It Began with Babbage (17 page)

An artifact, then, is the product of satisficing choices. When the artifact is then tested “in the field,” its limitations, flaws, and weaknesses will be revealed. There will inevitably be such limitations because satisficing is a matter of compromise. A new cycle of design may then begin with the current design as the starting point. A new aspiration level may be identified and new goals established that correspond to it, and the designer proceeds to a new, hopefully improved, design. And so it goes. Artifacts lead to new artifacts as their “descendants.” Over time, a phylogenetic family of artifacts will appear. Or, over time, one phylogenetic pathway may spawn the beginning of an entirely new one.

Of course, even satisficing is constrained by bounded rationality. The satisficing goals may not be attainable after all; all kinds of cultural factors (the “environment”) may intrude, including unforeseen, unanticipated changes in the cultural environment. In the case of computing machines, new technologies may appear, for example.

Thus it was with the Bell Laboratories series of machines. Thus it was with the computers that followed the Harvard Mark I.

Physical technology, an aspect of computing's cultural environment, played a vital role in this evolution. Electronic components as digital devices—the “offspring” of analog electronic devices, such as vacuum tubes long used in radio engineering—became a rival to relays. The speed at which the vacuum tube could switch states (ON to OFF or vice versa) was much faster than in the relay. However, digital electronic components were still regarded as unreliable by some, and so Aiken, having parted ways with IBM after the Mark I project, embarked, in 1945, on the development of the Mark II using electromagnetic relays.
⁵⁷
The storage capacity of this paper tape-controlled machine was larger than that of the Mark I. The latter had 72 storage registers
⁵⁸
; the former could hold 100 decimal floating-point numbers (see Section II, this chapter).
⁵⁹
Computational power was also greater; there were two adders and four multipliers that could operate simultaneously, and several more input and output devices.

This machine became operational in 1947. Because of its use of relays, it was also faster than the Mark I. Like its predecessor, the Mark II was a decimal machine; however, it used the binary coded decimal notation (described in an early form first by Stibitz in 1940) for representing decimal numbers.

As for IBM, with two of the ASCC/Mark I engineers, Lake and Burfee as members of the design team, the company built the Pluggable Sequence Relay Calculator that, in the proper IBM tradition of its electromechanical tabulating machines, used plugboards to input “programs.” On the other hand, like the Mark II, it used relays. This machine became operational at the end of 1944
⁶⁰
—in fact, the same year the Mark I, although
completed in 1942, also became operational. In 1945, a design team led by Eckert (now director of pure science in IBM as well as director of the Watson Astronomical Computing Laboratory in Columbia University), and Hamilton from the Mark I project, began work on the IBM Selectric Sequence Electronic Calculator. Electronics had arrived at IBM. However, well before the dedication of this machine in 1948, electronic computing had turned up in Philadelphia 2 years earlier. The Holy Grail seemed to be within reach.

World War II played two different roles in the annals of computing. On the one hand, it stimulated the development of automatic computing; on the other, the normal channels of scientific communication so deeply cherished by scientists working in different countries and so profoundly necessary to maintain the openness of the scientific enterprise were, “for reason of state,” blocked—especially between scientists of the Allies and those of Germany.

We have seen the first of these effects in the case of the Bell Laboratories' Ballistic Computer. The Harvard Mark I was also used, initially, by the U.S. Navy.
⁶¹
The Mark II was also intended for the U.S. Navy, although by the time it was completed in 1947, the war was over.
⁶²
The first “copy” of IBM's Pluggable Relay Sequence Calculator was sent, in 1944, to the Aberdeen Proving Ground in Maryland, the place where the U.S. Army conducted its testing of weapons systems. By the end of World War II, the military–scientific complex was firmly established, and the computer was undoubtedly in the thick of this complex.

