John von Neumann and the Origins of Modern Computing

by William Aspray

John von Neumann (1903-1957) was unquestionably one of the most brilliant scientists of the twentieth century. He made major contributions to quantum mechanics and mathematical physics and in 1943 began a new and all-too-short career in computer science. William Aspray provides the first broad and detailed account of von Neumann's many different contributions to computing. These, Aspray reveals, extended far beyond his well-known work in the design and construction of computer systems to include important scientific applications, the revival of numerical analysis, and the creation of a theory of computing.

Aspray points out that from the beginning von Neumann took a wider and more theoretical view than other computer pioneers. In the now famous EDVAC report of 1945, von Neumann clearly stated the idea of a stored program that resides in the computer's memory along with the data it was to operate on. This stored program computer was described in terms of idealized neurons, highlighting the analogy between the digital computer and the human brain. Aspray describes von Neumann's development during the next decade, and almost entirely alone, of a theory of complicated information processing systems, or automata, and the introduction of themes such as learning, reliability of systems with unreliable components, self-replication, and the importance of memory and storage capacity in biological nervous systems; many of these themes remain at the heart of current investigations in parallel or neurocomputing.

Aspray allows the record to speak for itself. He unravels an intricate sequence of stories generated by von Neumann's work and brings into focus the interplay of personalities centered about von Neumann. He documents the complex interactions of science, the military, and business and shows how progress in applied mathematics was intertwined with that in computers.

Genres: Biography, Nonfiction, History, Computer Science, Mathematics, Science, Computers

394 pages, Hardcover

First published December 7, 1990

About the author

William Aspray is Senior Research Fellow at the Charles Babbage Institute, University of Minnesota, Twin Cities. He earned his BA and MA in mathematics from Wesleyan and a Ph.D. in history of science from the University of Wisconsin, Madison.

Reviews

[Ushan · Rating 3 out of 5]

John von Neumann (1903-1957) was a scientist of the caliber of Albert Einstein, though far less famous. He did pioneering work in the mathematical foundations of quantum mechanics, set theory, game theory and much more. It was von Neumann who came up with the idea of using shock waves from shaped charges to compress plutonium to a critical mass, which proved to work near Alamogordo and over Nagasaki. Manhattan Project spurred von Neumann's interest in computing. Computations for atomic bombs were done by a team of about twenty mostly female employees using mechanical calculators, and by IBM electromechanical punched-card equipment. There was demand for more computing power for such problems as solving the partial differential equations describing shock implosion. In 1944 IBM produced an electronic calculator known as the Harvard Mark I, which von Neumann told the Los Alamos scientists about; calculations were done in parallel on it and on punched-card equipment, but the latter won for a time. In early 1945 von Neumann was involved in the design of EDVAC, an early electronic computer that stored both its instructions and its data in the main memory, and wrote a report on it. This design is now called the von Neumann architecture, while the design where the instructions and the data are stored in separate units, as they were in the Harvard Mark I, is called the Harvard architecture. 67 years later, all modern general-purpose computers share a variation on the von Neumann design. A modern Xeon may have eight control units and sixteen processing units instead of one each, and its innermost cache may be split between that used for instructions and that used for data, but the basic design of an active processing unit reading from and writing to passive memory that holds both instructions and data still stands. Under von Neumann's direction, the Institute of Advanced Study at Princeton built a machine, called the IAS computer, that formally launched in mid-1952, but was informally available for use since early 1951. It had 1,024 40-bit words stored in cathode ray tubes, with secondary memory on a magnetic drum, and 3,000 vacuum tubes; addition took 31 microseconds, and multiplication 620 microseconds; its power consumption was 28kW. The book lists 17 clones of the IAS computer built around the world between 1952 and 1957 including the Soviet BESM, but I am unconvinced that it was a clone: the IAS computer used fixed point arithmetic, and the BESM floating point; the IAS computer had one-address instructions and the BESM three-address. The IAS computer was of course used for the design of hydrogen bombs, but also for almost 50 other problems, from stellar evolution to X-ray diffraction, and even simulation of biological evolution! Now that science had such a powerful tool, numerical analysis needed to be reinvented so it could be used effectively; von Neumann wrote papers on numerical linear algebra and numerical solution of partial differential equations. The Monte Carlo method was invented by Stanislaw Ulam at Los Alamos; von Neumann was an early user, and he invented a pseudorandom number generator for it: square an 8-digit integer, take the middle 8 digits of the square; repeat. Donald Knuth once wrote a paper on von Neumann's first program, which was Merge sort. When George Danzig invented the simplex method for linear programming, von Neumann came up with a different method that allowed the user to estimate the necessary number of steps in advance (strange: the first polynomial-time algorithm for linear programming was invented by Naum Shor and Leonid Khachiyan in the 1970s; how was it different from von Neumann's work?). In addition to all this, von Neumann also invented cellular automata, including a self-reproducing one, and described building a reliable computer from unreliable parts using multiplexing.

On the one hand, this is a book on a rather narrow topic: history of computer science and technology in the 1940s and the 1950s. On the other hand, no reader will fail to be amazed: how could a single scientist, however brilliant, have done so much!

[Krollo · Rating 4 out of 5]

A very scholarly work, rather interesting in parts, though it can pall a bit. Nevertheless a good way to learn about a whole variety of topics, and how this enigmatic chap muscled his way into them.

[Timothy · Rating 2 out of 5]

Too slow. Stopped after a few chapters.

[Dave Schroeder · Rating 1 out of 5]

Why did I buy this?