Computing with Quantum Cats (6 page)

The re-assembled Colossus broke its first message on 5 February 1944. It was ten times faster than Heath Robinson, and, equally important, more reliable. Orders for more Robinsons were canceled, and Flowers was asked how quickly Dollis Hill could produce more “Colossi.”

One of the Wrens
¹³
who worked on Colossus, Betty Houghton (née Bowden), now lives in a neighboring village to us. She was fourteen when the war broke out, and three years later joined the WRNS. She was told that there were two kinds of jobs available—cook/steward or “P5.” Having no wish to be a cook/steward, she asked what P5 was. “That's secret,” she was told, and promptly volunteered. She ended up
as a Watch Leader in Hut 8 at Bletchley Park, working on Tunny, and recalls Turing as “a very nice man, very quiet; a bit daft, like most of them.”

Colossus was the first electronic computer. It was also programmable, in a limited sense, because Flowers had deliberately designed it so that it could be adapted to new requirements by switches, and by plugging cables linking the logic units in different arrangements. The crucial difference from a modern computer, though, is that it did not store programs in its memory, the way Turing had envisaged; the programming had to be done literally “by hand” at the switches and plugboards. Even so, this adaptability proved an enormous asset, and Colossi could be adapted to use new codebreaking methods as they were invented, carrying out tasks that its designer could not have imagined.

Flowers was asked to have an improved Colossus up and running at Bletchley by June 1, 1944. He was not told why, but the urgency was stressed. The tight deadline was met by having the machine, containing 2,400 valves and running 125 times faster than electromechanical machines, assembled and tested on site. It began operating on June 1, as requested; although Flowers did not know it at the time, this was intended to be D-Day, the date of the invasion of German-occupied France. Bad weather delayed the invasion, and as it continued there were serious doubts about whether the Allies would be able to ship enough men and matériel across the rough English Channel to support the invasion against a counter-attack. But on June 5 Colossus II was instrumental in breaking a message which revealed that Hitler had completely fallen for the Allied deception plan (Operation Fortitude), which led him to believe that the invasion would strike at the
Pas de Calais, with a diversionary raid in Normandy. In the intercepted Tunny signal, he ordered Rommel to hold his forces in the Pas de Calais area to repel the “real” invasion, due five days after the expected Normandy landing. It was this piece of information, combined with a forecast of slightly improving weather, that clinched Eisenhower's decision to go ahead on June 6, knowing that even in bad weather five days would give his forces time to build up the beachhead.

By the end of the war in 1945 eight more Colossi had been installed at Bletchley Park, and Eisenhower himself later said that without the work of the codebreakers the war would have lasted at least two years longer than it did. The two men who did more than anyone else to make all this possible were Turing and Flowers. They should each have been knighted at the end of hostilities, and given every support to develop their ideas further. But that isn't the way it happened.

ANTICLIMAX: AFTER BLETCHLEY

Harry Fenson, a member of Flowers' team, has said that he was well aware at the time that Colossus was “a data processor rather than a mere calculator, and rich in logical facilities.” It had the potential to manipulate many kinds of data, “such as text, pictures, movement, or anything which could be given a value.” It contained “all the elements to make a general-purpose device”—a Turing machine.
¹⁴

At the end of the war, Bletchley Park could (and I would say, should) have become a scientific research center, equipped with ten Colossi and a world lead in computing. Instead, on the direct orders of Winston Churchill (who did many questionable things to set alongside his greater moments), all but two of the machines were physically broken
up and most of their components smashed. This was part of a successful attempt to hide the success of the codebreaking work which had substantially shortened the war, so that the British could carry on reading the coded traffic of other nations without being suspected. The “other nations” included the Soviet Union, which used captured German Tunny machines long after the war. In April 1946, the codebreaking headquarters moved to Eastcote, a London suburb, and changed its name to the Government Communications Headquarters (GCHQ); GCHQ moved on to Cheltenham, its present home, in 1952. In both these moves, it took with it the two remaining Colossi (“Colossus Blue” and “Colossus Red”); the work they did is still classified. One was dismantled in 1959, the other in 1960. But all is not quite lost; a replica Colossus has been built at Bletchley Park, which is now a museum, and can be seen there in all its glory.

