen. Buck said he would need to be provided with a monthly allowance to support another child, and it went no further. Correspondence with Glenn’s caseworker details an exceedingly suburban life centered upon school, scouts and impeccable church attendance. Glenn soon had his own bike, and a dog.
In spite of this sudden immersion into an adult world of parenting, white picket fences, and PTA meetings—and the pressures that must have arisen from coercing his sister to succumb to the same lifestyle—Buck seems to have arrived on campus rigidly focused on his work.
Although his job was related to the input and output system for the Whirlwind device, he was soon assigned to working simultaneously on the core technology driving the machine—its memory systems. As if that was not enough to keep himself busy, Buck also remained loyal to his former commanding officers in the Navy—keeping them appraised of new technologies he was discovering in the MIT research labs, and passing along research papers as they were published.
A few days after Buck strolled into the lab at MIT for the first time, he asked for permission to relay information back to Washington. Jay Forrester had written a paper outlining some of the Whirlwind team’s work in the field of computer memory, Digital Information Storage in Three Dimensions Using Magnetic Cores.
Having had recent access to the leading military thought in this field, Buck could see that Forrester’s paper was breaking new ground. He asked if it would be possible for a copy to be sent to Joe Eachus at Seesaw. Although he was one of the most junior members of staff, Buck’s suggestion led to a copy of the paper being sent to Washington.
There was nothing unusual in this flow of information. In 1950, ideas moved fluidly across academia, the military, and a handful of emerging computer giants—such as IBM and Raytheon—who earned their keep mostly on the back of government contracts. An industrial cooperation agreement bound them all together with the government and ensured that the best technology was shared. The line that marked where the state ended and the private sector began was blurry at best. Buck was operating right in the middle of the gray area.
Buck kept close to his old connections. When he took a vacation, it was quite often to go to Washington. When Eachus came to Boston, he would take his old office junior out for lunch. Buck’s diary entries show that such appointments were relatively regular, even from his first arrival at MIT.
Eachus and his other wartime codebreaking friends were now running vast areas of the intelligence network, which was increasingly obsessed with building new computers. These war heroes were all mathematicians, however, who had been recruited initially for their ability to compute vast calculations by themselves. Buck brought a practical knowledge of electronics, physics, and chemistry. He could do things that they could not, and saw problems from a different perspective.
Eachus, in particular, had seen how Buck’s naïve optimism ensured that he never assumed that the existing way to do something was the right way. Throughout Buck’s career, his lab books were full of doodles mapping out often outlandish ideas that never got off the ground. Once, during his first year at MIT, he sketched out a theoretical design for a system that could manipulate ripples in Earth’s gravitational pull as a way to communicate. Gravitational waves—which had been theorized by Albert Einstein—were only proven to exist in 2016, earning the team of a thousand or so physicists the Nobel Prize in 2017. Buck never attempted to make the sketch come to life, but it gives an indication of how his mind worked.
His arrival at MIT coincided with a period in history where no idea was deemed too wild. As the quest for the ultimate computer gained pace within the intelligence community, it soon became clear that the central problem was to build better computer memories. Projects were set up across America, all working on slightly different ways to form a zero and a one that could translate into binary code, and thus be used to program a computer. Nobody had settled upon the best way to create these switches that were needed to form the fundamental building blocks of every operation that a computer could theoretically perform.
All computers work on the same basic principle. Any calculation, command, or action—no matter how complicated—can be stripped back to a simple iteration of steps that progresses from one to the next. That brutal logic—Boolean logic, as it is called—underpins the very existence of computers.
Once ideas and concepts have been broken down into their constituent parts, they can then be translated into binary code, a numeric system that allows the simplified instructions to be expressed as zeros and ones, positives and negatives, or yes or no answers.
Whether a computer is building a nuclear missile or sending an email, at its heart it is expressing everything in long chains of zeros and ones. Although computers are programmed using different languages and codes, at the base level everything is communicated in this same way. For the computer to work, those zeros and ones need to be created, stored, and read somewhere inside the machine. That’s what the computer’s memory is for.
In computing parlance, these zeros and ones are called bits—an abbreviation of binary digit. Eight bits make a byte. Today’s computer memories run to gigabytes and terabytes: a terabyte is one trillion bytes. Petabytes (a thousand terabytes) are even becoming part of common parlance as the “big data” revolution advances.
In the early 1950s, however, the best brains in computing science were still fretting over the quickest, easiest, and most reliable way to create and store a simple bit. The wartime machines at the British codebreaking center Bletchley Park had used electric valves, similar in principle to light bulbs, to produce the requisite ones and zeros. Their Colossus Mark 1, for example, had fifteen hundred of these valves, which could either be on or off. The ENIAC at the University of Pennsylvania, built shortly afterward, had 17,500 similar valves inside.
It was an imperfect system. Just like light bulbs, the valves got very hot and burned out relatively quickly. If just one valve blew, the computer went down.
