The Computer

Significant theoretical progress had been made in the theory of computation. In 1936 the Cambridge University student Alan Turing wrote one of the most influential papers of all time: "On Computable Numbers, with an Application to the Entscheidungs Problem" (published while he was visiting Princeton University). This thought experiment is about a machine capable of performing logical reasoning by manipulating symbols (like a mathematician does), which came to be known as a "Turing machine". He then described the "universal Turing Machine", capable of simulating any Turing machine by reading a symbolic description of the Turing machine to be simulated. It was a purely abstract concept, but it proved the feasibility of a machine capable of solving all problems; in technical words, a "programmable" machine.

In 1937 Claude Shannon at the MIT finished his thesis titled "A Symbolic Analysis of Relay and Switching Circuits", that showed how electronic circuits could implement Boolean algebra.

For practical purposes, though, the "differential analyzer" was still the most advanced computing device. It was a mechanical analog computer, capable of performing sophisticated computations as fast as a professional mathematician. This was an old idea: William Thomson (better known as Lord Kelvin, the same scientist who contributed to the birth of thermodynamics) had already designed in 1876 a mechanical device to solve sets of differential equations. Vannevar Bush, who had become rich with his vacuum-tube manufacturer Raytheon, constructed it in 1931 at the MIT, and copies were installed in 1935 at the Moore School to help the Ballistics Research Laboratory of the Army calculate the trajectory of bombs, the first major collaboration between these two organizations. Bush had a big advantage over Lord Kelvin: in 1925 Henry Nieman (an engineer of the Bethlehem Steel Company in Pennsylvania) had invented the torque amplifier.

The "soldiers" assigned to perform computations for the artillery were usually women, and were called "computers". Other copies of Bush's machine were installed at General Electric's Schenectady laboratories in New York, at the Aberdeen Proving Ground in Maryland, and at the University of Manchester in Britain.

In 1934 Douglas Hartree and Arthur Porter at Manchester University built a differential analyzer using Meccano parts (Meccano being a kit to build toys, invented in 1901 in Britain by Frank Hornby).

The differential analyzer replaced human computers in one of the tasks that for centuries had been associated with human intelligence: no wonder that the press started talking about the possibility of "thinking machines".

The Aberdeen Proving Ground in Maryland had assembled a large number of scientists under Oswald Verben to work on weapons for World War I. Verben had hired an impressive group of mathematical prodigies (human computers), including a young Norbert Wiener. While it had not been finished in time to contribute to that war, it remained operational and it set an important precedent for the USA on how to conduct wartime science.

From the beginning the Bell Labs had featured a department for applied mathematics, led by Thornton Fry. Here George Stibitz had the idea of using relay circuits (the same kind of circuits that AT&T was deploying for telephone systems) to perform binary arithmetic computations. In 1937 Stibitz built a relay-based binary calculator, the "Model K", assembled at home out of telephone relays. Bell Labs then assigned a team in New York to build a Complex Number Calculator (to calculate complex numbers, ubiquitous in electrical circuits) and in September 1940 Stibitz carried out the first major demo in the history of computers: he let mathematicians convened at Dartmouth College for a conference play with this new computer located in New York via commands typed into a teletype and transmitted over telegraph lines, the first case of remote access to a computer.

After this computer made of 450 telephone relays, Stibitz would continue building electromechanical (relay-based) computers at Bell Labs, mostly for military use during the war and then notably the Model 5 of 1946 that was fully programmable and was deployed at Aberdeen and at Langley Field (in Virginia). This machine (made of 9,000 relays) was the first dual-processor computer and featured sophisticated input-output equipment. Its processor was actually not a calculator at all: addition was done by looking up the result in a built-in table of additions, subtraction was done by adding the complement, multiplication was done as a sequence of additions, and division was done as a sequence of subtractions. The Model V was slow but it was very reliable.

In 1940 John Atanasoff and his student Clifford Berry at Iowa State College built the prototype of a special-purpose relay-based digital computer for solving systems of linear equations. It was not programmable and it was electromechanical.

A similar project was underway at NCR's Dayton laboratories, directed by Robert Mumma and Joseph Desch. In 1942 their laboratory was taken over by the military, turned into the Naval Computing Machinery Laboratory at the top-secret Building 26, and used to make electronic machines for cryptoanalysis, basically the US equivalent of Bletchley Park in Britain.

Konrad Zuse built the first program-controlled computer at the end in 1941 in Germany, the "Z3", the first hardware implementation of the Turing machine. It was programmable and digital, but it still employed electro-mechanical relays. Before the Z3, Zuse had painstakingly built the Z1 by hand at home in 1938, a mechanical machines that already used binary logic and was programmed by a punched tape made of photographic film. Zuse therefore (just like Stibitz in the USA) had the fundamental intuition of computer science: it is easier to build mechanical (and, later, electronic) elements that can only assume one of two states rather than elements that can assume one of ten states, and then Boole's algebra can be used to transform our ordinary decimal calculations into the computer's binary operations. Unlike future developments in the USA and Britain, Zuse was an independent working at home. The price he paid was to be excluded from the know-how and the supplies of vacuum tubes. He built his arithmetic unit (the unit that performed the binary operations) out of second-hand telephone relays (more than 2,000 of them). The Z3 was never used for practical applications: its memory was too limited (64 words) and its calculations were too slow (four seconds for a multiplication), and it was destroyed during a bombing raid in 1944. The only computer built by Zuse that was used during the war was the special-purpose S1, used to simulate the dynamics of unmanned flying bombs.

At the beginning of World War II, Germany was still a leader in electrical inventions. For example, the magnetic tape was invented by Fritz Pfleumer in 1927, and the first tape recorder (later dubbed Magnetophon) was by AEG in 1935 (using tape made by IG Farben, also manufacturer of the lethal gas used in Nazi extermination camps) and perfected into a high-fidelity system in 1941. It was only at the end of World War II that a USA soldier, Jack Mullin, brought this novelty back from occupied Germany, and it was only in 1947 that Alexander Poniatoff's Ampex (based in the Bay Area) introduced a tape recorder in the USA.

In 1948 Ampex installed a Model 200A tape recorder in the studio of a radio station, the first such machine used for regular broadcasting. (Bing Crosby purchased 20 of these models)

Computing for World War II

During World War II the US government heavily influenced the course of the office-machine industry. Very few new products were introduced, but IBM's and Remington Rand's punched-card tabulators were widely used by the wartime bureaucracy. Several companies had to change business. NCR, forbidden in 1942 to make cash registers (steel was rationed) and forbidden or unable to sell its machines abroad, hosted the Naval Computing Machinery Laboratory at its Ohio factory. Burroughs built equipment for air warfare. IBM was the lucky one: thanks to Thomas Watson's connections in academia, it was involved in the development of the early computers. The number of employees doubled during the four years of World War II. All three benefited because they had to invest more in electronics. The army also employed variations on the differential analyzers invented by Vannevar Bush at the MIT for ballistic analysis.

Besides computers, the other great invention of World War II was the radar. Several countries were workingon the technology. In 1934 Robert Page at the Naval Research Laboratory (NRL) near Washington had conducted the first test of the technology that the lab renamed RADAR (RAdio Detection And Ranging) in 1939. Robert Watson-Watt in Britain had installed a radar in 1936. Pavel Oshchepkov had designed his own version of the radar in the Soviet Union, which led to the Redut of 1940, but (luckily for the West) in the meantime he had falled victim to Stalin's "purges" and sent to a gulag where he spent ten years of his life.

In 1940 Vannevar Bush, who had become the director of the Carnegie Institution, convinced the US government to create the National Defense Research Committee (NDRC), later renamed the Office of Scientific Research and Development (OSRD). Among the founding members was also Bells Labs' president Frank Jewett.

Vannevar Bush presided over a number of top-secret projects, notably the Manhattan Project to develop the atomic bomb. In 1941 Bush also helped Alfred Loomis (a Wall Street investment banker turned benefactor of science) to create the Radiation Laboratory (the "RadLab") at the MIT, that lasted until 1945, the laboratory where the most advanced radar technology was developed (ten of its physicists would win a Nobel Prize and the lab grew to almost 4,000 workers by the end of the war).

In 1942 Bush placed Fred Terman in charge of the group working on electronic warfare, the Radio Research Laboratory (RRL) that moved to Harvard University. Bush's old company Raytheon had become a major defense contractor (ironically, the most important invention to come out of Raytheon during World War II was the microwave oven, invented by its engineer Percy Spencer who noticed that a particular electron tube was capable of warming up food quickly).

Both the army and the navy were interested in cryptoanalysis to decipher the secret code of the Enigma, the machine used by the Germans to encrypt their messages. This effort was mainly led by British and Polish mathematicians in Britain, at Bletchley Park's top-secret mansion. In 1939 Alan Turing in person (who had returned to Cambridge in 1938 for his graduation) designed the electromechanical Bombe, based on a machine designed in 1938 in Poland by Marian Rejewski; but in 1941 Siemens and Lorenz equipped Hitler's Germany with much faster encoding machines, generating an enormous amount of traffic with which the Bombes were unable to keep up. At the end of 1943 the Colossus Mark 1 debuted, a machine designed by telephone engineer Tommy Flowers (working for the Post Office Research Station in north London). This was a digital electronic computer, and the first electronic computer to be used extensively. It was "electronic" because it used vacuum tubes. Flowers deliberately set out to replace the electromechanical relays (the same ones used in telephone exchanges) with vacuum tubes in order to achieve higher speeds of computation. On the other hand, Flowers did not appreciate the revolutionary concept in Turing's 1936 paper. The machine was programmed manually by plugs and switches. There was no software in those days. By the end of the war, there were ten Colossi at Bletchley Park, progressively more sophisticated (from 1,600 vacuum tubes of the first one to 2,400 of the tenth one). The existence of these machines was kept secret until 1970.

The closest thing in the USA was the result of a joint project between IBM and Harvard University: a prototype, code-named Automatic Sequence Controlled Calculator (ASCC) that Harvard preferred to call "Harvard Mark I", designed by Howard Aiken of the Harvard Computation Lab and built by Claire Lake at IBM's Endicott Labs (reporting to James Bryce), completed in February 1944 after five years of work.

This was a monster, 15-meters long, overflowing with hundreds of kilometers of wiring: it was built out of IBM accounting machines. The press was duly stunned and started reporting the rise of the "electronic brains".

It was the first computer programmed by punched paper tape, but it still employed electro-mechanical relays, just like Zuse's Z3. The paper tape was a significant innovation: all the previous computers required human intervention during the calculation, while the paper tape contained all the instructions needed to run without any human help. The Harvard Mark I printed the output on an Electromatic. The Mark I was immediately monopolized by the US navy. One of its "programmers" was Grace Murray-Hopper, a female officer dispatched by the US navy to the Harvard Computation Lab. She wrote the 500-page manual.

The machine's weakest point was that it was elecromechanical, i.e. it had a lot of moving parts, and was therefore very slow (and noisy), capable of no more than 200 operations per minute. Secondly, it did not implement the conditional branch (the "if... then"), despite the fact that Lovelace had described it in 1843. The Mark I was not Turing-complete and it was decimal, not binary. IBM's version, the ASCC, deployed in May 1944 at the Bureau of Ships, was only different in appearance: Thomas Watson commissioned industrial designer Norman Geddes to give it an appealing look.

