IEEE Annals of the History of Computing (April-June)

102 Pages·2004·2.62 MB·English

Volume 26 Number 2 April–June 2004 http://www.computer.org
• The Chess-Playing Turk
Machine
• Early Radio Broadcasts on
Computing
• Arlington Hall Station—
The Lesser-Known WWII
Cryptoanalytic Center
History in Perspective
IEEE Annals of the History of Computing
Contents
Vol. 26, No. 2
April–June 2004
http://computer.org/annals
2 From the Editor’s Desk
David Alan Grier, Editor in Chief
Features
3 Five 1951 BBC Broadcasts on
Automatic Calculating Machines
Allan Jones
16 Datamation’s Glory Days
Robert V. Head
22 Sydis and the Voice/Data Terminal Craze of 1984
Albert J. Henry and Glenn E. Bugos
34 Recollections of the Philco Transac S-2000
Saul Rosen
48 Computer Sciences at Purdue University—1962 to 2000
John R. Rice and Saul Rosen
62 People, Languages, and Computers: A Short Memoir
Keith Smillie
The Chess-Playing Turk
of Baron Wolfgang von
Kempelen.
Published by the IEEE Computer Society ISSN 1058-6180
Departments
Editor in Chief
David Alan Grier
75 Biographies Senior Consulting Editors
Donald Knuth Thomas J. (Tim) Bergin
Paul Ceruzzi
Robert W Floyd, in Memoriam
Consulting Editors
84 Events & Sightings Janet Abbate
Jennifer Light
David Alan Grier and Dan Campbell
Associate Editors in Chief
“Echec”: The Deusches Museum reconstructs
Anne Fitzpatrick
the Chess-Playing Turk Mary Croarken
Editorial Board
86 Anecdotes
William Aspray, Per Brinch Hansen,
Laurie Robertson, Acting Editor Dan Campbell, Martin Campbell-Kelly,
Alan Clements, James W. Cortada,
Arlington Hall Station: The US Army’s World War II
Nathan Ensmenger, Denise Whitson Gürer,
Cryptoanalytic Center Thomas Haigh, Ulf Hashagen,
Luanne Johnson, Peggy Aldrich Kidwell,
Michael S. Mahoney, Arthur L. Norberg,
90 Reviews
Brian Randell, Laurie Robertson, Raul Rojas,
Raul Rojas, Editor Keith Smillie, Dag Spicer, Christopher Sterling,
James E. Tomayko, Eric A. Weiss
96 Think Piece
Staff Editor: Tammi Titsworth
Greg Downey Magazine Assistant II: Alkenia Winston
Contributing Editor: Louise O’Donald
Jumping Contexts of Space and Time
Executive Director: David Hennage
Publisher: Angela Burgess
Assistant Publisher: Dick Price
Business Development Mgr: Sandy Brown
Sr. Advertising Coordinator: Marian Anderson
Member/Circ. Promo Mgr: Georgann Carter
Computer Society Information, p. 93
Magazine Operations Committee
Bill Schilit (Chair),
Jean Bacon, Pradip Bose,
Doris L. Carver, George Cybenko,
John C. Dill, Frank E. Ferrante,
Robert E. Filman, Forouzan Golshani,
David Alan Grier, Rajesh Gupta,
Warren Harrison, M. Satyanarayanan,
Nigel Shadbolt, Francis Sullivan
Publications Board
Articles appearing in this journal are abstracted and Michael R. Willaims (Chair),
indexed in Historical Abstractsand America: History Michael Blaha, Mark Christensen,
and Life. Sorel Reisman, Jon Rokne,
Bill Schilit, Linda Shafer,
Steven L. Tanimoto, Anand Tripathi
Submissions
See Author Guidelines at
http://computer.org/annals
Circulation: IEEE Annals of the History of Computing(ISSN 1058-6180) is published Copyright and reprint permission: Copyright © 2004 by the Institute of Electrical
quarterly by the IEEE Computer Society, 10662 Los Vaqueros Circle, P.O. Box 3014, Los and Electronic Engineers, Inc. All rights reserved. Abstracting is permitted with
Alamitos, CA 90720-1314, USA; phone [phone]. IEEE Computer Society credit to the source. Libraries are permitted to photocopy beyond the limits of US
Headquarters, 1730 Massachusetts Ave. NW, Washington, D.C. 20036-1992, USA; IEEE copyright law for private use of patrons articles that carry a code at the bottom of
Headquarters, 345 East 47th St., New York, NY 10017, USA. Subscription rates: IEEE the first page, provided the per-copy fee indicated in the code is paid through the
Computer Society members get the lowest rates and choice of media option— Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923, USA. For
$31/25/40 US print/electronic/combination. Go to http://computer.org/subscribe to other copying, reprint, or republication permission, write to Permission Editor,
order and for more information on other subscription prices. Back issues:members, IEEE Annals of History of Computing, 10662 Los Vaqueros Circle, P.O. Box 3014, Los
$10; nonmembers, $20. This magazine is also available in microfiche form. Alamitos, CA 90720-1314, USA.