The machines that Konrad Zuse made in his parents’ apartment in Berlin used relays—electromagnetic switches that had been popularized by the advance of the telephone—to perform the same job. The switch could only be in one of two positions: zero or one. Relays were smaller, and comparatively sturdy, but much slower—even relative to valves.
The transistor, created at Bell Telephone Laboratories in 1947, was lighter, cheaper, and used less power. Yet it wouldn’t be until 1954 that the device was produced on a commercial scale, and not until 1960 would an iteration emerge that could be lashed together into an integrated circuit, creating the basic microchip we know today.
Given the lack of consensus on the best way to create binary digits, and the large research budgets available, there was a willingness to listen to alternative suggestions for how best to form a zero or one.
Some researchers believed sound waves could be used to make a computer memory. The $500,000 Electronic Discrete Variable Automatic Computer (EDVAC), the second machine built by the University of Pennsylvania, and completed in 1949, trapped sound inside pools of liquid mercury as a way of storing data. At one end of the tiny pool of mercury there was a speaker, at the other a microphone. Pulses of electricity were converted to sound by the speaker which caused a vibration that traveled through the mercury. Above a certain frequency it represented a one, and below a zero. The sound would then be read at the other end of the mercury pool by the microphone. The process was repeated on a constant loop.
All the zeros and ones were stored in a perpetually cycling chain of sound waves. The problem was that if a particular piece of information was required, the machine had to wait until the next time the right segment of the wave passed through the microphone. Nonetheless, the EDVAC could add numbers in 864 microseconds.
At the opposite extreme of complexity there were punch cards. A one or zero could be determined by whether or not there was a hole in a piece of card. It was old technology—French textile mills had been using perforated paper tape to send instructions to their mechanical looms since about 1725. Charles Babbage, the inventor of the first mechanical computer, had also played around with them.
Punch cards were extremely reliable, and relatively cheap, yet they were also chronically slow. They would be used until the 1980s as a means of storing documents and processing large volumes of data. Yet they were far too slow to use as the memory for a computer’s brain, certainly not for a machine like Whirlwind, which had to respond instantaneously to the threat of nuclear-armed Soviet bombers.
Then there was magnetic tape. A number of early machines in the 1940s used reel-to-reel tape recorders as a form of memory. Like punch cards, it was a technique that survived for a long time as a means of electronic archiving, but it was also too slow.
In an effort to create a faster version of magnetic tape, the magnetic drum storage system emerged. Rather than have information spooled from reel to reel, the zeros and ones were stored magnetically on the inside of a large revolving drum, not dissimilar to a washing machine. Unlike the reel-to-reel tapes, the information did not have to be read sequentially—at least not in theory.
The laws of physics were still a limiting factor. To store a sufficient amount of information the drum had to be about thirty-four inches in diameter. It had to rotate at a speed of about 115 miles per hour, or 3,600 revolutions per minute, to access data at a fairly sluggish speed of 16,600 microseconds.
At the University of Manchester, the British center of excellence for computing immediately after World War II, they developed a much quicker memory system. The Williams-Kilburn tube, named after its creators Freddie Williams and Tom Kilburn, was based on a cathode ray tube, similar to the first televisions.
Rather than creating a television picture, the electron gun inside the tube generated rows of dots representing zeros and ones. The data in the electrically charged dots were then read by a metal plate that covered the end of the tube, detecting the tiny voltage generated from the screen. The system was used in the Manchester Mark 1 machine, which went on to be commercialized as the Ferranti Mark 1, and then in a handful of other pioneering machines on both sides of the Atlantic.
It was a version of this tube that was installed in the first Whirlwind machine, the computer that had dazzled with its performance on CBS television. Yet the team at MIT was not 100 percent satisfied with the Williams tube design, so they took it apart and created their own version. Rather than use a metal plate to read the output, they decided to use a second electron gun inside the tube to read what the first one had written.
Like most things to do with Whirlwind, the tube was complicated, expensive, and groundbreaking. The tubes were all built on-site; a glassmaking factory was set up on the MIT campus. Whirlwind remained imperfect, however. The new data tubes burned out almost as quickly as the old valves did on the wartime machines. And while Whirlwind was now quick, it was not quick enough. The specification that had been agreed upon with the navy and air force required data to be retrieved in six microseconds. The tube system could only do so in about thirty microseconds—still incredibly quick for its time, but five times slower than was needed.
The problem of how to find a better, quicker memory soon became one for Buck to tackle. After his productive spell working on the display systems for Whirlwind, he was reassigned to the team trying to increase the speed and reliability of this new machine by finding a better memory system.
For years Forrester had believed it would be possible to use electromagnets as a way of storing data. It had to be possible to use electricity to create a magnetic field and turn that into a means of storage, he posited. Variations on the concept had been kicking around since the mid-1940s, but it had never been developed properly. All it needed was an electric switch of some kind to move the polarity of a magnet from north to south. Each magnet could then represent a one or a zero.
As early as November 1949 Forrester had alerted his military sponsors to the possibility that magnets could be the future. Before he had even perfected the expensive, complex storage tubes he was developing, he was telling his sponsors about how he would create their replacement.