The other effect of the war meant that work going on in Germany on computing (as in atomic physics) during the war years was unknown in the Allied countries until well after the war ended. Maurice Wilkes (1913–2010), whom we will encounter at length later in this story, tells us in his autobiography that it was not until well after the war that he first heard the name of Konrad Zuse (1910–1995).
⁶³

During the early 1930s, as a civil engineering student at the Technische Hochschule Berlin, Zuse, like so many others we have encountered in this story, was frustrated by the tedium of manual calculation. Like others, he, too, began to think about automatic computation. In 1936, he filed a patent for the design of a paper tape-controlled mechanical binary computer.
⁶⁴
In his patent application, he described a memory device comprised of “cells” that could hold both input numbers entering into a computation and the results of computations. He wrote of a “computational plan,” punched onto a paper tape, that would specify the arithmetic operations to be performed, the locations of the memory cells that contained the input data, and the locations of cells where the computational results would be placed. The computational plan would initiate the necessary operations automatically.
⁶⁵

Zuse's design included the possibility of multiple arithmetic units, memories, tape readers, and punches so that several operations could be performed simultaneously (in present-centered language, the possibility of parallel processing).
⁶⁶
He recognized that certain computational situations might demand data that would be constant across parts of a process, in which case, rather than be input to the machine every time they were needed, they could be specified as part of the computational plan
⁶⁷
(in present-centered terms, these constants are called “literals”).

Zuse recognized that a distinction could be made between a human performing computations and a computing machine. Typical “human habits” could be cast off, and simpler mechanisms, conducive to automatic computation could be used instead. An example was the use of binary arithmetic. Zuse also recognized that scientific and technical computations (in contrast to accounting calculations) may need to deal with numbers ranging from the very small (for example, the coefficient of thermal expansion, e = 0.000012) to the very large (for example, the modulus of elasticity, E = 2,100,000 kg/cm
²
), with both kinds perhaps appearing in the same computation.
⁶⁸

To accommodate such variation, he proposed the use of “semilogarithmic notation”
⁶⁹
—in present-centered language, floating-point representation. Thus, Stibitz was not the
original
inventor of floating-point notation for numbers (see Section II, this chapter), although it is quite likely that the American did not know of Zuse's patent application, which meant that he independently (re)invented the concept.

Zuse's proposal was the basis for his first computer called the Z1, completed in 1938. It was a purely mechanical machine, with a 16-word binary memory, and was the progenitor of his next machine, which combined a mechanical memory with an arithmetic unit made of some 200 electromagnetic relays, and was called the Z2.
⁷⁰
The Z2 was a perforated paper tape-controlled machine that could calculate “certain simple formulae” and demonstrate the “principle of program control,” but it was not a practical computer.
⁷¹

Zuse's further work was interrupted temporarily when he was called up for military service. During that time an associate, Helmut Shreyer (1912–1984), an engineer–inventor, began building an electronic version of the Z1 that, using vacuum tubes (“valves” as Shreyer called them), was able to compute at much higher speeds than relays.
⁷²

The electronic components in this design were a combination of valves and neon tube diodes. Shreyer built a small binary arithmetic unit with about 100 valves. Unfortunately, the unit was destroyed by war damage.
⁷³

A positive outcome of Shreyer's memorandum was that Zuse was released from military duty and given government backing to pursue his computer research.
⁷⁴
The first fruit of this was the Z3—an electromechanical machine built in Berlin between 1939 and 1941, and financed mainly by the German Aeronautical Research Institute. Unfortunately, it was destroyed in an air raid in 1944.
⁷⁵
Zuse does not tell us the extent to which it was actually in productive operation, save to comment laconically: “A series of interesting programs was tested and calculated on the machine.”
⁷⁶
This machine was controlled by commands punched on paper tapes; its two “parallel” arithmetic units used floating-point
binary, 22-bit representation of numbers (a sign bit, a 7-bit exponent, and a 14-bit mantissa).
⁷⁷
The machine not only performed standard arithmetic operations, but also square root extraction and multiplication by common factors such as 2, 1/2, 10, 0.1, and –1. Data were input in decimal form through a keyboard and converted internally to binary, and the reverse was done to produce output results in the form of a “lamp display with four decimal places and a point.”
⁷⁸
A total of 2600 relays were used.