Postmaster. Send undelivered copies to IEEE Annals of the History of Computing, Editorial: Unless otherwise stated, bylined articles and reviews reflect the author’s
IEEE Service Center, 445 Hoes Lane, Piscataway, NJ 08855, USA. Periodicals postage opinion. Inclusion in IEEE Annals of the History of Computingdoes not necessarily
paid at New York, NY, and at additional mailing offices. Canadian GST Registration constitute endorsement by the IEEE Computer Society. All submissions are subject
No. 125634188. Canada Post Publications Mail Agreement Number 0487880. to editing for style, clarity, and space considerations.
Printed in the United States.
From the Editor’s Desk
David Alan Grier
Editor in Chief
Baron von Kempelen’s Chess-Playing Turk—featured on Turing, von Neumann, or any of the others who con-
the cover of this issue of Annals—was not really a com- tributed to the modern computer.
puter in any sense of the word. In the bluntest language, The story of the Chess-Playing Turk reminds us that
it was a con game, a means of separating fools and their the vision of the computer has always preceded the
money. In his 1970s novel on the chess-playing machine machine itself. We can see what we want to do before we
(King Kill, Random House, 1977), author Tom Gavin por- do it. As many readers of this magazine know, we have
trays von Kempelen as a peer of Mark Twain’s Duke and often announced these visions before we know how to
Dauphin, the sharpsters in Huckleberry Finnwho work the implement them. In the 1960s, computer manufactures
towns that spread along the Mississippi River. The Events announced new models with powerful features long
and Sightings column in this issue reviews the history of before they had designed such machines. In the 1980s,
the Turk and explains how the intelligence behind the the trade press defined the concept of “vaporware”, soft-
Turk had a very human origin. ware that existed as an idea in some designer’s mind but
In a more generous light, the Chess-Playing Turk is an had yet to be written in code. “Where there is no vision,”
example of the aspirations of the Industrial Age. As inven- wrote King Solomon, “the people perish.” The designers
tors began to build more complicated machinery in the of computing machines have had visions from the start
late 18th and early 19th centuries, they pursued goals and these visions have often run far ahead of practice.
that remain part of our life in the 21st century. They In this issue, we see how visions have moved from the
wanted machines that would relieve them of the tedium ethereal plane of ideas to the concrete reality of actual
of “mental labor,” as Charles Babbage described it. They machines and working software. We have
wanted to gather and disseminate information over large
distances. They wanted to build a machine that could • the story of computer science education and research
think or at least mimic the actions of a thinking person. at Purdue;
In King Kill, Gavin suggests that von Kempelen was • a memoir of a pioneer;
intrigued by the idea of building a thinking machine—at • a study of the rise and fall of Sydis;
least enough intrigued to speculate on how such a • a treatment of an early leader in the computer industry
machine might be built. Von Kempelen had real mechan- at Philco; and
ical talent. He was recognized by no less an authority • two pieces on how computers were explained to a larg-
than Ludwig von Beethoven for the design of a er public—the story of how the BBC presented com-
metronome. Had he speculated more effectively on the puting machines in broadcasts and a recollection of
nature of thinking machines, he might be renowned for how the trade journal Datamationhelped shape the
more than swindling. new field of data processing in the 1950s.
In all fairness, it should be noted that he lived in an
age before Charles Babbage, George Boole, and others This last article is particularly touching as its author,
who laid the foundation for the modern computer. Von Mr. Robert Head, died between the time he submitted the
Kempelen may have speculated about machines that article and its publication in this issue of Annals. Not only
would appear nearly two centuries after his time, but he do our visions precede our acts, they can also outlive our
was clearly a figure of his time, and not a foreshadow of deeds.
2 IEEE Annals of the History of Computing Published by the IEEE Computer Society 1058-6180/04/$20.00 © 2004 IEEE
Five 1951 BBC Broadcasts on
Automatic Calculating Machines
Allan Jones
Open University
In May and June 1951, five leading figures of British computing—
Douglas Hartree, Max Newman, Alan Turing, Frederic (“Freddie”)
Williams, and Maurice Wilkes—spoke about their work on BBC radio.