In a letter to Captain J. B. Pearson, deputy director of the Office of Naval Research, Forrester explained,
Recent advances in magnetic materials and the theory of solid state and electron physics make it apparent that new and improved high performance storage devices will become possible. Basic research in this field should be maintained at a modest but steady pace because one can expect that all forms of storage tubes now in use will become obsolete within a decade. Research in the first year of this program would be on certain magnetic storage and switching devices which show promise of becoming highly competitive with electrostatic tubes. Other new forms of storage could be added to the program of investigation later.
Forrester came up with the idea of running electric wire through tiny coils of magnetic wire arranged in a grid formation. Each magnetic coil would have two wires passing through it—one vertical, one horizontal. To switch the magnet from north to south would require both wires to be electrified—one wire alone would not have enough power to make the switch.
Each individual coil—or core, as Forrester preferred to call it—could be identified by the grid coordinates on the x and y axes of the mesh of wires, so each core could be easily found. He then added a third set of wires that ran diagonally across the grid, creating a third reference point.
It was this broad idea that was mapped out in the paper by Forrester that Buck asked to have sent to Eachus. Forrester had not worked out a way to make such a device, but the concept was there. Bill Papian, a graduate student in search of a thesis, was told to find the right materials and design to make this idea work. Forrester, meanwhile, went back to the giant administrative task of running the Whirlwind project.
Papian started to build a team around him. Buck was given the job of testing every type of material they could possibly find to make these miniscule magnets while also trying to improve the technology more generally.
Initially they tried to use a type of extremely fine magnetic tape called Deltamax. They would wrap it around a miniaturized ceramic bobbin, then feed the electric wire through the middle. The magnetic ribbon had to be wound by hand, so Buck invented a gadget that could be used to wind it more accurately. Eventually they were able to make a working memory cell that could hold sixteen bits of data, but it was a bit bulky.
After that they started dabbling with different types of materials. A specialist ceramics lab was built on the MIT campus just to make different magnetic compounds, all of which were tested by Buck. They baked doughnut-shaped ceramic magnets in electric ovens. Specialist ceramics manufacturers around the country were also contracted to make the tiny magnets.
A steady production line developed, producing more and more versions of the equipment. They were churning out two thousand miniature magnets a day. Each type of magnet was tested to see whether it could create a switch that flipped quicker than the last. They were strung together with different types of electrical wire in wooden grids about six inches square.
It got quicker and quicker—especially after Buck solved one of the key problems with Forrester’s design. Under the original plan, each time a zero or one was read it was destroyed. As a result, the machine was constantly repeating the same work, rewriting what had just been read. Buck designed a system that allowed the numbers to be read without destroying them, meaning that the computer now had more time to process instructions.
Although the new memory was being designed for use in Whirlwind, the lab team was barred from hooking its experiment up to the prized machine until it had been sufficiently tested. The team members were forced to build a separate computer just to trial the memory system, which they called the Memory Test Computer. As they got their heads around their new inventions, the machine kept getting faster and faster.
What they had invented was the first reliable random-access memory (RAM). With RAM, data could be retrieved more quickly than it could on tape or any of the other systems that existed before. It could be accessed ad hoc; there was no need to look through the whole tape, or wait until the sound wave passed the microphone again. The acronym RAM soon stuck, and is still used today as an indication of a computer’s processing speed.
On September 1, 1953, the first of these magnetic core memory units was installed in the Whirlwind machine. In a report to Eachus on the benefits of the new memory system, Buck explained that the Whirlwind machine now ran four times faster on a typical program. And it only broke down about once a week. It could perform twenty-five thousand multiplications sums per second. It could add numbers in 0.05 milliseconds—four times quicker than the wartime codebreaking machines or ENIAC at the University of Pennsylvania.
Thanks to Buck, Whirlwind was finally up to specification. Having started as a gigantic science experiment, it was now playing a part in protecting America from a Soviet nuclear threat. It occupied twenty-five hundred square feet of floor space total on two floors of the Barta Building at MIT.
Whirlwind only went out of service in 1959, when it was replaced by a fleet of machines built by IBM. All of those computers, and every other successful commercial computer at that time, used the same magnetic memory technology created by the Whirlwind team. The basics of the idea had been shared among different universities and corporations, courtesy of the industrial cooperation agreement on new computer technology that ran through Eachus at Seesaw.
Magnetic RAM would go on to become standard in all large computers of the 1960s and 1970s. It was so dependable and resilient to damage that magnetic RAM was still being used in the US space shuttle program as late as the 1980s, even though other technology had come along to supplant it. It all came down to the work done by Buck, Papian, and the rest of the team that had taken Forrester’s idea and made it work.
Buck never assumed that the solution they had devised would be the end point, however. While making and testing tiny ceramic magnets for Forrester, he was continuously dabbling with other means of creating zeros and ones. He had more than just one alternative idea as to how that could be achieved.