Besides Hopper (later at Univac), Aiken's group formed a number of influential early computer specialists: An Wang (later of Wang fame), Kenneth Iverson (later at IBM), Frederick Brooks (later at IBM), Richard Bloch (later at Raytheon). In 1949 his Chinese students An Wang and Way-Dong Woo invented the "pulse-transfer controlling device" that later enabled magnetic-core memories.

The Office of Scientific Research and Development (OSRD) identified computers as a strategic technology and funded a project at the Ballistic Research Laboratory of the Aberdeen Proving Ground, 100 kms from Philadelphia, that worked closely with the Moore School. John Mauchly of the Moore School was painfully aware that calculating technology was lagging behind the development of new weapons and in 1942 he proposed the construction of an electronic computer, a computer using vacuum tubes ("The Use of High Speed Vacuum Tube Devices for Calculating") and engaged electronic (not electric) engineer Presper Eckert (who had worked with Philo Farnsworth) to build one.

Herman Goldstine, the mathematician in charge at BRL (Ballistic Research Lab) of preparing the firing tables of each new weapon, who was desperately hiring human "computers" to do the job and using two differential analyzers, appreciated the idea of replacing all the gears and wheels of the differential analyzer with electronic components. The vacuum tubes had a life expectancy of 3,000 hours but the gain in speed was worth the trouble of replacing a tube every few minutes. "Project PX" was started in 1943 and completed in November 1945, and in February 1946 the ENIAC (for "Electronic Numerical Integrator and Computer") was unveiled, although when World War II had already ended. It contained 18,000 vacuum tubes, occupied about 160 square meters of space, and could perform 5,000 operations per second (a thousand times faster than the Harvard Mark 1). The ENIAC was Turing-complete, but each new "program" meant that skilled engineers had to manually reconfigure it by means of plugs and switches. Eckert, who had worked on radars, invented a kind of non-electronic memory that was a variation on the acoustic delay line used to store radar pulses, an idea originally pioneered by William Shockley at Bell Labs and first applied during the war in 1942 using water as the delay medium. In 1943 Eckert and Kite Sharpless at the Moore School, still working on radar signals, had developed a delay line that used mercury instead of water as the delay medium, and in 1944 Eckert figured out how to use this same technology for the memory of a computer. This memory was very limited (and, by definition, sequential rather than "random") and so it was not suited to storing a program internally: a group of programmers (mostly women) had to organize switches and wires to "enter" the program into the computer. Also, the ENIAC was a decimal computer, not binary. Luminaries such as Vannevar Bush and Howard Aiken were critical of the project.

The ENIAC was "programmed" (a task that required more hardware than software skills) by a group of female "computers" from the Moore School: Kay McNulty, Betty Jennings (aka Jean Bartik), Betty Snyder, Marlyn Wescoff, Fran Bilas and Ruth Lichterman.

Whatever the merits of these early machines, they provided the momentum to fulfill Turing's dream. Therefore, one of the major outcomes of World War II, as far as technology goes, was the computer. The effect of WWII was that governments could assemble the best brains in computer science and electronic engineering to solve big problems instead of having them scatter among competing corporations. Once the military secrets were lifted, these scientists were free to spread the know-how at will: there were no patents and no industrial secrets. The war had thus dramatically accelerated research in electronics, mainly for radio and radar applications, and the computer was an indirect beneficiary of that progress. The MIT's differential analyzer and Aiken's Harvard Mark I suddenly looked obsolete. Electronics percolated into office machines. For example, in 1946 IBM demonstrated an electronic version of its ten-years old Model 601 with a dramatic increase in speed.

Computing for the Future

In 1930 Abraham Flexner (the man who had written the article "The Usefulness of Useless Knowledge" in 1939 in defense of pure scientific research) had founded the Institute for Advanced Study (IAS) in New Jersey, next to Princeton University, with the ambitious goal of recruiting the best physicists and mathematicians in the world.

The new institution had quickly hired from the German-speaking countries, where antisemitic sentiment, peaking with the purges of the German universities in 1933, had convinced scores of scientists to emigrate: Albert Einstein (a German Jew, in 1932), Hermann Weyl (a German whose wife was Jewish, in 1933), John VonNeumann (an Austrian-Hungarian Jew, real name Janos Neumann, in 1933), Kurt Goedel (an Austrian who had studied with Jewish mathematicians, in 1933), Wolfgang Pauli (an Austrian son of a Jew, in 1935), Stanislaw Ulam (a Polish Jew, in 1935), Pal Erdos (another Hungarian Jew, in 1938), etc. Thanks to this institute, Princeton replaced Goettingen (and the USA replaced Germany) as the world's center of excellence for Mathematics. In 1930 Jeno Wigner (another Hungarian Jew) had moved to nearby Princeton University, and in 1936 Turing was studying at Princeton too.

In 1943 John VonNeumann, who had become famous by providing mathematical foundations to Quantum Mechanics in 1932 and was about to publish his theory of games, joined the Los Alamos team working on the atomic bomb. He needed a better machine to solve his partial differential equations and realized that the ENIAC was almost what he needed. In fact, ENIAC's first program (about one million punched cards) was "written" for Los Alamos. The result of discussions held at the Moore School with Herman Goldstine and the ENIAC team was a 101-page document titled "First Draft of a Report on the EDVAC", published in June 1945. EDVAC stood for "Electronic Discrete Variable Automatic Computer" and was meant to be a "stored-program computer": a machine storing in its memory both its data and its instructions. It was also designed to use binary logic. VonNeumann loved the biological metaphor: input, memory, reasoning, output. For the record, the EDVAC paper quoted only one paper: Warren McCulloch's and Walter Pitts' "A Logical Calculus of the Ideas Immanent in Nervous Activity". VonNeumann's network of vacuum tubes was inspired by their network of binary neurons.

VonNeumann called for an electronic memory based on the principle used by Farnsworth and Zworykin to build a television camera (the electron scanning tube, that Zworykin had originally called "iconoscope" in 1923, and that had first been demonstrated by Farnsworth in September 1927).

In August 1946 the Moore School organized a series of lectures titled "Theory and Techniques for the Design of Electronic Digital computer" that spread the word about what came to be known as the "VonNeumann architecture".

Mauchly and Eckert had begun design work on a stored-program version of the ENIAC at the Moore School even before the ENIAC was finished, but quit before the Moore School started building it as the renamed EDVAC. The first EDVAC was eventually delivered by the Moore School's team (now led by Richard Snyder) to the Ballistics Research Laboratory of the Aberdeen Proving Ground in September 1949, but was not operational until (i.e. it did not run its first program until) October 1951 because of technical problems (and only in January 1952 was it considered reliable enough to run a big program), when Manchester had already built a stored-program computer. The EDVAC was never used intensively at Aberdeen because the faster ORDVAC was about to be completed and Richard Clippinger of the BRL had turned the ENIAC into a stored-program computer in September 1948 (Herman Goldstine's wife Adele Goldstine of the Moore School had written the first program for this improved ENIAC). The ENIAC died in 1955, the EDVAC in 1961.

The agreement between the Moore School and Aberdeen was that the EDVAC was to be a preliminary version of the machine, whereas the Institute for Advanced Study would build the full-fledged machine. VonNeumann returned to the Institute for Advanced Study, where Julian Bigelow (a former IBM Endicott engineer who had the MIT in 1943 had coauthored with Norbert Wiener the historical paper "Behavior, Purpose and Teleology" that launched Cybernetics) was hired in May 1946 to lead the development of the stored-program computer. The IAS computer commissioned electronic components to the RCA Research Laboratory, located nearby, and headed by the Russian-born pioneer of television Vladimir Zworykin. In that laboratory the Polish-born engineer Jan Rajchman had invented the "computron" (co-developed with Richard Snyder) and the "selectron", two devices made of vacuum tubes that worked, respectively, like an arithmetic processing unit and a random-access memory. By the end, however, Bigelow chose the Williams tube (just developed in Britain) over the selectron.

Luckily, the US army decided that the EDVAC (or "VonNeumann") architecture was to remain in the public domain, i.e. nobody owned it. It was legal for anybody to build a stored-program computer.

The secret military project "Manhattan Project" had successfully ended World War II with the atomic bombs dropped on Hiroshima and Nagasaki. Those were "fission" bombs. They had been designed at the Los Alamos National Laboratory in New Mexico under Robert Oppenheimer (of UC Berkeley) after Enrico Fermi had achieved the first controlled nuclear chain-reaction at the University of Chicago at the end of 1942. Fermi's team and Oppenheimer's team had used IBM punched-card tabulators and dozens of female human computers armed with electromechanical calculators (ironically, Fermi's favorite calculator was German, made by Brunsviga). At the end of World War II the scientists convened one more time in April 1946 at Los Alamos for a conference on the "Super" bomb, an idea cultivated mainly by Ede Teller (yet another Hungarian Jew who had emigrated to the USA after Hitler's rise to power in Germany). Teller's proposed TX-14 (or the "Alarm Clock") was go be a new kind of atomic bomb: a fusion bomb, a bomb that used hydrogen.

The calculations for this one were much more complex. Robert Richtmyer set out in 1947 to carry out machine calculations of a nuclear fission, the so-called "Project Hippo", that he would eventually run on an IBM SSEC in New York in 1950 (nonstop for several days). In 1946 the Polish-born mathematician Stanislaw Ulam came up with a different method to simulate and estimate (not calculate exactly) a nuclear explosion: the Monte Carlo method. This method provides approximate solutions for intractable mathematical problems ("intractable" because the time required to find the exact solution grows exponential).

VonNeumann understood that this method is perfectly suited for the electronic computer. His wife Klara VonNeumann and Nicholas Metropolis of the Los Alamos lab programmed the ENIAC of Aberdeen in 1948 to run the Monte Carlo method, first on fusion and then on fission problems. Ulam and Metropolis published the first paper on the Monte Carlo method in 1949.

The IAS computer was operational in the spring of 1951. It was much smaller than the ENIAC because because it contained "only" 2000 vacuum tubes compared with ENIAC's 16,000. The memory was implemented in Williams tubes. While the ENIAC at Aberdeen in 1950 was being used by meteorologist Jule Charney for weather simulations, the IAS machine helped Los Alamos simulate the hydrogen bomb.

The research on the hydrogen bomb provided further impetus for the development of the IAS machine, and the IAS machine allowed the simulation that led to the hydrogen bomb designed by Ulam and Teller that was tested in November 1952 on one of the Marshall Islands.

The IAS was still a very difficult machine to program. It was only in 1955 that Hans Maehly developed of a set of reusable subroutines, the FLINT (FLoating point INTerpretative) system, which could basically play the role of a programming language.

Several "clones" of the IAS machine spread the gospel throughout the nation: the ILLIAC I (Illinois Automatic Computer, 1952) at Chicago's University of Illinois; the ORDVAC (the exact same computer, but delivered to the Ballistics Research Laboratory at Aberdeen Proving Ground); the MANIAC I (1952) at Los Alamos Scientific Laboratory in New Mexico, designed by Nicholas Metropolis' team; the ORACLE (Oak Ridge Automatic Computer and Logical Engine, 1952 at the Oak Ridge National Laboratory in Tennessee; the AVIDAC (Argonne Version of the Institute's Digital Automatic Computer, 1953 at the Argonne National Laboratory in Chicago); the JOHNNIAC (1953) at RAND Corporation (RAND stands for "Research and Development") in the Los Angeles area (a non-profit "think tank" spun off from the Douglas Aircraft Company to work on military projects), the only computer that ended up using RCA's selectron. Despite the fact that all of these machines were built for military purposes, their designs were widely publicized within the academic community, helping create momentum for a computer industry.

Von Neumann tested the limits of his new computer by trying to solve a problem that no human being had been able to solve: predicting the weather. While the IAS became famous for simulating the hydrogen bomb of 1952 (or, better, for proving that Ulam's and Teller's design of the bomb was correct), it was the simulation of air flows that had proven the power of the IAS. In 1952 the IAS machine carried out in ten minutes the weather forecast that took the ENIAC 36 hours.

Meanwhile, in Britain in February 1946 Alan Turing, who had been commissioned an electronic computer by Britain's National Physical Laboratory, delivered the design for the Automatic Computing Engine (ACE): "Proposed Electronic Calculator", written shortly after VonNeumann's EDVAC paper. A key difference between the "VonNeumann architecture" and the Turing architecture was that a VonNeumann computer executed instructions in the sequence in which they were located in memory, whereas Turing's ACE used instructions that contained the address of the next instruction (which could therefore be located anywhere in the memory). The Pilot ACE, an incomplete prototype deployed in may 1950, had 1,450 thermionic valves (vacuum tubes). However fast (Turing had actually designed a much faster computer than VonNeumann's), Pilot ACE was not practical for programming, and Turing's template was quickly abandoned in favor of John von Neumann's architecture.

Eckert and Mauchly, having left the Moore School, tried to turn computers into a business under the moniker Electronic Control Company (ECC), formed in 1946 in Philadelphia, and later renamed Eckert-Mauchly Computer Corporation, hiring scientists from the Moore School. The government signed a contract for a machine called Univac ("Universal Automatic Computer"), that was supposed to use magnetic tape for storage, for the 1950 census but EMCC never delivered because Mauchly fell victim of right-wing senator Joseph McCarthy's anti-communist persecution and the company was forbidden to work for the government.

Mauchly was not the only victim of McCarthy's paranoia. Edward Condon was forced to resign in 1951 from the National Bureau of Standards (NBS) that was supporting several computer projects, and another supported of computer projects at the NBS, John Curtiss, was forced to resign in 1953.

The electromechanical ASCC aged quickly. In fact, in 1946 IBM already introduced the Model 603, an "electronic multiplier", a much smaller and cheaper product that had a fully-electronic arithmetic circuit of vacuum tubes and easily outperformed the ASCC. In January 1948 IBM (that had fallen off with Aiken) unveiled an improved version of the Harvard Mark I, IBM's SSEC (Selective Sequence Electronic Calculator), another colossal machine built at IBM's Endicott laboratories by Frank Hamilton's team in collaboration with Columbia University's Wallace Eckert: it contained 12,500 vacuum tubes and 21,400 relays. It was not quite a stored-program computer, and therefore already obsolete when introduced, but at this point the race was on to build bigger and bigger machines. It was so big and expensive that it was only used at Columbia and only for scientific calculations (mainly for Eckert to calculate the positions of the planets). Between 1948 and August 1952 (when it was dismantled) the SSEC raised a generation of programmers at Columbia University, notably John Backus and Edgar Codd at Columbia.

The SSEC was also one of the first (real) computers to be featured in a film (Alfred Werker's 1952 "Walk East on Beacon"). It is telling that the machines showed in most movies, such as Irving Pichel's "Destination Moon" (1950), Rudolph Mate's "When Worlds Collide" (1951) and Fred Sears' "Earth vs the Flying Saucers" (1956), were not computers but differential analyzers. Bush had left MIT in 1938, but work on his new Rockefeller Analyzer (code-named RDA2) had continued and this (electromechanical) machine had become operational in 1941 just in time to calculate firing tables for World War II and to be used by the MIT Radiation Laboratory. It was, in fact, the most powerful and strategic machine in existence until the ENIAC was unveiled. Osaka Imperial University had built one in 1944. In 1947 the big news in commercial computing was not an electronic computer but the first differential analyzer sold by General Electric (to UCLA), one of six that would be built. It was advertised as a machine capable of doing in two weeks the calculations that would take a skilled mathematician 17 years.

Ironically, the difference engine that Babbage had envisioned for astronomers and mathematicians ended up becoming the differential analyzer used to computing ballistic firing tables for cannons.

The main customer for IBM's early computers was the Naval Surface Weapons Center at Dahlgren (in Virginia): they bought the electromechanical Mark II in 1948 (another dual-processor machine) and then the semi-electronic Mark III in 1951 (5,000 vacuum tubes and 2,000 relays), both designed again by Howard Aiken like the Harvard Mark I and both serviced by a team that included Grace Murray-Hopper.

For the Mark III (completed in 1949) Aiken invented an expedient to help programmers write their programs. Hopper had programmed the previous machines by hand, entering the binary code that the machine would understand. Aiken figured that a major productivity gain was to be obtained by letting programmers write in their usual algebraic language and designed a "programming" machine that translated such algebraic language into machine code. This machine punched the code into a tape that was then fed into the Mark III. This represented the first step towards a "compiler". In 1952 Heinz Rutishauser in Switzerland, a user of Zuse's Z4 machine, imagined that a stored-program computer could be programmed to be its own programming machine: first translate the user's commands into machine code and then execute that code. Neither Aiken nor Rutishauser could implement a real compiler because their machines were not stored-program computers. The real world did not use the digital computer. Wallace Eckert of Columbia University designed IBM's PSRC (Pluggable Sequence Relay Calculator), delivered in December 1944 to the BRL at Aberdeen. Thanks to a clever configuration of cables and plugboards, this electromechanical relay-based calculator read numbers from punched cards, performed a sequence of calculations on them (up to 48 steps), and punched the results on other cards. It wasn't just a tabulator or a multiplier, but a computer (albeit electromechanical and limited to 48 steps) capable of any arithmetic combination.

In 1946 IBM applied vacuum tubes to its line of multipliers (the glorious 601) and introduced the electronic multiplier 603 (the first electronic product sold by IBM). In 1948 IBM's 604 calculator was basically a PSRC with vacuum tubes (borrowing the 603 circuits) instead of relays. Finally, in 1949 IBM introduced the Card Programmed Calculator (CPC), a machine designed at the Endicott laboratories, that connected a calculator (the Model 605) to an accounting machine (the Model 407) and a card punch; a machine that removed the need for human operators to carry decks of punched cards from one machine to another, that was programmed via punched cards instead of plugboard wires, and that printed the result on regular paper pages rather than punched cards. This machine was capable of 60 steps of calculation. These were the IBM machines that remained popular until the mid-1950s.

Trivia: The Original App

The Transistor

In december 1947 the brand new discipline of computer science was given a boost by the invention of the transistor: AT&T Bell Labs' engineers John Bardeen, William Shockley and Walter Brattain demonstrated the principle of amplifying an electrical current using a solid semiconducting material.

As the name implies, a semiconducting material lies in between electrical conductors and insulators. Germanium and silicon are two of the most common semiconducting materials. Unlike conductors, which always conduct, and unlike insulators, which never conduct, the behavior of a semiconducting material can be customized by "doping" it, i.e. by disturbing it with an electromagnetic field or by heating it. In other words, pure germanium and silicon conduct poorly, and "dopants" are added to increase conductivity: either extra electrons (negative-type dopant) or extra holes (positive-type dopant). The trick in the fabrication consists in joining "positive-type" semiconductors with "negative-type" semiconductors, where positive or negative refers to the majority electrical charge. That "p-n junction" is the elementary unit of a semiconducting device. A transistor is simply made of two "p-n junctions" (usually n-p-n). The first transistor was truly made by Bardeen and Brattain, using germanium, but Shockley was the one who immediately realized the potential of the transistor: it was a better amplifier than the vacuum tube and it was easier to mass manufacture.

The first transistor radio, the TR-1, was introduced in 1954 by Regency using Texas Instruments transistors, but the early bipolar junction transistors were far too slow for wireless communications. Sony developed its own improved transistors for its TR-55 (1955) and especially for the smaller TR-63 (1957) that became the first mass-produced transistor radio and launched the vogue for transistorized consumer electronics. For the same reason RCA introduced in 1956 the 2N247 drift-transistor after hiring Herbert Kroemer who had come up with the idea while working at the telecommunications laboratory of the German Postal Service.

Case Study: the Bell Labs

Founded in 1925 by AT&T, the Bell Labs in New Jersey had become a magnet for engineers from every part of the USA. It enjoyed immense funding because AT&T had the monopoly of telephone communications in the USA. Bell Labs was, in fact, an odd case of a private research laboratory funded by the customers with approval from the government: the government de facto allowed AT&T to charge the cost of the research on the customer's phone bill.

The Bell Labs were mostly working on long-term projects. They belonged to an organization whose products had a life expectancy measured in decades. The Bell Labs themselves were not making products that would be sold in stores, i.e. that marketing campaigns could turn into front-page news. The job at Bell Labs was similar to tenure faculty, but without the obligation to teach and with greater accountability. Throughout their history they had virtually no connection with nearby aristocratic Princeton University.

In 1936 Mervin Kelly had become the director of the labs. His fame was due to the vacuum tube, that his group had perfected, but he was keenly aware of the limitations of the device: it was unreliable and it sucked a lot of power. In 1945, at the end of the war, he established the solid-state team (unusually for those times, an interdisciplinary group) under the supervision of the relatively young William Shockley, an MIT graduate who had been with the company for ten years. Kelly had created an environment that valued collaboration over cooperation: it was anathema to compete with fellow scientists. Cooperation among Shockley, John Bardeen and Walter Brattain had led to the invention of the transistor. Then Shockley migrated to the San Francisco Bay Area and transferred the know-how to what would become Silicon Valley.

Meanwhile, AT&T shared transistor technology with the rest of the country out of two considerations, one purely political (they had to avoid being accused of being the monopoly that they were) but the other one moral: as internal correspondence shows, they realized the importance of the discovery and felt a moral duty to share it with the scientific world. After the transistor, other notable inventions to come out of the Bell Labs would be: the solar cell (in 1954 by Daryl Chapin, Calvin Souther Fuller and Gerald Pearson), the laser (in 1958 by Arthur Schawlow and Charles Townes), the communications satellite (the Telstar in 1962), the touch-tone telephone (in 1963 by Leo Schenker and others), digital imaging (in 1970 by Willard Boyle and George Smith), the Unix operating system (in 1971 by Kenneth Thompson and Dennis Ritchie), and the cellular phone system (first tested in Chicago in 1978). In the late 1960s the Bell Labs had about 50,000 people.

The Bell Labs won several Nobel Prizes: Clinton Davisson (1937), the three inventors of the transistor (1956), Phillip Anderson (1977), and Arno Penzias and Robert Wilson (1978), who discovered the cosmic background radiation (and would win two more in the 1990s). However, the Bell Labs remained mainly focused on refining a giant communication platform: not disruptive innovation, but sustainable evolution. They ended up missing packet switching and fiber optics, two of the most important innovations in their field.