This article examines surviving texts of their broadcasts, and the
speakers’ principal points are summarized through quotations and
commentary. The broadcasts are placed in the context of
contemporary developments in computing and the particular BBC
service on which they were broadcast.
Researchers of Britain’s early postwar history of • ACE (Automatic Computing Engine), at the
computing have known for some time that a National Physical Laboratory, designed by
series of five British Broadcasting Corporation Turing, launched in 1946 and experimen-
(BBC) radio broadcasts under the general title of tally operational in a pilot version in 1950,
“Automatic Calculating Machines” was broad- although not completed until late 1951.6
cast on the BBC’s Third Programme radio serv- • EDSAC (Electronic Delay Storage Automatic
ice in May–June 1951. In these broadcasts, five Computer), at Cambridge University, designed
British pioneers of computing spoke about their by Wilkes, begun in 1947 and operational in
work. In the order of their broadcasts, they were May 1949.7
Douglas Hartree, Max Newman, Alan Turing, • Mark 1 Prototype at Manchester University,
Frederic (“Freddie”) Williams, and Maurice associated with Newman, Williams, and
Wilkes. Apart from Turing’s broadcast, which (from 1948) Turing. Operational from April
has been discussed by B. Jack Copeland1 and 1949 to August 1950, having evolved from
Andrew Hodges,2these broadcasts have received an earlier “Baby” test machine (operational
little attention from historians of computing. June 1948) and replaced in February 1951
No sound recordings of the broadcasts sur- by the Ferranti Mark 1.8
vive, although they all were recorded on
acetate phonograph discs prior to transmission. Table 1 (see p. 4) gives the titles and broad-
However, texts of all five broadcasts survive as cast dates of the talks, and the computers that
BBC transcripts, which were taken from the the speakers were associated with at the time of
recordings shortly after they were made. These the broadcasts.
transcripts are held at the BBC’s Written
Archives Centre in Caversham, near Reading, As Table 1 shows, the Cambridge and
and are the basis for this article. Manchester projects were well represented in
In addition to the existence of the five BBC the five broadcasts. The National Physical
transcripts, three of the speakers’ scripts are Laboratory’s ACE computer project was repre-
known to have survived. These are Turing’s, held sented only indirectly, via Turing, who was no
at the Alan Turing archive at King’s College, longer associated with it when he made his
Cambridge, and those of Wilkes and Newman, broadcast. This machine was not, in any case,
copies of which are held by Wilkes. Turing’s fully completed at the time of these broadcasts.
script has been published,3although curiously
not in the Collected Works of A.M. Turing,4and is The Third Programme
also available on the Word Wide Web.5None of Before discussing the content of the broadcasts,
the other scripts has been published. I should mention the BBC’s Third Programme,
All five of the speakers in this series were, or on which these five talks were transmitted. This
had been, involved with one or more of the idiosyncratic radio service—so unlike almost
three major computing projects in the UK in anything in modern-day broadcasting—occu-
the immediate postwar period: pied an important position in Britain’s intellec-
IEEE Annals of the History of Computing Published by the IEEE Computer Society 1058-6180/04/$20.00 © 2004 IEEE 3
Five 1951 BBC Broadcasts
Table 1. “Automatic Calculating Machines” broadcasts.