As machinery was physically destroyed, so papers were burned and the codebreakers were all sworn to secrecy—and they all kept their secrets, in many cases taking them to the grave. The attitude that wartime secrets should not be inquired into was shared by people outside Bletchley Park. When I asked Betty Houghton what she had said to her parents when they inquired about her war work, she replied, “They never asked.” The story of Enigma did not emerge properly until the 1970s, and that of Colossus became known in detail only after a crucial document, called
General Report on Tunny
and written in 1945, was released in 1996 under the American Freedom of Information Act. Deliciously, it is now available online to anyone with a Turing machine.
¹⁵

Tom Flowers, the man who designed and built the first electronic computer, never imagined that the secrecy would
last so long. Although he was granted £1,000 by the government at the end of the war, this did not cover his personal expenditure on Colossus, so he was actually out of pocket as a result of his work. Flowers was also awarded the MBE (the same honor later awarded to The Beatles), for work designated simply “secret and important”: no details were given. His career was hamstrung by the fact that he could not reveal anything about his wartime work, and so was unable to persuade his superiors to pursue the development of electronic telephone exchanges in the post-war years. This may sound trivial, but in these days of instant global communication it is hard even for those who were around at the time to remember how primitive telephones were even in the 1950s, when “long-distance” calls (that is, anything out of town) still had to be connected by a human operator plugging leads into the appropriate sockets. It was ten years after the end of the war before the Post Office began to move into the electronic era, missing out, apart from anything else, on the opportunity to boost British exports at a time of economic hardship. But Flowers lived just long enough to see the importance of his work beginning to be recognized by the computing community. He was able to give a talk in Boston in 1982 which lifted a corner of the veil of secrecy, and in 1997, on the occasion of his own eightieth birthday, Bill Tutte gave a talk detailing the way Tunny was broken. Thomas Flowers died in 1998 at the age of ninety-two.

Unlike Flowers, Alan Turing was able to pick up the threads of his wartime work after the completion of the Delilah project in 1945. He too was “honored” by the government, with the award of an OBE—one step up from an MBE, but such an inadequate recognition of his true worth
that when Max Newman was also offered an OBE he refused it in protest at Turing's “ludicrous” treatment.
¹⁶

In October 1945, less than ten years after the publication of “On Computable Numbers,” Turing joined the National Physical Laboratory (NPL) at Teddington, in charge of a project to design and build an electronic “universal computing machine.” He was, in fact, head-hunted for the post by John Wormersley, the head of the mathematical research division at NPL, who had been an admirer of Turing's work since reading “On Computable Numbers.” The first fruit of this project was a report by Turing, produced before the end of the year, called “Proposed Electronic Calculator.” This contained the first full description of a practical stored-program computer—one in which the program is stored in the computer's memory, rather than being plugged in by hand. Each program, remember, can be a virtual machine in its own right, so a single computer can simulate other computers; when you open an app on a tablet or smartphone, you are actually opening a stored program that is itself equivalent to a computer. Turing's plan, as set out in this document, was more far-reaching than the work of his contemporaries in the United States (discussed in the
next chapter
). He was interested in developing an adaptable machine that could, through its programming, carry out many different tasks; he suggested that one program could modify another; and he understood better than his contemporaries the use of what we now call subroutines. Unlike modern computers, Turing's machine did not have a central processing unit, but worked in a distributed way, with different parts working in parallel with one another; also, instead of the instructions in a program being followed one
after another in order, the program (or the programmer!) was to specify which instruction to go to next at each step. All of this made his planned computer faster and more powerful than those planned by his contemporaries; but it would require very skilled programmers to operate it. Why did Turing follow this route? As he wrote to a friend: “I am more interested in the possibility of producing models of the brain than in the practical applications to computing.”
¹⁷