The labs also had a history of racial discrimination (particularly against Jews), which clearly did not help diversity, and were not as good at importing foreign immigrants as California would be.

Automata

Other revolutionary ideas revolved around the concept of information. In 1945 Vannevar Bush at the MIT envisioned the Memex, an electromechanical device capable of accessing archives of microfilms (which at the time were the most widely used format of storage after paper), of creating paths of navigation by linking pages, and of combining microfilms with annotations.

In 1943 the MIT mathematician Norbert Wiener founded Cybernetics, having realized that machines and animals shared two fundamental concepts: control and communication (the first paper was titled "Behavior, Purpose and Teleology", co-written with physiologist Arturo Rosenblueth and engineer Julian Bigelow).

In 1943 the self-taught mathematician Walter Pitts (who had never finished high school) and the neurophysiologist Warren McCulloch (25 years older than McCulloch) at the University of Chicago (in the paper "A Logical Calculus of the Ideas Immanent in Nervous Activity") introduced the concept of a "binary neuron", a neuron that can only be on or off and that is part of a network of similar "logical" neurons, and proved that this network is equivalent to a Turing machine. They managed to model the brain with logical calculus, and therefore provided the first mathematical explanation for how the human brain can perform such powerful tasks. From that point one, the brain would be conceived as an information processor.

Opening a panel titled "The Design of Machines to Simulate the Behavior of the Human Brain" during the Institute of Radio Engineers' Offsite Link Convention, held in New York in 1955, McCulloch confidently stated that "we need not ask, theoretically, whether machines can be built to do what brains can do" because, in creating our own brain, Nature already showed us that it is possible.

In 1945 Claude Shannon at AT&T's Bell Labs was researching how much information could be sent on a noisy phone line. In the process, he came up with a mathematical definition of "information", founded Information Theory, and used the term "bit" (coined by John Tukey) to describe the fundamental unit of information. Shannon's definition implied that an unexpected occasional event contains more information than a regular behavior. (His original manuscript "Mathematical Theory of Cryptography" was kept secret in 1945 and eventually expanded and published as the article "A Mathematical Theory of Communication" in 1948, which became the core of the book cowritten with Warren Weaver, "The Mathematical Theory of Communication," 1949).

According to Rob Goodman, co-author of the Shannon biography "A Mind at Play" (2017), Shannon planned a memorial parade for his own funeral, featuring a jazz combo, a 417-instrument marching band, acrobats, a computer chess game and juggling robots (all things that he had done himself in his lifetime). Shannon was also responsible for introducing entropy in information theory. His entropy measures uncertainty in information. This concept of entropy is useful because it ends up expressing the fact that in information processes whatever happens next depends on whatever just happened.

The founders of cybernetics regularly convened from 1946 until 1953 at the Macy Conference on Cybernetics, organized by the Macy Foundation of New York when Willard Rappleye was its president. These conferences, that sometimes occurred twice a year, were truly interdisciplinary. Speakers at the first conference were: John von Neumann (computer science), Norbert Wiener (mathematics), Walter Pitts (mathematics), Arturo Rosenblueth (physiology), Rafael Lorente de No (neurophysiology), Ralph Gerard (neurophysiology), Warren McCulloch (neuropsychiatry), Gregory Bateson (anthropology), Margaret Mead (anthropology), Heinrich Kluever (psychology), Molly Harrower (psychology), Lawrence Kubie (psychoanalysis), Filmer Northrop (philosophy), Lawrence Frank (sociology), and Paul Lazarsfeld (sociology).

In 1947, at the Second Cybernetic Conference, Warren Pitts (by then a graduate student at the MIT) announced that he was writing his doctoral dissertation on probabilistic three-dimensional neural networks. (Unfortunately, he burned his unfinished doctoral dissertation).

In 1947 John Von Neumann advanced the notion of self-reproducing automata, expanding Turing's concept of the "universal machine" into the concept of a "universal constructor" that can build copies of itself. His colleague at Los Alamos, Stanislaw Ulam, who was simulating the growth of crystals, came up with the idea of employing "cellular automata" (discrete units of space whose state changes over discrete intervals of time is determined by the state of neighboring units). VonNeumann proved under which circumstances a pattern makes endless copies of itself. In 1948 VonNeumann gave a lecture titled "The General and Logical Theory of Automata", and in 1953, one month before James Watson's and Francis Crick's paper on the structure of DNA, VonNeumann delivered a series of four lecturers at Princeton University titled "Machines and Organisms". The other center of research on self-reproducing automata was the the University of Michigan, where Arthur Burks (who had worked at the IAS with VonNeumann) in 1946 founded the Logic of Computers Group to expand VonNeumann's theory of automata. Burks published von Neumann's seminal papers in 1966.

Around the same time the Italian-born biologist Nils Barricelli published "Numerical Models of Evolutionary Organisms" (1948) and then, mentored by VonNeumann, simulated a population of self-reproducing organisms in 1953 at the IAS.

In 1948 Alan Turing submitted a report titled "Intelligent Machinery" that described how an intelligent machine could be built consisting of binary neurons connected in a random manner. And he wrote: "My contention is that machines can be constructed which will simulate the behaviour of the human mind very closely."

The first robots appeared: in 1949 William Grey-Walter built the Elmer and Elsie robots.

And in 1951 Claude Shannon conceived his maze-solving robots ("electronic rats").

In 1950 the same Claude Shannon explained how to program a computer to play chess.

In 1950 Turing even proposed a test (albeit as a mere teaser) to determine when a machine could be considered intelligent, which came to be known as the Turing Test: if a human observer, asking all sorts of questions, cannot tell whether the agent providing the answers is human or mechanical, then the machine has become intelligent.

Human imagination was jumping way ahead of human technology.

The First Computers

What really made the difference was the memory: the Williams tube allowed memory to extend enough so as to hold not only the data but also the program. The world's first stored-program computer, officially named Manchester Small-Scale Experimental Machine (SSEM), but nicknamed "Baby", ran its first program in June 1948.

The "Baby" was the blueprint for the Manchester Mark 1, that was ready in April 1949 with the addition of a "drum memory" (a rotating form of magnetic memory invented by Andrew Booth in 1948), and in February 1951 British defense contractor Ferranti unveiled its commercial version, the Ferranti Mark I, the first programmable electronic computer that could actually be purchased and used by a customer (Vivian Bowden was the first computer salesman). It had a RAM of 256 words of 40 bits each, and a drum memory of 16,000 words. Almost immediately the engineers set out to use the Ferranti computer for playing board games: in 1951 Christopher Strachey wrote the first checkers-playing program and Dietrich Prinz wrote the first chess-playing program.

Another early stored-program electronic computer from Britain was the Electronic Delay Storage Automatic Calculator (EDSAC), that debuted in May 1949, built by Maurice Wilkes at Cambridge University using as input device a British Post Office telegraph tape and as output device a Post Office teletype printer.

His software engineer David Wheeler also implemented the first library of "subroutines", of software that can be used by multiple programs. In 1951 Wilkes, Wheeler and Stanley Gill published the seminal book "The Preparation of Programs for an Electronic Digital Computer", one of the first textbooks about software.

A version of the EDSAC called LEO (Lyons Electronic Office) was deployed by John Pinkerton at food conglomerate Lyons and ran its first program in September 1951: it was the first business application of an electronic computer, and the beginning of "office automation". In 1954 Lyons started selling LEO computers to other organizations. For the record, the EDSAC 2 of 1958 would be the first bit-sliced computer, its processor consisting of a set of modules.

The first stored-program electronic computer to be deployed in the USA for general use was the Standards Eastern Automatic Computer (SEAC) in May 1950, built by the National Bureau of Standards (NBS) in Washington. In fact, the SEAC, originally meant as an "interim" computer when it was conceived in 1948, was simply a scale-down version of the EDVAC. It was also the first computer to use semiconductor devices (10,500 germanium diodes) for its logic circuits. Its magnetic disk storage (designed by Jacob Rabinow) actually predated IBM's magnetic disk, although it was never sold. It had a memory of 512 words of 45 bits each (the equivalent of about 2.5 kilobytes). A program was entered via teletype or paper tape as a string of hexadecimal characters. The console to control the computer had no keyboard but just switches and dials. The computer was born because the NBS lost faith in the various projects underway in the USA, all of them being delayed multiple times: the EDVAC, the Whirlwind, the IAS, the ERA. The application for which a computer was urgently needed was Project SCOOP (Scientific Computation of Optimum Problems), the project for which in 1947 the mathematician George Dantzig had invented the method later known as "linear programming".

The NBS trained a relatively large staff of programmers. The "interim computer" was designed by a huge team of 33 people, organized in two groups: the engineers, headed by Sam Alexander and featuring Ralph Slutz (who had worked with VonNeumann on the IAS), Samuel Lubkin (who had worked on the Univac), Bob Elbourn, Ruth Haueter (the sole female engineer) and Sidney Greenwald; and the mathematicians, headed by Ed Cannon and featuring Ida Rhodes, Ethel Marden and Joe Levin.

The SEAC computer would remain in use until 1964. In 1957 Russell Kirsch at the National Bureau of Standards built the first image scanner and fed it into the SEAC, the first time that a digital photograph was stored in a computer.

In July 1950 UC Los Angeles (UCLA) completed a computer for the government's National Bureau of Standards (NBS), code-named Standards Western Automatic Computer (SWAC) and designed by Harry Huskey, who in 1947 had been a member of Turing's ACE team: it contained 2,300 vacuum tubes and used Williams tubes (later magnetic drums). It was meant as the West Coast counterpart of the SEAC, but deliberately conceived as a different beast altogether so that different computer architectures could be compared. Huskey actually wanted to use the SWAC to work on machine translation.

The staff of the National Bureau of Standards (NBS) was valuable to the computer industry of the West Coast because many of its programmers accepted to move west and joined the aviation industry.

Meanwhile, in 1945 IBM had founded a Watson Scientific Computing Laboratory at Columbia University under the direction of the astronomer Wallace Eckert. In 1946 Columbia became the first university in the world to offer a course on electronic computers (Eckert's "Machine Methods of Scientific Calculation") and in 1947 a group of computer scientists (including Wallace Eckert and Grace Hopper) was invited at Columbia University by Edmund Berkeley to form a society that would evolve into the Association for Computing Machinery (ACM).

Interest in electronic computers was skyrocketing, although only a handful of people had actually seen one and even fewer were capable of using them. The main centers for research on electronic computing were Boston (Harvard University and MIT), Philadelphia (Moore School, BRL), New York (IBM and Columbia University), and New Jersey (Bell Labs, RCA Labs). Coincidence or not, Boston had been the printing center of the American colonies since 1639 and Philadelphia had been the papermaking center of the colonies since 1690.

The excitement in the press was exceptional. The first book on electronic computing was written by Edmund Berkeley and titled "Giant Brains or Machines that Think" (1949). In this book Berkeley, who had worked on the Univac when employed at the insurance company Prudential and co-founded the Association for Computing Machinery (1947), also described Simon, a small computer built out of relays for educational purposes. Quote: "We shall now consider how we can design a very simple machine that will think". In 1950 Berkeley began publishing detailed instructions in Radio-Electronics magazine on how to build the machine. While never used other than for fun, the idea would be influential on future designers of small computer. In fact, "Giant Brains" was the only book that explained how to actually build a computer. The only other source of information for aspiring computer inventors was the journal Mathematical Tables and Aids to Computation published by the American Mathematical Society that from the very first issue in 1943 sported a section on "Mechanical Aids to Computation". In 1950 Berkeley himself launched the first computer magazine, Computing Machinery Field, later renamed Computers and Automation.

