Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Mechanical intelligence (collected works of A. M. Turing)
Ince D. (ed), North-Holland Publishing Co., Amsterdam, The Netherlands, 1992. Type: Book (9780444880581)
Date Reviewed: Aug 1 1993

Given the modern tendency to compartmentalize disciplines, it is especially refreshing to find such a wide range of topics discussed by a single person. Mechanical intelligence, the second of four volumes of the collected works of Alan Turing, addresses an astonishing array of subjects: computer architecture, artificial intelligence, neural networks, structured programming, and program verification. As this book clearly shows, Turing recognized no compartments. Ince, the editor of this volume, admirably sums up Turing’s universality in his introduction:

Alan Turing’s contribution to computer science was immense; not only in terms of depth, but also in terms of breadth. Today, his name tends to be frequently invoked in philosophical discussions about the nature of artificial intelligence. However, it is often forgotten that he was also a pioneer researcher in the areas of computer architecture and software engineering.

The editor has done an excellent job in compiling this volume, although I did not agree with several of his interpretations or explanations of Turing’s text. In the introduction, Ince nicely places each paper in context, and he has copiously annotated the papers themselves with references to his unobtrusive endnotes. This volume comprises seven papers that range in length from 3 to 86 pages, appearing in the sequence in which they were originally written (between 1945 and 1954.)

In 1945, Turing was invited to join the National Physical Laboratory to design Britain’s counterpart to ENIAC, which had just been built in the United States. Turing’s report on the Automatic Computing Engine (ACE) is the first, and longest, paper in this book. The influence of the Turing machine on the architecture of ACE is clear.  Turing  goes into much detail on the design of the computer, which includes logic diagrams for the main components, the logical control, and the central arithmetic part. He also discusses the technology available to implement the machine, devoting several pages to the mathematics of mercury delay lines to be used for memory.

Of more enduring interest than the technological details is Turing’s presentation of the relationship between hardware and software. Turing, clearly believing that programming was the key to the versatility of the machine, argued for a simple processor (the principle on which RISC machines are based). Complex operations would be broken down into “subsidiary operations,” or subroutines as we now call them, supported by a stack. Turing’s thoughts on programming are revealing:

Instruction tables will have to be made up by mathematicians with computing experience and perhaps a certain puzzle-solving ability.…The process of constructing instruction tables should be very fascinating. There need be no real danger of it ever becoming a drudge, for any processes that are quite mechanical may be turned over to the machine itself.

Today’s compilers and program generators are examples of tools that have turned mechanical processes over to the machine itself. As well as giving a few example programs, or “instruction tables,” Turing contemplates the applicability of ACE. He describes ten problems that ACE (or any other discrete-state computer) would or would not be able to compute. An interesting section on hardware and software checking is evidence that correctness was no small concern for Turing.

The second essay is the text of a lecture Turing gave to the London Mathematical Society in 1947, in which he gave an overview of ACE and elaborated some of his ideas on subroutines. He also introduced the topic of machine learning, which was to become more important in his later work. Indeed, in the next paper, he provides further explanation of the nature and role of machine learning.

It had been some years since I had read the next paper, “Intelligent Machinery,” originally published as a report in 1948 (and subsequently reprinted in AI compendia). What a joy it was to rediscover it! Here is the genesis for “Computing Machinery and Intelligence,” the work for which Turing is most widely known in AI circles. I strongly advocate that this paper be read first, as it explains Turing’s philosophy regarding the nature of mind and the nature of the machine--there is a nice link between logical computing machines (Turing machines) and practical computing machines (digital computers).

The fourth paper is the shortest. “Checking a Large Routine” was presented in 1949. Here, Turing demonstrates a mathematical approach to proving the correctness of a program through a series of assertions. “Computing Machinery and Intelligence,” the paper mentioned above, in which the “Turing Test” was proposed, follows. This paper is so widely known and available that I will not describe it here, but it is worth repeating that I got more out of this paper after having read Turing’s earlier works. For example, I had long wondered exactly why memory availability was so central to Turing’s prediction about the future of computers. The other papers have answered this. For Turing, the size of the memory of the machine is the most significant measure of its power: the larger the memory, the larger the program.

The penultimate essay, “Digital Computers Applied to Games,” is taken from Faster than thought, a book published in 1953. Turing focuses here on several games: chess, draughts (checkers), and nim. In discussing strategies for these games, Turing clearly foresaw many of the developments in computer game-playing strategy during the subsequent decades. He includes brief accounts of early game-playing programs in Britain.

Finally, “Solvable and Unsolvable Problems,” published in Science News in 1954, is a fairly successful attempt to provide an accessible account of the Entscheidungsproblem articulated by Kurt Gödel in 1931. Gödel’s work led to  Turing’s  own treatment of computability, in which he developed the concept of the Turing machine, first published in 1937.

My only regret about this volume is that it does not contain all of Turing’s work. For example, his work on computability, referred to above, appears in a companion volume. Such is the nature of his work, and such was his intellect, that his contributions to one domain cannot be separated from his contributions to other domains without a sense of loss. This interconnectivity is clear even within this book, a volume much greater than the sum of its parts. I am now eager to read the other volumes in this series: Pure mathematics (already published) [1], Mathematical logic, and Morphogenesis.

This book will appeal to anybody who professes an interest in computer science. Its significance is not just historical, as Turing’s work supplies a perspective still relevant today. We have a tendency to believe that advances in technology, such as RISC architectures, program verification, expert systems, and neural networks, are evidence that we have become smarter. These were all part of Alan Turing’s plan almost 50 years ago. From this volume alone, it is clear that “It is not in dispute that A. M. Turing was one of the leading figures in twentieth-century science,” as P. N.  Furbank,  a friend of Turing and executor of his estate, writes in his preface. It is also clear that, had Turing lived beyond his brief span of 42 years, his phenomenal intellect would have continued to shape, influence, and unify the discipline of computer science.

Reviewer:  Paul A. Luker Review #: CR116616
1) Britton, J. (Ed.) Pure mathematics (collected works of A. M. Turing). North-Holland, Amsterdam, 1992.
Bookmark and Share
 
Alan Turing (K.2 ... )
 
 
Bounded-Action Devices (F.1.1 ... )
 
 
Hardware (K.2 ... )
 
 
Mechanical Theorem Proving (F.4.1 ... )
 
 
Theory (K.2 ... )
 
 
Applications And Expert Systems (I.2.1 )
 
  more  
Would you recommend this review?
yes
no
Other reviews under "Alan Turing": Date
Pure mathematics (collected works of A. M. Turing)
Britton J. (ed), North-Holland Publishing Co., Amsterdam, The Netherlands, 1992. Type: Book (9780444880598)
Feb 1 1993
Alan M. Turing: centenary edition
Turing S., Cambridge University Press, New York, NY, 2012.  194, Type: Book (978-1-107020-58-0), Reviews: (1 of 2)
Aug 22 2012
Alan M. Turing: centenary edition
Turing S., Cambridge University Press, New York, NY, 2012.  194, Type: Book (978-1-107020-58-0), Reviews: (2 of 2)
Aug 9 2012
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy