Issue No. 03 - July-September (2006 vol. 28)
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/MAHC.2006.55
Andrew Hodges , University of Oxford
James W. Cortada , IBM
A.M. Turing (ed. B. Jack Copeland), The Essential Turing, Oxford Univ. Press, 2004, 613 pp., $24.95, ISBN 0198250800.
Few will dispute the value of making Alan Turing's writings more accessible. The selection of an "essential" oeuvre, and the writing of critical annotations, is less straightforward. These 613 pages show the choices of the philosopher B. Jack Copeland. His attention centers on the philosophy of computation and the cognitive sciences, but the scope is wide, including a didactic elucidation of Turing's classic 1936 work, historical Enigma material, and an account of the digital computer's origin.
Oxford University Press bills the result as "the first purchasable book by Turing," but this volume is neither the first, nor the largest, edition of Turing's papers. A four-volume Collected Works of A.M. Turing,1 annotated by distinguished mathematicians and computer scientists, was published between 1992 and 2001. An inattentive reader of The Essential Turing, missing the small print on pages 409, 510, and 581, could remain unaware of it. The reader cannot tell what has been left out because the book does not provide a list of Turing's works. Readers would have been better served by a general critical bibliography of earlier Turing editions (which certainly are open to criticism). This omission is typical—Copeland's approach is energetic, clear, and detailed but has blind spots.
Thus, the book does not include a discussion of the Bayesian inference methods using "weight of evidence" that Turing and I.J. (Jack) Good developed for crypt-analysis. In the Collected Works, this work was well accounted for with papers by Good. Turing's theory was not limited to Enigma (being akin to Claude Shannon's theory of communication) and was central to the story of Bletchley Park as a scientific enterprise. But no discussion of it appears in this edition, even though Copeland says that here "the full story of Turing's involvement with Enigma is told for the first time" (p. 2).
This omission also weakens Copeland's story of the wartime genesis of artificial intelligence. Copeland argues that the serial trial by the Bombe of a million or so rotor positions lies behind Turing's 1948 idea of search as a key concept. But Turing's mechanization of judgment through his statistical methods would make a much more substantial link with the program and techniques of AI.
More generally, the hinterland of mathematics and science is weak in this version of "the ideas which gave birth to the computer age," as the front cover promises. Copeland never mentions complex analysis, the machine Turing built in 1939 to calculate the Riemann zeta-function, nor his 1950 computer work superseding it. Although the author deems AI essential, other advances in digital computation are not. Copeland omits Turing's numerical analysis of matrix inversion and his 1949 "program proof," which anticipates the development of modern computer science.
Of course, Copeland is right to emphasize that in 1945 Turing's central thrust was "to build a brain" (p. 374). However, the word "build" raises the question of why Turing, a Cambridge mathematician, thought himself capable of building anything. The inclusion of more material documenting his knowledge of advanced engineering (for example, the 1944 report on his speech scrambler or his report on the American Bombes) would have given a clearer picture of Turing's technological place in the computer's origin.
The book's cover mentions Turing's "ground-breaking design for an electronic stored-programme computer." Yet, the author omits Turing's ACE report. 2 Various bits are quoted, but the reader is unable to assess Turing's proposals for the scope of the machine, subroutines, floating-point coding, a higher level language, the future of software, possible storage media, and general independence of von Neumann.
The converse question of what von Neumann owed to Turing is addressed from the outset in the preface to Turing's 1936 work. Some readers will criticize Copeland for using the expression "stored programme computer," which only was employed much later, in connection with the 1936 "universal machine." This makes it rather easy to assert that in 1945 von Neumann became the first to outline a "practical version of the universal machine," although the best explicit evidence for the link is still the statement Brian Randell got from Stanley Frankel in the 1970s (p. 22). Copeland also gives a detailed discussion of the origin of the 1948 Manchester computer in response to the Manchester group, who in 1998 marginalized the role of M.H.A. (Max) Newman and Turing.
A general problem in both discussions of origins is that not everyone will agree that Turing's abstract universal machine is the single critical idea in the emergence of the computer (the Manchester school certainly does not). A more subtle problem is that while Turing was clear about the universal machine as the basis for the digital computer, he did not identify the common storage of data and instructions as its crucial feature. Indeed, he wrote that Charles Babbage had "all the essential ideas" (p. 446). Something not brought out by Copeland is that Turing did in practice exploit the modifiability of the stored program immediately in 1946. In his championing of Turing's priority, it is also surprising that Copeland does not mention the vigorous support offered from the fresh perspective and experience of the distinguished logician Martin Davis. 3
Copeland gives detailed attention to Turing's early (1948) neural nets, although strangely Christoph Teuscher's work on implementing them 4 goes unmentioned. Regarding Turing's morphogenesis (which Copeland's preface, at p. 515, rather confuses with genetic algorithms), some readers will be disappointed that only Turing's 1952 published paper is included. However, this area of applied mathematics is obviously remote from Copeland's philosophical center of interest.
Here, Copeland's main contribution, in earlier papers, has been his insistence that the Church-Turing Thesis, on what constitutes effective calculability, was never intended to apply to "machines." Rather, it must be limited to the model of a human being working to rule. It is highly questionable whether Alonzo Church and Turing recognized this strict distinction. Church gave a characterization specifically in terms of machines, 5 and Turing's own formal statement 6 referred simply to "what could be carried out by a machine."
From this dubious premise, Copeland has derived theories about Turing allowing for machines affecting computations that would go beyond Turing computability. 7 Copeland, along with a coauthor in another article, 8 claimed that what Turing imagined was that the uncomputable "oracle" 6 could be physically embodied—for example, by measuring electricity to infinite precision—and that "the search is under way" for a new revolution in computing. Fortunately, Copeland has largely excluded these and various other extraordinary assertions from The Essential Turing.
Further original work of Copeland's in this text consists of squeezing as much as possible from Turing's late semipopular items, with nuances on the "imitation game" and the question of learning machines. Yet, he makes little of his own pertinent observation that Turing really did refer to the possibility of uncomputability in nature—in the reduction process of quantum mechanics. 9 The reason seems to be Copeland's lack of interest in physics. The book doesn't refer to Turing's last year of intense interest in fundamental physics at all. With the advent of quantum computing, the physical basis of computability is of growing importance. Essentially, Alan Turing left this central question for the future.
This selection of Turing's papers, made accessible in paperback form, is a valuable resource. It is not, however, a complete account of Turing's work, and its form is influenced by the editor's particular philosophical agenda.
References and notes
Steven T. Usdin, Engineering Communism: How Two Americans Spied for Stalin and Founded the Soviet Silicon Valley, Yale Univ. Press, 2005, 343 pp., $40.00, ISBN 0-300-10874-5.
Joel Barr and Alfred Sarant were two Americans who became Communists in the 1930s, spied for the Soviet Union in the 1940s as part of the Rosenberg espionage ring, moved to the USSR in the 1950s, and remained there until the end of the Cold War. These men were engineers who worked on engineering projects in the US during World War II and parlayed that experience in the Soviet Union into engineering and later computing projects. The majority of this well-written, interesting book, in which Steven Usdin tells their story, is about spying and daily life in the Soviet Union. It is also an indictment of the FBI's inability to track the whereabouts and work of these two spies over the years, despite repeated attempts to do so.
Barr and Sarant convinced Nikita Khrushchev to build a small research community called Zelenograd to work on computers. The book recounts the establishment and operation of this research facility during the 1960s and 1970s, the work of other computer projects at the Leningrad Design Bureau between 1965 and 1973, and efforts to develop a microchip minifab between 1975 and 1990. What is remarkable, and new information, is that for a brief period two American born and educated engineers led the Soviet Union's primary microelectronics initiatives. Usdin provides brief descriptions of key projects, such as the construction of the UM-I, a computer small enough to fit on an aircraft in the late 1950s and early 1960s, and the larger KB-2.
However, the value of this book's 100 some pages devoted to computing lies in the details provided on how high-tech projects were managed in the Soviet Union, providing insights on political activities, administrative and operational practices, and commentaries regarding individual engineers and policy makers.
The book is based on interviews with Barr and on his personal records, in addition to the growing body of spy literature on the Cold War, but it is more about Cold War spying than a history of Soviet computing. Annals readers will find new material on the development of computing in the USSR, although not the full story because there were multiple work streams underway beginning in the 1950s that involved other groups. The material provided is credible because it was based on diaries kept in the period, and some of the details have been corroborated by other publications beginning to appear on the topic. The book certainly adds to our rapidly growing understanding of Soviet computing.James W. Cortada, , firstname.lastname@example.org