With the early computers came the first computer games: in 1951 Christopher Strachey, a physicist at the National Physical Laboratory, developed a draughts game on the Manchester Electronic Computer Mark II, and in 1952 Sandy Douglas developed developed OXO (a variation of noughts and crosses) on the EDSAC computer of the University of Cambridge.

The government continued to be the main investor in the high-tech industry because of the Cold War with the Soviet Union. The market for computers was actually quite large, but it was almost entirely military.

Another British pioneer was Andrew Booth at Birkbeck College in London, who in 1951 developed the Hollerith Electronic Computer 1 (HEC 1) and in 1954 the APE(X)C, or All Purpose Electronic (X) Computer series.

Continental Europe was lagging far behind Britain and the USA. In 1950 the only computer in continental Europe was the Z4, built by Zuse (a more mechanical version of the Z3) and deployed at the Federal Polytechnic of Zurich (ETH), and it would remain in service till 1960.

The first VonNeumann-architecture computer in Europe was built in 1950 in Ukraine's capital Kiev, then part of the Soviet Union, by Sergei Lebedev's team, and it was called MESM. Trevor Pearcy built the CSIR Mark 1 (later known as CSIRAC) in 1952 in Australia. (The Z4 and he CSIRAC are the only surviving stored-program computers of that age).

Whether by coincidence or not, in 1953 another fundamental discovery set the foundations for a major scientific revolution: James Watson and Francis Crick published a paper about the structure of DNA. Oswald Avery had discovered that genes are made of DNA in 1944, and it was obvious that DNA contained some kind of "code". Crick and Watson showed that this genetic code is organized as a double helix. The analogy with the computer was not lost on the thinkers of the age: here is the program (the genes) and here is the output (the human body); here is the hardware (the cell) and here is the software (the genetic code). Nature and fate were helping computer scientists imagine the universe in terms of Turing-style computation. (For the record, the genetic code would be cracked in 1961 by Marshall Nirenberg and Heinrich Matthaei when they discovered how the 4-letter code of DNA gets translated into the 20-letter language of proteins).

Office Machines

Game Theory

Game Theory was born with John VonNeumann's essay "On the Theory of Games of Strategy" (1928), written when he was still working on Quantum Mechanics in Germany. This essay contains his proof of the "minimax" theorem. In 1944 he published the book "Theory of Games and Economic Behavior", co-written with the German economist Oskar Morgenstern at Princeton University. Game Theory appealed immediately to military strategists of the USA at a time when the Cold War was basically a war game between the USA and the Soviet Union; and in 1948 the RAND Corporation had already designed a war game named SAW (Strategic Air War). VonNeumann's book launched the field in academia, especially among mathematicians. For example, in 1949 the mathematician John Nash, also at Princeton University, proved that any competitive game has at least one scenario of equilibrium, which came to be known as Nash Equilibrium. In 1950 Merrill Flood and Melvin Dresher at the RAND Corporation introduced the "prisoner's dilemma", the most famous of the scenarios studied by game theorists. In 1951 the Air Force in collaboration with the University of Chicago set up the first research laboratory for simulation of flight control systems. In 1954 RAND constructed a war gaming facility, Project Sierra. Its wargames were manual games. RAND specialized in strategic games. The statistician Alexander Mood wrote the paper "War Gaming as a Technique of Analysis" (1954). SAW evolved into STRAW (STRategic Air War, 1953), a game played at the Air War College in Alabama to analyze the the economic effects of atomic bombings. In 1948 the Army had started a collaboration with Johns Hopkins University, the Operations Research Office. In 1953 this joint group launched a computer game about air defense that also happens to be the first computer simulation in the history of operations research. In 1956 this project delivered the first digital computer game, Carmonette (Computerized Monte Carlo Simulation). In 1955 the RAND Corporation presented the Air Battle Model (ABM), a simulation game of global nuclear war, at the Air War College. In 1955 the Naval War College designed a manual strategic game named "School of Naval Warfare Strategic Game" the simulated a global conflict between two superpowers. In 1956 the American Management Association, based in New York, hired consultants from RAND and IBM to translate war gaming into business gaming. "The Top Management Decision Simulation" running on an IBM 650 computer was demonstrated to 20 corporations in 1957. In 1957 the Air Force established an Air Battle Analysis Center to coordinate all research on games. By then SAGE had become the center of mass for computerized simulations. In 1958 a computer for simulation of war games was installed at the Naval War College, the Navy Electronic Warfare Simulator (NEWS).

In 1961 Dresher published the book "Games of Strategy", that summarizes RAND's approach at a time when games were routinely programmed into computers, like STRAT (Strategic Air Planner) running on an IBM 7090 computer.

Data Processing

In 1950 the three big data-processing companies (Remington Rand, NCR and IBM) all had about 30,000 employees, while Burroughs had 15,000; but for the other data-processing companies catching up with IBM and Remington Rand in the computer business was not trivial: it was a tiny market (basically, just government bureaucracy and military applications) and it required a huge investment.

The government further helped IBM, albeit involuntarily, by sueing it again in 1952 under the Sherman Antitrust Act. It was a reasonable lawsuit because IBM indeed was becoming a monopoly in the tabulator business, but the net effect was that IBM's next boss (Watson's namesake son Tom) steered the company away from tabulators and towards electronic computers.

Remington Rand had to acquire the computer skills that it didn't develop in-house: in 1950 it acquired the Eckert-Mauchly Computer Corporation (about 12 people), and in 1952 it acquired Engineering Research Associates (ERA). The original Univac was finally delivered to the US government (to the Census Bureau) in March 1951 (actually, that first Univac I remained hosted in Univac's plant until the end of 1952 by which time a second Univac I had already been deployed at the Pentagon for the project SCOOP). The Univac I was up and running more than 80% of the time, which was a very good rate by the standards of vacuum-tube machines. It printed output at the speed of ten characters a second on a teletype that was an enhanced Remington Rand typewriter. In 1952 the fifth Univac I correctly predicted the results of the presidential elections (based on partial results and previous trends), a major publicity stint. In April 1953 that machine, purchased by the Atomic Energy Agency (AEC), was installed at the Lawrence Livermore Labs to carry out the usual weapons design but it had just proven that computers could be used for a broad range of applications. Eventually, 46 Univac I machines were built. Despite its first customer, the Univac was conceptually a different machine: unlike the IAS, that focused on speed of calculation for scientific applications, Eckert and Mauchly had focused on fast input-output, not speed of calculation. It was the spirit of the old tabulators, of the business world. In fact, the Univac was better than punched-card tabulators at sorting data, one of the most crucial tasks for businesses. In 1952 Betty Holberton, one of the ENIAC programmers who was now at Univac, wrote the first sort program.

Businesses also loved the magnetic tape, that represented a major revolution in the workplace: instead of people carrying decks of punched cards from one machine to another, the Univac was able to read data from the magnetic tape and write data on the same tape. It saved time and reduced the staff needed in the data-processing room. Univac used vacuum-tube amplifiers in the Uniservo tape drive, one of the first cases of electronic amplifiers for servomechanisms.

At the end of the war (1946) the US navy had dispatched a group of veteran Washington code-breaking scientists (notably William Norris) to set up a company at John Parker's Northwest Aeronautical Corporation (NAC) in Minneapolis, Engineering Research Associates (ERA), with (among other things) the mission to build its top-secret computer, the stored-program Atlas (operational in December 1950 but only available to the Navy), sold commercially in December 1951 as the ERA 1101 (only one unit was sold). This was one of the first machines to use magnetic drum storage. It displayed results on a 6x6 character display.

When Remington Rand acquired ERA, the ERA 1101 was renamed Univac 1101. In February 1953 ERA commercialized the Atlas II as the Univac 1103, and this time it sold 20 to the general market. The Univac 1103, introduced commercially a few months before the IBM 702, was the first commercial computer in the USA equipped with Random Access Memory (RAM) a` la Manchester Mark 1. In fact, in 1954 an experimental version of these machines used the magnetic core memory invented at the MIT, and in 1956 the Univac 1103A definitely replaced the Williams tube memory with core memory. The ERA team included a young Seymour Cray.

Therefore in 1952 Remington Rand owned the most complete product line of any computer companies: the Univac machines for business applications and the ERA machines for scientific applications. The Univac computers (and the IBM computers that followed them) fulfilled Wallace Eckert's dream of automating the entire data-processing room, not just improving one of its machines.

The biggest legacy of the ERA machines was perhaps the magnetic drum storage, that they had perfected before the Remington Rand acquisition and even licensed to IBM in 1949.

The magnetic drum storage allowed computer startups with limited financial resources to build computers. During the 1950s several of these startups were born (CRC, Consolidated, LPG, Bendix, etc), although most of them were acquired by the bigger traditional office-equipment manufacturers.

The Computer Research Corporation (CRC) was formed in 1950 by the team that had worked at Northrop (an aircraft manufacturer of the Los Angeles area) on a special-purpose digital computer, MADDIDA (MAgnetic Drum DIgital Differential Analyzer), designed in 1949 by one of them, Floyd Steele. Max Palevsky then created the commercial version of 1952 that was purchased by six customers (a record for an electronic computer). This was the last differential analyzer: that age had ended.

CRC introduced their 102A in 1953, the commercial version of the CADAC that they had built for the Air Force, one of the several computers of the era based on drum storage and equipped with a Flexowriter for input/output functions.

The second customer of a digital electronic computer after Lyons (not counting government agencies and military) was General Electric, that bought a Univac I for its model Kentucky plant to perform payroll, accounting and logistics. In 1954 other Univac machines were shipped to three life insurance companies (in New York, Ohio and Los Angeles), US Steel (in Pittsburgh), Du Pont (in Delaware), Sylvania (in New York), and to Consolidated Edison (the successor to the Edison Illuminating Company, in New York).

In 1951 Grace Murray-Hopper, a pupil of Aiken at Harvard who had joined the Eckert-Mauchly Computer Corporation in 1949,developed a program on the Univac to handle subroutines that she called the "A compiler" (available to users in late 1953). This was therefore not a real compiler, but a program similar to Wheeler's system. The first compiler, that translated algebraic formulas into machine code, was developed in 1953 by Hal Laning and Neal Zierler for the Whirlwind computer. Aiken's programming machine, Rutishauser's intuition, Hopper's "compiler", Wheeler's system and Laning-Zierler compiler constituted a step towards software engineering at a time when software was still in its alchemic stage. Programmers were virtuosi who worked with machine code exploiting every possible shortcut to make the machine do what they wanted, with the result that their code was impossible to understand by anyone else. The only help for programmers were the assembly languages, that simply translated a symbol into a machine code. Hopper's first compiler for the Univac was the B-0 or Flow-matic (1955).

Computers were still programmed with punched cards. The programs were mostly typed on punched cards by women, who were typing on a keyboard similar to a typewriter but with a key difference: they could not see what they were typing.