The trouble was that, as usual, Turing was far ahead of everyone else, and wanted to build an artificial intelligence before anyone had built a really effective electronic calculator. The project, dubbed ACE (for Automatic Computing Engine), was too ambitious, and Turing's strengths did not lie in project management. He wanted Flowers to work with him, but because of the secrecy surrounding Flowers' wartime work could not explain why his presence would be vital; Flowers stayed at Dollis Hill, collaborating with Turing at arm's length, but was soon ordered to concentrate on his proper job. Things stumbled along, with a great deal of testing but very little computer building, until September 1947, when Turing (whose father had died the previous month) left the project, initially for a year's sabbatical in Cambridge, then moving on to Manchester University. But he left a legacy of programs, a kind of software library, prepared in the expectation of the completion of the project; when ACE eventually was built, its immediate success was largely based on this legacy.

While at King's, Turing developed his running to such an extent that he was planning to enter the trials for the Marathon squad in the 1948 Olympic Games; the plan fell through, according to John Turing,
¹⁸
when as a result of a bet
“Alan dived into a lake in January, contracted fibrositis, and thereby put himself out of the Wembley Olympics.”

A cut-down version of ACE, called the Ace Pilot Model, or Pilot Ace, was completed at the NPL after Turing left, and ran for the first time on May 10, 1950. It contained a thousand electronic valves, and used just a third of the amount of electronic equipment of contemporary British computers, but ran five times faster than them. The design was adapted and taken up by the English Electric Company as DEUCE, and thirty-three DEUCE machines were built and used commercially in the 1950s and 1960s—the last one was shut down in 1970. The first “personal” desk-side computer, housed in a cabinet about the size of a tall kitchen refrigerator, was also based on the ACE design. Marketed by the American Bendix Corporation as the G15, it went on sale in 1954. But even by then, the mainstream of computer design was flowing in a different channel, although the idea of a personal computer was an indication of things to come.

In Cambridge, computer development consciously jumped off from the work in the United States which I will discuss in
Chapter 2
; even the name of the first Cambridge computer, EDSAC (Electronic Delay Storage Automatic Calculator), was deliberately chosen to show its relationship to the American EDVAC. Turing, whose philosophy was to minimize the amount of hardware by maximizing the use of software, described it as “in the American tradition of solving one's difficulties by means of much equipment rather than by thought.”
¹⁹
He was right, but did not appreciate that the average computer user could not think as well as he did, or that the cost of equipment would fall so dramatically, so that today any idiot can use a computer.

The team Turing joined in Manchester was headed by Max Newman, by now Professor of Mathematics; Newman had actually taken unidentifiable bits of a dismantled Colossus to Manchester with him, where some of the pieces were incorporated in the first Manchester computer. It had the distinction of being the first to run a successful program on a stored-program electronic computer. The date was June 21, 1948, and the computer was the “Manchester Baby,” with a random access memory (RAM) equivalent in modern terms to 128 bytes. But it worked. The Baby was the forerunner of the Manchester University Mark I computer, for which Turing developed the programming systems. Audrey Bates, one of the MSc students using the Manchester computer under Turing's supervision in 1948–9, asked if she could go on to work for a PhD; she was told by Newman that he thought it unlikely that anyone would ever get a PhD for working with computers.

Input and output for the Mark I used a system familiar from Bletchley Park days—teleprinter paper tape with a five-bit code punched into it. This kind of tape was still in use for communicating with computers well into the 1960s. As an undergraduate taking a very basic computer course as part of my physics degree I had to prepare programs in this way, before the tapes were taken off to another institution and fed into a computer I never saw (Sussex University didn't have its own computer in those days); the output would be returned a couple of days later as another roll of punched tape (usually with errors caused by the incompetent programming). The Mark I was developed into another commercial machine, the Ferranti Mark I, which in the early 1950s was the most powerful “supercomputer” around—with a RAM of 1 kilobyte.
It used 3,600 valves, housed in two bays each 17 feet long and 9 feet high, and consumed 25 kilowatts of electricity.