In present-centered language, the Z3 was a single-address machine, with each instruction on the program tape specifying the storage address of a single operand and an operation. Presumably, the second operand was “implied,” an internal register in the arithmetic unit.

The immediate successor to the Z3 was the Z4. Its design, planning, and construction began “immediately” after the completion of the Z3.
⁷⁹
In general architecture, it was “fundamentally” identical to Z3, but with some changes. The word length (in present-centered terminology) was increased to 32 bits; the store was mechanical, and there were “special units for program processing,” although Zuse did not specify what these were. However, he mentioned the addition of conditional branch commands so that, one may presume, these special units included the capacity to execute conditional branches. There were also “various technical improvements.”
⁸⁰
The Z4 was completed in 1945, to the extent that “it could run simple programs.” It was the only one of the wartime Z computers that survived.
⁸¹
It was also the last of Zuse's wartime computers. In 1950, after German recovery and reconstruction, the Z4 was transferred to the Eidgenössiche Technische Hochschule (ETH) Zurich,
⁸²
but the Z series phylogeny continued on with the Z5, built in the 1950s and still a relay machine, and the later development of electronic successors.
⁸³

We will encounter the highly original Zuse later in this chronicle in another context: the development of notation for communicating with the computer—in present-centered language, the development of programming languages.

World War II harbored innumerable secrets. One was the development of an evolutionary series of computers in Bletchley Park, a manor house and estate in what is now the town of Milton Keynes not far from London. Bletchley Park now houses Britain's National Museum of Computing. During the war, it was the site of that country's cryptanalytical center, responsible for decrypting codes and ciphers used by the Axis countries. Computing machines played a critical role in Bletchley's wartime mission. Naturally, the work carried out there was highly classified and remained so long after the war was over.

Like other major technoscientific centers created specifically for the war effort, Bletchley Park was populated by mathematicians, scientists, and engineers, many of whom had either already achieved distinction or would do so in later life. Mathematician Max Newman, whose lectures in Cambridge on Hilbert's problems had been the catalyst for Turing's work
on computability (see
Chapter 4
), was one of the team leaders. There was William T. Tutte (1912–2002), who would achieve great distinction for his contribution to combinatorial mathematics. There was Thomas H. Flowers (1905–1998), an electronics engineer who had conducted research on the use of electronic valves (vacuum tubes, in American parlance) in telephone switching networks almost a decade before he entered Bletchley Park.
⁸⁴
There was the eclectic Donald Michie (1923–2007), a classics scholar who, after the war, would turn into a mammalian geneticist before self-transforming into one of Britain's leading figures in a branch of computer science called
artificial intelligence
. There was the mathematician and statistician Irving J. Good (1916–200), a student of the renowned Cambridge mathematician Godfrey H. Hardy (1877–1947). There was Allan Coombs (1911–1993), an electronics engineer; and the Welshman Charles Wynn-Williams (1903–1979), a physicist whose doctoral research in the Cavendish Laboratory, Cambridge, was supervised by Lord Ernest Rutherford (1871–1937), and who became especially known before the war for his work on electronic instrumentation for use in nuclear physics and radioactivity research. Among his prewar contributions was the invention of the binary counter, which became a standard component in digital systems, including the digital computer. And there was Alan Turing.

As it turned out, Turing was involved in some of the early computer developments at Bletchley, but not the later work that produced the most significant products.
⁸⁵
Turing's contributions lay, rather, in actual cryptanalysis—the analysis and deciphering of codes—and with the Enigma—the generic name for a type of electromechanical encryption machine (invented by a German engineer after World War I) and used by such civilian organizations as banks in peacetime, and by both Allied and Axis intelligence during the war. Turing was concerned with deciphering intercepted code produced by the German Enigma.
⁸⁶