New York and Boston now controlled the computer industry. Philadelphia, whose Moore School had been so instrumental in inventing the computer, was rapidly left behind. In 1948 Howard Aiken had killed any prospect for Eckert's and Mauchly's startup of receiving money from the government when he coauthored a negative report on computers (the one where he estimated that there was a market for only about five computers in the world). To make matters worse, the University of Pennsylvania was hostile to collaborations between academia and industry (exactly the opposite approach that Fred Terman was advocating at Stanford and that the MIT was advocating in Boston), one of the reasons of Eckert's and Mauchly's departure.

In 1952 John Diebold of Harvard's Business School published a book titled "Automation" that coined a new discipline, although Diebold was more influenced by Wiener's cybernetics and its notion of "feedback" than by Turing's and VonNeumann's computers. In the same year Kurt Vonnegut published the sci-fi novel "Player Piano" that revived the spectres of Orwell's "1984" and Huxley's "Brave New World" in the more realistic setting of a fully automated society, thereby starting a debate that would rage for decades about machines stealing people's jobs and controlling people's minds.

Alan Turing (a persecuted homosexual) committed suicide in 1954. Von Neumann died in 1957, ironically of a cancer caused by nuclear radiations. The era of the computer as a university research project had come to an end. The future belonged to the manufacturers of office machines.

(More pictures are here)

The West Coast: Los Angeles

Located far from the research centers of the large office automation companies (IBM, NCR, Burroughs, Remington), peripheral to the strategies of the giants of the electronic industry (General Electric, RCA, AT&T) and left out of the loop of the large government-funded computing projects (Boston's Lincoln Lab and Philadelphia's Moore School), on the West Coast the computer industry was limited to serve the needs of the booming aviation industry of Los Angeles, namely Northrop, Douglas and Hughes.

Northrop, however, did not believe in electronic computers. In 1950 some of its engineers quit to form Computer Research Corporation (CRC), the company that was later acquired by NCR in 1953. The rest of its computer lab was sold to Bendix, a maker of appliances and radios based in Indiana. In March 1956 Bendix introduced their first digital computer, the Bendix G-15, designed by SWAC designer and Turing's former collaborator Harry Huskey according to Turing's ACE architecture (and the team included a young David Evans). It took input from paper tape or punched cards, stored data on a magnetic drum, and sent its output to a typewriter (about ten characters per second) or a pen plotter. It was much smaller than the monsters created by IBM and Univac, a premonition of the minicomputers to come.

In 1957 some of the former Northrop engineers led by Max Palevsky quit Bendix and joined Packard Bell (a Los Angeles-based maker of consumer radios) to open their computer labs.

In 1953 Boston-based Raytheon (having hired engineers from Howard Aiken's Harvard lab) developed the code-named Hurricane (later RAYDAC, Raytheon Digital Automatic Computer) for a Naval Air Missile Test Center near Los Angeles, a computer that was supposed to replace the SWAC at the NBS.

Finally, there were at least two spin-offs of the California Institute of Technology (CalTech): the Librascope division of General Precision (that built another "small" computer, the LPG-30, designed by Stanley Frankel at CalTech, yet another "drum machine" using a Flexowriter, of which 400 units were sold); and the Electrodata division of instrument maker Consolidated Engineering Corporation (CEC), whose main customer was CalTech, and whose chief scientist was Clifford Berry. ElectroData built the ElectroData 203, delivered to CalTech's Jet Propulsion Laboratory in 1954,

The West Coast: Bay Area

At the end of World War II, Fred Terman had returned to Stanford University as the dean of the engineering school, and used his connections on the East Coast to start and fund a new Electronics Research Lab (ERL). The beneficiaries of DARPA investment in electronic technology, following following the Korean war and the Soviet Union's launch of the Sputnik, were mainly the MIT on the East Coast and Stanford University on the West Coast.

In 1946 Stanford spun off the Stanford Research Institute (SRI), an industrial research center to use Stanford's high-tech know-how in industrial (mainly military) projects. One of the SRI's very first assignments was to improve and shrink the ENIAC by replacing old electronics with the latest one. SRI pioneered a new business model: a consulting firm with roots in and links with a major university but working for the military and for large corporations.

In 1950 the Stanford Research Institute was hired by Bank of America to design a computer for the automation of cheque processing. In September 1955 the prototype of the ERMA (Electronic Recording Machine Accounting) was ready and the following year the machine was manufactured by General Electric and NRC, one of the earliest transistorized computers. It was the first successful use of computers for a banking application, and even pioneered optical character recognition. For this project General Electric's Patrick Hanratty helped SRI and Bank of America standardize the digits and characters printed on cheques so that they could be easily read by a machine, and this would become a national standard after being adopted in 1958 by the American Bankers Association in 1958.

Another pioneer of the strategy of close interaction between academia and industry that had been so successful during the war was Barney Oldfield of General Electric, another MIT alumnus who had worked on radar technology. In 1950 he established the Advanced Electronics Center at Cornell University and in 1952 he set up the Microwave Laboratory at Stanford. The Cornell lab worked on missile defense systems while the Stanford lab worked on microwave devices. He was also instrumental in forging the partnership with SRI and Bank of America to build the ERMA, and later would be instrumental into starting General Electric's computer business.

Meanwhile, in 1951 his Stanford counterpart Terman had convinced Stanford to open the Stanford Industrial Park, partially to accommodate the rapidly growing Varian (founded by two former Terman students), whose revenues increased more than tenfold during the Korean War, but in general to lease its unused land to high-tech companies: Varian (1953), Eastman Kodak (1953), Hewlett-Packard (also founded by former Terman students), General Electric (1954), Zenith (1956), and Lockheed (1956).

The IPOs (Initial Public Offering) of Varian (1956), Hewlett-Packard (1957) and Ampex (1958) signalled the commercial emergence of the Bay Area's high-tech industry.

In 1954 Sylvania opened an Electronic Defense Lab (EDL) in the south of the Bay Area, in Santa Clara Valley (the future Silicon Valley), directed by Stanford's alumnus Bill Perry. In 1958 IBM shipped its first transistorized 709 (originally built with vacuum tubes) specifically for this project. In 1959 Sylvania was bought by General Telephone to form General Telephone and Electronics (GT&E). By then Sylvania's EDL had become one of largest companies of the Bay Area.

Lockheed too benefited from huge military projects: a submarine-launched ballistic missile (the Polaris, first deployed in 1960), and a satellite to spy on the Soviet Union (the Corona, first launched in 1959). By 1960 Lockheed had become the main high-tech company of the south Bay Area.

The first stored-program computer developed in the San Francisco Bay Area was the CALDIC (California Digital Computer), yet another military project, completed in 1954 at UC Berkeley by Paul Morton employing former ENIAC staff from Pennsylvania and local students (including a young Doug Engelbart and Albert Hoagland). It was one of the first computers to bet on magnetic data storage.

In 1952 IBM opened its first West-Coast laboratory in San Jose. This is the lab that changed the world of data storage and retrieval. Magnetic tape was more efficient than punched cards but it was still a sequential kind of storage: records were stored sequentially and constituted a file. When someone wanted information about the files, most likely the programmers needed to write a sorting program. The output was a printout and sometimes also a new file that kept records in a different sequence. Sorting a big file could take a long time if the internal memory could hold only a few records: the program had to sort out a block of records, then write the sorted records on the tape, sort another block of records, write it on the tape, and so on, and then sort the sorted blocks. In september 1956 this lab unveiled the Random Access Method of Accounting and Control (RAMAC) 305, another vacuum-tube computer and the first to use magnetic-disk storage, invented by Jacob Rabinow at the National Bureau of Standards in 1954 but improved by Berkeley engineers who had worked on the CALDIC project (such as Albert Hoagland). This computer shipped with a processing unit, a card-punch machine, a console (card feed, typewriter, keyboard), a printer and (in 1957) the world's first hard-disk drive. The 350 hard-disk drive, the first commercially available moving-head hard-disk drive, was made of 50 magnetic disks for a total capacity of five megabytes. In 1958 IBM installed the RAMAC at the World's Fair in Belgium, one of the influential "demos" of all time because it reached the general public. The RAMAC was already obsolete when it was introduced (it still used vacuum tubes) but the hard disk would revolutionize the computing world, since it truly allowed for direct access to data instead of the "batch" processing of data that had prevailed since the invention of punched-card equipment.

Some companies of the Bay Area contributed to progress in electronic goods. In 1956 Charles Ginsburg at Ampex, heading a team in Redwood City that included a young Ray Dolby, built the first practical videotape recorder (the VRX-1000), a device that changed the way television programming worked (previously, all programs had been broadcast live, and, obviously, at the same time in all time zones). In theory, recording video is not all too different from recording audio, but the amount of information encoded in video (even just black-and-white video) is orders of magnitude higher than the amount for an audio signal. The VRX-1000 made it possible to record one hour of broadcast-quality video on a reel of magnetic tape.

Friden was still the leading manufacturer of high-end calculators. In 1949 it had introduced a fully-automatic calculator, the Model STW.

In 1957 Friden acquired the Commercial Controls Corporation, manufacturer of the tape-driven Flexowriter electric typewriter that had been used as the output device (the "teleprinter") of the Harvard Mark I and could be attached as well to Friden calculators to produce invoices automatically.

In 1963 Friden would introduce the first fully-transistorized calculator, the model EC-130 designed by Bob Ragen. In 1965 Friden was acquired by Singer.

The Bay Area companies were still small by the standards of the East-Coast conglomerates: in 1956 both General Electric and RCA had revenues of over $700 million, while Varian (the largest of the native Bay Area companies) barely reached $25 million. HP had more employees (901) than Varian but smaller revenues ($20.3 million).

And the defense industry was still the main employer of the Bay Area.

Decline of the Giants

NCR, whose products didn't even use punched cards, too had to acquire computer know-how from outside. In 1953 NCR acquired the Computer Research Corporation (CRC).

Burroughs entered the computer business with the E101 in 1954, an odd desktop machine made by ElectroData, and Burroughs acquired this company in 1956. In 1956 ElectroData made the Datatron (1200 vacuum tubes and 3000 crystal diodes), designed by Ernst Selmer and known as both the ElectroData 205 or Burroughs 205, a machine that used magnetic tapes and offered floating-point arithmetic.

In 1953 Underwood acquired the Electronic Computer Corp, founded by Samuel Lubkin in 1949 in New York, that in 1952 had introduced a computer with a magnetic tape drive and magnetic drum memory, the Elecom 110. One of ECC's engineers was Evelyn Berezin, future inventor of the word-processor.

Magnetic Core Memory

In December 1952 IBM introduced its first electronic digital stored-program computer, dubbed IBM 701, designed by Jerrier Haddad and Nat Rochester at a laboratory north of New York (the Poughkeepsie Engineering Laboratory, set up during World War II) by the team assembled by Ralph Palmer, a former electronic engineer of the Navy who had no experience with the Harvard project but had just introduced IBM's line of electronic calculators (starting with the 604 calculator of 1948 followed in 1949 with the project for the electronic computer that would become the TPM). Palmer had hired Von Neumann as a consultant when VonNeumann was building the same kind of machine at the Institute for Advanced Studies (IAS) in New Jersey. The stored-program 701 was actually born as a "Defense Calculator" when the USA entered the Korean War of 1950. When it was first sold in 1953, it was still meant as a scientific "calculator" (IBM still avoid the term "computer" that referred to human mathematicians), not as a business machine like the Univac: in fact, the first customer was the Los Alamos laboratory and all the other 18 units that were built in the following years went to military research centers. Nonetheless, this was the first computer to be called a "data-processing machine" (a term coined by IBM's marketing executive James Birkenstock). The 701 was superior to the Univac, having Williams tubes for the memory and being four times faster. The IBM 701 came out before Remington Rand's ERA 1103 because the 1103 was a top-secret project until 1953. (Tom Watson famously predicted that only 5 of the 701 computers would be sold and is often misquoted as saying that only 5 computers would be sold in the entire world, which in fact had been said by Aiken in 1948). In 1956 the IBM 701 repeated the 1952 feat of the Univac by correctly predicting the winner of the presidential election.

The more publicized and more expensive Model 702, designed by Charles Bashe, the first computer introduced by IBM that was meant for data processing rather than for scientific calculations (i.e. for business applications like the Univac), also used Williams tubes for memory, but replaced the punched cards with magnetic tapes (hence the internal name of Tape Processing Machine or TPM). The magnetic tape was the real innovation for IBM's customers. While announced in 1953 to compete with the Univac (the first computer to use magnetic tapes), the IBM 702 was first delivered only in 1955 and only 14 units were built. It was Palmer's Poughkeepsie lab that launched IBM in the business of commercial electronic computers.

Gene Amdahl architected the IBM 704 of 1954, another scientific computer that used ideas originally developed for yet another military contract.

In 1950 a government agency, the Navy Bureau of Ordnance, commissioned a superfast computer for the Naval Surface Weapons Center at Dahlgren in Virginia, which IBM named Naval Ordnance Research Computer (NORC). Built between 1950 and december 1954 in collaboration with Eckert at Columbia University (Byron Havens was the chief engineer) this decimal, vacuum-tube, computer, that used Williams tubes for the RAM, was indeed the fastest computer of its time. It also offered floating-point arithmetic. The RAM of the IBM 704 as well as the RAM of the upgraded Univac 1103A, IBM 701 and IBM 702, replaced the Williams-tube storage with a new kind of magnetic storage, the magnetic core invented by Jay Forrester at the MIT. The NORC also introduced an architectural innovation, a rudimentary "operating system": an input-output subsystem that executed while the computation continued, to minimize in-between time. This feature would be transferred to the commercial computer IBM 709 in 1960 and would then become a standard for all computers.

However, IBM's first mass-produced computer, designed at the Endicott laboratory, was the low-cost 650, also introduced in 1954, but it was still decimal, not binary, and had a rotating drum memory (a technology acquired from ERA). Magnetic drums were slow but cheap. IBM ended up selling about 2,000 units of this "Magnetic Drum Calculator" (as it was known internally), a staggering number for those days. The Model 650 was also deployed (at a discounted price) on campuses nation-wide, so that thousands of students got introduced to computers via this machine (that wasn't even called a "computer"). It was also the first computer acquired by the Ames military research center in the Bay Area, originally set up in 1940, which in 1958 became part of a new government agency called National Aeronautics and Space Agency (NASA). It was also the first computer acquired by the Internal Revenue Service (IRS) to catalog and process tax returns, replacing the Friden, Burroughs and NCR calculators and accounting machines (millions of tax returns written on paper had to be punched into cards by an army of women and millions of tax returns were stored in magnetic tapes that could be searched only sequentially). But even this computer hardly made a dent into IBM's revenues. The IBM 407 accounting machine, introduced in 1949, was still infinitely more popular among customers. Electromechanical punched-card accounting machines like the 407 were not only cheaper than an electronic computer, but also more reliable than vacuum-tube machines, and, of course, didn't require the specialized operators required by electronic computers.

In 1955 a group of IBM users met at RAND Corporation in Los Angeles and formed SHARE, the first user group for sharing experiences and know-how about a computer product. The product was the IBM 701. When the 704 came out, this group was instrumental in publicizing it, and even developed its assembler. Univac users responded with the Univac Scientific Exchange, or USE, formed at the end of 1955.

The era of commercial computers had begun with a race between Remington Rand (Univac) and IBM. Univac had acquired the technology from outside. IBM had grown the technology internally.

IBM probably won out (in the long term) because of its involvement in yet another military project, the Semi-Automatic Ground Environment (SAGE). Between 1948 and 1951 an MIT team led by Jay Forrester built the "Whirlwind" computer, the first real-time system and the first computer to use a video display for output (the beginning of computer graphics). Forrester's lab at the MIT had originally been commissioned in 1944 by the US navy to build a flight simulator, the Airplane Stability and Control Analyzer (ASCA): the flight simulator was never built but the project evolved into the Whirlwind after learning of the ENIAC at the seminal Conference on Advanced Computation Techniques organized by Raymond Archibald in 1945 at the MIT and Harvard. For the Whirlwind, Forrester had to invent a faster type of random-access memory because of the real-time requirement: magnetic core memory, first installed on the Whirlwind in 1953. This memory was not volatile, unlike the Williams tubes and the delay lines that required continuous supply of power, and was therefore ideally suited for RAM. Norman Taylor had worked on the Whirlwind from its inception in 1948 and Jack Gilmore had written the assembler program.

In 1949 the Soviet Union exploded its first atomic bomb. This event created fear in the USA of a Soviet nuclear attack. The alarm increased in 1950 when the USA entered the Korean War. Air defense became a strategic goal of the USA. In 1951 the US government created the Lincoln Laboratory within the MIT (but located in the town of Lincoln, more than 100 kms away) specifically to improve the technology of air defense. The Air Force wanted a system to rapidly process the data coming from a network of radars in case of a Soviet attack against the USA. The Lincoln Laboratory was de facto the Cold War version of the old RadLab. The Whirlwind project (that had finally produced a working machine in 1951) was moved to Lincoln and in 1954 a new project, code-named SAGE (Semi-Automatic Ground Environment), was assigned to the Whirlwind team with the goal to create a system for monitoring and intercepting enemy rockets. IBM was in charge of adapting the Whirlwind computer to the task, and the result was the AN/FSQ-7 computer, which (first delivered in 1958) still remains the largest computer ever built: it weighed 275 tons and covered 2,000 square meters of floor space, and contained 55,000 vacuum tubes.

On this project IBM refined its magnetic drum technology as well as its magnetic core technology. This project also expanded Whirlwind's display technology: the operator was even able to point at objects on the 19" CRT screen using an electronic "light gun".

On this screen in 1956 an unknown programmer first drew a human picture, the black and white picture of a pin-up girl, the beginning of computer art.

Another innovation was that the radars would send data to the computer in digital format by modem over telephone lines (a feature developed by AT&T that jumpstarted its digital telecommunications business).

Half of IBM's computer-related revenues of the 1950s came from two military contracts: the guidance computer for the B-52 bomber, and SAGE. SAGE alone accounted for sales of $500 million in the 1950s. At the time the MIT was the single largest recipient of military contracts, and IBM basically parasited on that bonanza. For the record, SAGE, designed to intercept enemy bombers, and extremely influential in the history of computers, became obsolete before being completed because the Soviet Union tested the R-7 Semyorka, the first intercontinental ballistic missile (ICBM), in 1957, thereby making it pointless to defend the USA from bombers.

IBM also successfully leveraged its position as the leading manufacturer of punched-card equipment. Last but not least, in 1956 the young Tom Watson succeeded his father Thomas at the helm of the company and restructured it to become a high-tech company with heavy investment in research (led by newly hired Emanuel Piore). In 1957 IBM started publishing its Journal of Research and Development.

The IBM 610, instead, was a failed experiment: a "small" computer that could fit in a normal office. But it was for this computer that in 1954 John Lentz at Columbia University built the first video terminal (keyboard plus screen) and invented the cursor to keep track of the current position on the screen.

The transition from the early task-oriented computers to the general-purpose IBM and Univac computers had largely been enabled by the development of magnetic-core memory, the first successful and affordable implementation of Random Access Memory. Two groups had contributed to its refinement: Chinese-born physicist An Wang at Harvard's Computation Laboratory under Aiken since 1949 (and at his own Wang Laboratories since 1951) and Jay Forrester's Whirlwind project at the MIT (first installed in 1953). Wang can also be credited for introducing the idea of outsourcing hardware manufacturing to the Far East, as magnetic-core memories were probably the first computer component whose price declined rapidly thanks to cheap labor in the Far East.

Within ten years the computer industry had undergone a dramatic transformation. Initially computers were wartime government projects. Then small companies (mainly working for government agencies) tested the new technology and the market for it. Finally, the large office automation players (Remington Rand, IBM, NCR, Burroughs) entered the field. The electronic megacorporations (General Electric, RCA) were followers, not leaders: their first computers were respectively the ERMA (1956) and the BIZMAC (1956). The latter was a failure (despite being one of the first commercial ones to use core memory), the former was a special-purpose computer designed by the Stanford Research Institute (SRI) and Bank of America to process cheques. They certainly had the capabilities: in 1953 General Electric's Electronics Laboratory in Syracuse built the general-purpose computer OARAC (Office of Air Research Automatic Computer) computer for the Wright-Patterson Air Force Base in Ohio, but never thought of building a commercial version.

In 1957 Barney Oldfield, who had jumpstarted General Electric's labs at Cornell and Stanford, set up an Industrial Computer Department in Arizona to work on a commercial computer, despite the opposition of GE's chief executive officer Ralph Cordiner. That's how General Electric introduced in 1961 the GE-225, a 20-bit computer, designed by Arnold Spielberg (th father of future Hollywood director Steven) and Chuck Prosper.

In 1955 General Electric's revenues were $3 billion, whereas IBM didn't even reach half a billion: General Electric had the know-how, the engineers and the capital to dwarf IBM and Univac in the computer field; but it didn't. AT&T's revenues were even bigger, but AT&T was engulfed in an anti-trust lawsuit and was refraining from entering businesses other than telephony. RCA's revenues were about $1 billion. Thus the top three electronic companies in the world, for one reason or another, failed to enter the computer business. In 1955 IBM passed Remington Rand for number of installed computers and became the world leader in computers.

The computer had been invented by scientists interested in solving complex mathematical problems such as nonlinear differential equations and had found its first practical application in military-related tasks. The first companies to realize the non-military potentiality of the computer were the ones making typewriters, cash registers, adding machines and tabulating machines, not the ones making electronic components.

Jay Forrester's Whirlwind project (and the subsequent SAGE project) gave Boston a huge lead in computer science over the rest of the country. In 1951 the Air Force chose the MIT to create a state-of-the-art laboratory for computer science, the Lincoln Laboratory, which probably became the main center for training computer scientists in the entire world.

Another major center for computer innovation was the University of Illinois at Urbana-Champaign, where in 1951 a team develped the ORDVAC, based on Von Neumann's EDVAC, which became the second computer entirely built within a university (after the Manchester one) although on behalf of the Army, followed in 1952 by its more famous twin, the ILLIAC I (Illinois Automatic Computer), which remained to be used by the university. They boasted 5 kilobytes of memory (or, better, 1024 40-bit words). Incidentally, a member of that team, Saburo Muroga, returned to Japan to build that nation's second computer, the Musashino-1 (1957), following Okazaki Bunji's FUJIC (1956) at Fuji. In 1960 Donald Bitzer at University of Illinois at Urbana-Champaign used this very ILLIAC computer to create PLATO, the first computerized system for learning that inaugurated the field of computer-based education. Despite the limitations of input/output devices, Bitzer realized that graphics was to be crucial to using computers to teach.