Computers associated
Broadcast Repeat with at time
date date Speaker Title of broadcast
5 May 1951 24 June 1951 Douglas Hartree “Automatic Calculating Machines” Cambridge, EDSAC
8 May 1951 26 June 1951 Max Newman “Automatic Calculating Machines” Manchester Mark 1 (Ferranti)
15 May 1951 3 July 1951 Alan Turing “Can Digital Computers Think?” Manchester Mark 1 (Ferranti)
2 June 1951 4 July 1951 Freddie Williams “Automatic Calculating Machines” Manchester Mark 1 (Ferranti)
5 June 1951 10 July 1951 Maurice Wilkes “The Use of Automatic Calculating Machines” Cambridge, EDSAC
tual life, and an appreciation of its philosophy tune in for just the broadcasts that interested
sheds useful light not only on the broadcasts but them or which aroused their curiosity, and then
on the nature and size of the audience that switch off, or listen to another station. As the
would, or could, have heard them. BBC’s historian Asa Briggs noted, “The Third
The Third Programme was a national Programme set out not to meet the wishes of lis-
domestic radio service inaugurated by the BBC teners who would be engaged in continuous lis-
in September 1946 with an avowedly intellec- tening but rather to recruit ‘patrons’.”9
tual and cultural character. Two other national How many patrons the Third Programme
domestic radio services, the Home Service and had at the time of these broadcasts is hard to
the Light Programme, already existed—hence pin down. In the late 1940s, it was claimed to
the name Third Programme. Central to the be between 1.5 million and 2.5 million.9A typ-
activities of the Third Programme were broad- ical audience for any single Third Programme
casts of serious music, literature, and speech. broadcast would naturally have been much
Many leading thinkers of the day were invited smaller than this. In 1949, two years before
to speak on the service, and in general the style these five broadcasts, the audience for a Third
of presentation was for the speaker to deliver a Programme broadcast was estimated to be
scripted talk typically lasting around 30 min- around 90,000, and it appears not to have
utes. (All the talks in this series lasted about 20 grown during the next few years.10Indeed, the
minutes.) Interview-style presentations were percentage of BBC radio listeners tuning in to
unusual, although widely used on the BBC’s the Third Programme was generally 1 percent
Home Service. The fact that the five broadcasts of the total radio audience during the early
were made by the computing pioneers them- 1950s.11 The more popularly oriented Home
selves, rather than by journalists or commen- Service and the Light Programme would typi-
tators, is thus typical of the approach used on cally have audience figures of a few million for
the Third Programme and is what makes them their more popular broadcasts. The Third
particularly interesting as historical sources. Programme was subject, in any case, to techni-
The Third Programme had no regular cal constraints that restricted its coverage to the
timetable of program “slots”—there was no reg- more populous parts of the UK. Reception in
ular time of the week for science broadcasts, many parts of the country was poor, and in
poetry, or anything else. The only way for lis- remote areas nonexistent.
teners to find out about forthcoming broadcasts Though small in absolute terms, the Third
was to consult the program listings published Programme’s audience was nevertheless influ-
daily in newspapers or the BBC’s own weekly ential, as Britain’s academics, artists, and intel-
Radio Times. The five computer broadcasts dis- ligentsia were disproportionately represented
cussed here, therefore, did not form part of a among it. However, professional intellectuals
regular science and technology stream, and were by no means the Third Programme’s only
were not even broadcast at equal intervals or listeners. A 1949 survey reported that 35 per-
regular times. They nevertheless were conceived cent of the audience was working class,
of and presented as a series, and at the end of although it appears that working classwas then
each broadcast there was an announcement of defined more widely, and middle classmore nar-
when the next would take place. rowly, than would now be the case.12
The Third Programme operated only during Radio broadcasts relating to computers,
the evening, and listeners were not expected to cybernetics, and artificial intelligence (as we
spend the whole evening listening to the serv- would now call it) were by no means rare on
ice. Indeed, it was considered undesirable for the BBC in this period. Between 1946 and
them to do so. Rather, listeners were expected to December 1956, there were 24 such broadcasts,
4 IEEE Annals of the History of Computing
not counting repeats. Most of these broadcasts than out of chauvinism. Hartree and Wilkes, in
date from after 1950, and most were on the particular, were happy to pay tribute to these
Third Programme. Speakers in these other American pioneers in their writings.
broadcasts included Norbert Wiener, Colin It cannot be claimed that the broadcasts sig-
Cherry, Wolfe Mays, Frank H. George, and nificantly change our view of the history of
Christopher Strachey.13The extent of this cov- computing. The transcripts of them do never-
erage of computer-related matters is perhaps theless offer a valuable insight into the rela-
surprising given the widespread perception in tionship of the then emerging field of
the UK that press and broadcasting personnel computer technology to the public under-
are biased against (and ignorant of) science and standing of that technology, as revealed
technology. However, computer-related broad- through the mouths of its leading British prac-
casts on the Third Programme were probably a titioners. It is against this background that the
special case owing to the particular interests of broadcasts are most profitably viewed. Through
their producer, to whom I will later return. the broadcasts we get a sense of what the speak-
ers thought was significant in their work, what
The broadcasts might be comprehensible to a nonspecialist
Although much of the Third Programme’s audience, and where developments might lead.
output was broadly educational, the Third Significantly, we also get repeated reassurances
Programme was not part of the BBC’s educa- about where the work was not likely to lead—
tional service. Third Programme broadcasts were toward the “electronic brains” so frequently
therefore not didactic in the usual sense, and invoked in popular journalism of the time.