The Transistor Industry

An important decision was made by AT&T, the owner of the Bell Labs, to share the technology of the transistor with anyone who could improve it. Jack Morton organized a symposium to disseminate know-how about semiconductors among scientists and engineers from all over the world. The first one, held in september 1951, specifically targeted defense contractors, but the second one, in april 1952, was open to everybody who had purchased the license for the transistor technology. Several electrical companies understood the long-term potential of transistors: Sylvania, a vacuum-tube company based in Massachusetts that had expanded during World War II and that introduced one of the earliest transistors in 1949; Motorola, which opened its Semiconductor Division in Arizona in 1949; Texas Instruments, one of the companies that bought the license in 1952; and, of course, Western Electric itself, AT&T's manufacturing arm, that opened a factory in Pennsylvania (at Laureldale) to make transistors and diodes exclusively for the government. Texas Instruments had originally founded in 1930 by Eugene McDermott as Geophysical Service to conduct petroleum exploration and to manufacture military electronics. In 1946 it had hired an electrical engineer of the navy, Patrick Haggerty, to run a new division in charge of military electronics, and in 1951 it changed name to Texas Instruments.

Another "start-up" that understood the importance of the invention was Transitron Electronics, founded in 1952 near Boston to take advantage of Western Electric's transistor license, by David Bakalar, who had worked at Bell Labs on transistors. Bell Labs was working on a gold-bonded germanium diode and that became Transitron's first product in 1953. Their main customer was the military.

In 1950 Bill Bradley (an alumnus of the MIT Rad Lab) started a project at Philco, the old Philadelphia maker of batteries and radios which at that point was mostly a military contractor, to develop a proprietary transistor and hired a group of young engineers, including MIT graduate Robert Noyce from the MIT (reporting to former Italian artist Carlo Bocciarelli). Their main military contract was for a project underway at the Naval Ordnance Test Station (NOTS) at China Lake in California: the AIM-9 Sidewinder missile, first tested in 1952. In 1953 Philco introduced its "surface-barrier" transistor, the first high-speed, high-frequency transistors.

In 1954 RCA Princeton Laboratories hired German-born physicist Herbert Kroemer from the Central Telecommunications Laboratory of the Deutsche Post and within a few months he published his new invention, the "drift" transistor, the precursos of the double-diffused transistor.

The first applications of transistors had nothing to do with computers. Raytheon became the largest manufacturer of transistors by selling transistors used in hearing aids. What made "transistor" a household name was the first portable radios: the "Dick Tracey" wrist-radio (end of 1953) built by Paul Cooper at Fort Monmouth in New Jersey using Western Electric and RCA transistors; the Regency TR-1 (october 1954), which used Texas Instruments' transistors; and the TR-52 (march 1955) by the Tokyo Telecommunications Company (later renamed Sony), which was followed in 1957 by the TR-63 pocket-size radio. By then the cost of a transistor had been reduced to $2.50 and the Regency (that contained four transistors) was sold for $50.

The portable radio marked the birth of consumer electronics, a trend towards miniaturization and lower prices that would eventually bring ever more powerful appliances in every house and even every pocket.

In 1952 RCA even demonstrated a portable transistor television made with 37 transistors.

In 1955 Motorola introduced a high-power germanium transistor for car radios, and in 1955 Philco developed an all-transistor car radio for Chrysler.

The first all-transistor computer, the TRADIC (TRAnsistor Digital Computer), was built by Jean Felker at Bell Labs for the Air Force in January 1954, but AT&T was barred by the government from commercial computer business.

It was followed in 1955 by the Experimental Transistor Computer at Manchester University (whose prototype had been completed in November 1953), and by the CADET, built at the Atomic Energy Research Establishment in Britain by Edmund Cooke-Yarborough. The TX-0 was designed at the MIT by Wesley Clark (who had joined the Whirlwind project in 1952) and built in 1955 under the direction of a young Ken Olsen, basically a transistorized version of the Whirlwind (replete with CRT display and light-pen and utilizing Philco's faster transistors).

The Electrotechnical Laboratory (ETL) in Japan demonstrated the transistorized Mark III in 1956. In 1957 Burroughs built a transistorized computer for the guidance system of the Atlas intercontinental ballistic missile, the second one after the SOLO (but the existence of Philco's SOLO was kept top-secret). Sperry's ERA built an experimental transistorized computer for the Air Force in 1957, the Transistor Test Computer (TRANSTEC). Philco built the transistorized Solo in 1958 for the NSA. The first fully-transistorized commercial computer in the USA was Philco's Transac, a commercial version of the Solo, introduced at the same time that Elliott Brothers introduced the transistorized 800 in Britain. In 1959 Sylvania built the AN/MYK-1 (MOBIDIC), another military project (later commercialized as the Sylvania 9400). IBM introduced in 1955 the 608 transistor calculator but did not introduce a transistorized stored-program computer until 1960, the 7070 (meant as a replacement for the 650).

The role of the government in nurturing the nascent computer and semicondutor industries was significant and probably crucial. In 1956 both IBM and AT&T settled antitrust suits by accepting to license their technologies to their own competitors. This was just one notable example of how the stringent antitrust policies of that era contributed to the rapid diffusion of intellectual property throughout the computer and semiconductor industries; otherwise the barrier to entry would have been too high for small companies to compete with these giants.

Summary of Early Digital Computers

Zuse 3 (1941): Turing-complete, electromechanical, no stored program, no memory
Colossus (1943): electronic, not Turing-complete, no stored program, no memory
Harvard Mark 1 and IBM ASCC (1944): not Turing-complete, electromechanical, no stored program, decimal, no memory
ENIAC (1946): Turing-complete, electronic, no stored program, decimal (memory: delay lines)
Manchester Baby (1948): Turing-complete, electronic, stored-program (memory: Williams tubes)
Manchester Mark 1 (1949): Turing-complete, electronic, stored-program (memory: Williams tubes and drum memory)
Cambridge EDSAC (1949): ditto (memory: delay lines)
Pilot ACE (1950): ditto (memory: delay lines)
SEAC (1950): ditto (memory: delay lines) with semiconductor logic
SWAC (1950): ditto (memory: Williams tubes)
ERA 1101 and Atlas (1950): ditto (memory: drum memory)
Whirlwind (1951): ditto (memory: magnetic-core memory)
IAS EDVAC (1951): ditto (memory: Williams tubes)
Ferranti (1951): first commercial Turing-complete electronic stored-program computer (memory: Williams tubes and drum memory)
Univac (1951): ditto (memory: delay lines)
IAS (1951): ditto but not commercial (memory: Williams tubes)
IBM 701(1952): ditto, commercial (memory: Williams tubes)
IBM 704 (1954): first mass-produced electronic stored-program computer (memory: magnetic-core memory)
Bell Labs Tradic (1954): fully transistorized
MIT TX-0 (1955): fully transistorized

US Dominance

The story of the electronic computer was now a US story. There was little innovation coming from the rest of the world. This was surprising, given that British scientists such as Faraday and Maxwell had pioneered research on electricity, that Germany had led the world in electrification of cities, factories and homes, and that Britain boasted all the historical milestones in the development of the electronic computer (Alan Turing, the Colossus, the first stored-program computer, the first commercial computer, the first business computer, the first virtual memory, the first transistorized computer). The first British startup in computers was Elliott Brothers, whose first customer was the military. Ferranti introduced the highly influential Pegasus in 1956. The main tabulator companies of Britain, BTM and Powers-Samas, that had virtually no electronic technology, merged in 1959 to form the International Computers and Tabulators (ICT), which in 1962 also acquired Ferranti's computer business. The other computer manufacturers of Britain, Elliot and Leo, formed English Electric. In 1968 these two conglomerates were fused in International Computers Limited (ICL), but this new company owned no valuable technology.

Nazism and World War II had taken a huge toll on Germany: Germany had lost the Jewish scientist to Hitler's madness, and was destroyed and split in two after the war. The occupation government installed by the USA, Britain and France on West Germany de facto forbade German companies from building electronic computers until 1955. It wasn't until 1958 that Zuse's special-purpose Z-22 appeared on the market, and only in 1959 did Standard Elektrik introduce a (fully transistorized) general-purpose computer, followed in 1960 by Siemens' 2002 (also fully transistorized, demonstrated in 1958). Siemens eventually absorbed all other computer manufacturers. The first major computer startup of Germany (and Europe in general) was born in 1968: Nixdorf.

The Soviet Union had a vast population of great mathematicians but there was no link between the military complex and the private sector (that didn't exist anyway) so that the military inventions were never refined by market competition.

Explaining the British decline in computers is harder, but mostly it was probably due to the same arrogance displayed in car manufacturing and other industries: Britain was reluctant to adapt to the standards of other countries (whether right-hand traffic or the metric system), but it rapidly lost the empire and became a small country with a small market behaving like a big country with a big market. Initially, the British computer industry displayed the same dynamics of the US computer industry: government-funded projects in academia. But the transfer of this technology to the private industry did not work well.

France's situation was even worse: its first digital electronic computers appeared only in the late 1950s (SEA's CUBA and Machines Bull's Gamma Extension of 1958) and they were based on obsolete US technology.

The company that was going to become Europe's main computer manufacturer was Italy's Olivetti, that in 1960 introduced the fully-transistorized Elea 9003 and in 1965 would introduce the pioneering desktop computer P101.

Japan never developed an industry of vacuum tubes. Its office-machine industry directly jumped to the integrated circuit. And, yet, Japan succeeded where Europe failed. Its first digital computer was the electromechanical Mark 1 built by the Electrotechnical Laboratory (ETL) and Fujitsu, first demonstrated in 1952 and commercialized in 1954 by Fujitsu as the FACOM 100. The first electronic computers appeared in 1956: Fuji's FUJIC (that used vacuum tubes) and ETL's transistorized Mark III. Japan's NTT (Nippon Telephone and Telegraph) built a clone of the ILLIAC, the M1. But soon the US companies gained control of the Japanese market: Sperry Rand through Oki, Librascope through Mitsubishi, Honeywell through Nippon Electric (NEC), General Electric through Toshiba, etc. The real protagonists were the two companies that shunned this kind of collaboration: IBM was the only US manufacturer to make and sell its own computers in Japan, Fujitsu was the only Japanese manufacturer to make and sell its own computers. When in 1961 the Japanese government introduced protectionist measures, Fujitsu and IBM benefited. The other Japanese companies were totally dependent on US technology. In 1964 the Japanese government decided to foster a national industry of integrated circuits. In 1971 the government also forced Fujitsu and Hitachi to develop IBM-compatible mainframes. In 1974 Fujitsu began manufacturing mainframes for Amdahl, and thus the Japanese finally learned from Amdahl the latest mainframe technology. The combination of government involvement and Amdahl deal lifted the Japanese computer industry Meanwhile, in 1972 NEC, Fujitsu and Hitachi built the machines (the DISP computers) that NTT used to offer the first time-sharing service in Japan, and NTT was leading the development of a Japanese semiconductor industry.