speakers were encouraged to address the listener Once again, however, Turing was something of
as an equal who just happened not to be con- an exception.
versant with the speaker’s subject. Accordingly, In the space of this article, it is impossible to
none of the speakers in this series pitched his discuss each broadcast in depth. In the follow-
talk at a high technical level, and none adopted ing five sections, therefore, I summarize each
the style of a formal academic lecture where one broadcast through quotations and commen-
might expect a progression of ideas from funda- tary, taking the broadcasts in the order in
mentals to higher-level concepts. In this respect, which they were made. Because the transcripts
the style of these broadcasts was similar to that were made by nonspecialist clerical staff, there
used for other factual talks (not just science are occasionally places where the transcriber
talks) on the Third Programme at that time. In has clearly misinterpreted what the speaker has
general, the speakers confined themselves to said. In my quotations I have corrected such
fairly simple factual accounts of what comput- misinterpretations without comment, and in a
ers were and what they did. As so often, howev- few places I have adjusted the punctuation to
er, Turing was something of an exception. His something more appropriate for a written pres-
presentation, although not of a high technical entation. Occasional interpolations of my own
level, certainly made greater demands on the lis- are enclosed within square brackets.
teners’ comprehension.
With the exception of Williams, the speak- Douglas Hartree
ers said relatively little about the hardware, The first speaker in the series, Douglas
concentrating instead on software concepts Hartree, had broadcast about computers on the
such as programs, data, subroutines, and so on, BBC five years earlier, in December 1946, on
and also touched on the recurring theme of the Home Service. In his earlier broadcast, he
what a program in principle could and could had mainly been concerned with the ENIAC
not do. As regards the prehistory of computing, machine, which he had recently used during a
no speaker referred to wartime code-breaking visit to the US.
activities, although Williams did mention the At the time of his 1951 broadcast, Hartree
importance of wartime radar research for the was Plummer Professor of Mathematical
development of computers. The names of Physics at Cambridge University, although in
Charles Babbage and Lady Lovelace (that is, the immediate prewar period he had been asso-
Ada Byron, mathematician and associate of ciated with developments in analog computing
Babbage) are occasionally invoked as important at Manchester University, particularly the dif-
pioneers, but those of John von Neumann, J. ferential analyzer.14 His inaugural lecture at
Presper Eckert Jr., and John Mauchly are not Cambridge had been titled “Calculating
mentioned at all. Their absence was probably Machines: Recent and Prospective Develop-
more out of consideration for the listener, to ments,” and he had already published various
whom those names would have meant little, writings relating to digital computers, notably
April–June 2004 5
Five 1951 BBC Broadcasts
his account of the ENIAC machine in Nature15 which data was held on counters and instruc-
and in his book Calculating Instruments and tions on punched paper tape that was read as
Machines.16 the calculation proceeded, or the ENIAC, in
In his May 1951 broadcast, Hartree was con- which the “program” was assembled physically
cerned at a basic level with the differences by setting switches and by patching together
between computers and other sorts of machine. processing units via plugboards and cables. The
He described the parts of a computer, the rela- more modern machines not only held instruc-
tionship between data and information in tions and data in the same memory, but made
terms of the computer’s operation, and the no distinction in the way they were held:
tasks computers could be made to do (such as
calculating, playing games, and other appar- But in most of the recent machines there is no
ently human-like activities). He began by distinction between the form used for numbers
emphasizing three salient points about the and for instructions. The distinction between
machines that were to be the subject of this words representing numbers and words repre-
series of talks: that they were automatic, gener- senting instructions lies in the way in which they
al purpose, and digital. Only the first two of are used.
these three points were elucidated:
A consequence of this lack of distinction
By an “automatic machine” is meant one which between data and instructions is the possibili-
can carry out numerical calculations of any ty of self-modifying programs (something on
length without the attention of an operator, which more than one speaker was to com-
once the schedule of operations to be carried out ment):
has been supplied to the machine, in a suitable
form; and by a “general-purpose machine” is This possibility of modifying instructions as the
meant one which can be used for a large calculation proceeds provides the means of
number of different kinds of calculations, by instructing the machine to carry out much of the
supplying it with the appropriate schedules of discrimination and selection between alternative
operating instructions. procedures which a human computer would
exercise in doing the same calculation by pencil
The third of Hartree’s introductory points, the and paper methods.
digital nature of computers, was not expanded
(although Newman enlarged on it in the sec- Hartree raised here the contentious issue of the
ond broadcast). analogy between humans and computers.
The concept of a general-purpose machine From the announcement of the ACE project in
can be traced back to Babbage’s proposed ana- autumn 1946 (the first of the British comput-
lytical engine. Hartree was aware of Babbage’s ing projects to be announced publicly), the
work and mentioned it in passing as represent- press had had a tendency to refer to the new
ing the first conception of a general-purpose computers as “brains,” or “electronic brains.”17
digital machine. He then launched into the Hartree was anxious to correct what he viewed
anatomy of the modern (that is, von as a misapprehension:
Neumann) machine:
But do not jump to the conclusion that [in mod-
An automatic digital calculating machine con- ifying its own program] the machine is thinking
sists of five main parts, an arithmetical unit, a for itself. All these instructions for modifying
store, a control unit, an input unit, and an out- other instructions, and for evaluating and using
put unit. The purpose of the store is to hold the criteria of any discrimination, have to be
information, either numbers or operating thought out and programmed in detail. The
instructions, for as long as they may be required, machine only carries out literally and blindly
in the course of the calculation. In some of the and without thinking, the instructions which
older machines, the store consisted of two dis- the programmer has thought out for it.
tinct parts, one for numbers and one for instruc-
tions. But in most of the more recent machines, Turing, later in the series, took a different
the same store is used both for numbers and for view, as we shall see.
instructions. Martin Campbell-Kelly18 has written that
one of the distinctive features of computing as
By “the older machines” Hartree was refer- done at Cambridge at this time was the empha-
ring to machines such as the Automatic sis on building up a library of commonly used
Sequence Controller at Harvard University, in subroutines. Hartree alluded to this policy:
6 IEEE Annals of the History of Computing
… the machine and the process of providing it way of solving equations, and another a routine
with instructions may be such that groups of for playing bridge. [….] Problems that appear not
operations for standard processes, such as the to be arithmetical at all may often be made so, by
evaluation of square roots and cube roots, can be quite trivial changes in the way they are stated.
programmed once and for all. […] The main
work in the preparation of the calculations for However, the existence of a set of elementary
the machine may then be the programming of a operations is not by itself what gives a com-
master routine consisting mainly of instructions puter its power:
for calling in the subroutines in the proper
sequence. If an automatic computing machine really need-
ed a tape containing 100,000 instructions in
Hartree concluded by mentioning the pos- order to do 100,000 elementary operations,
sibility of computers playing games such as somebody would have to punch the tape; and
chess, but he would not regard this as evidence that “somebody” might be just as usefully
of thinking: employed in doing the 100,000 elementary sums
himself, with a pencil and a piece of paper.
… [playing games] would come very near what,
in ordinary speech, we would call thinking—an This is perhaps debatable. Even if a task with
aspect of those machines on which I understand 100,000 operations required a program tape
Dr. Turing will be speaking. But remember, that containing 100,000 instructions, there might
the sequence of operations for such a process still still be a benefit in creating the tape because
has to be programmed, and Lady Lovelace’s the program could be used many times to
words still apply—“the machine can only do process different sets of data. As far as Newman
what we know how to order it to perform.” was concerned, however, the utility of a com-
puter lay in the fact that a multistep operation
This remark of Lady Lovelace’s is a recurring can be specified in fewer steps than the opera-
theme in the first three broadcasts, with each tion itself would take:
speaker giving a different verdict on its veracity.
The machines that are the subject of this talk […]
Max Newman all have the essential property of being able to do
The mathematician Max Newman had a big job from a few instructions. … [The]
worked on breaking German Enigma–coded arrangements by which this is achieved are the
messages at Bletchley Park during World War II most characteristic feature of these machines,
along with Turing, who had been one of his and are the source of those complexities of
students at Cambridge University before the behaviour that give some colour to comparisons
war. At the time of his 1951 broadcast, with certain mental processes, […].
Newman was professor of mathematics at
Manchester University, and his interest in com- Newman went on to mention the “jump”
puting was mainly with a view to their use as a instruction as one technique for doing “a big
mathematics research tool. Although Newman job from few instructions,” by enabling a
had largely initiated the project to build a com- sequence of operations to be repeated:
puter at Manchester, he had little involvement
with the design of its hardware.19 The normal procedure, when the machine is start-
In his talk, Newman was concerned with ed off, is for the instruction in line 1 to be carried
what made computers so powerful. He located out first; then control passes to line 2, the instruc-
their power in the fact that they used a limited tion in it is carried out, control passes on to line
set of elementary operations, and any sequence 3, and so on. … [There] is a special type of instruc-
of operations could be repeated until a stipu- tion whose function is precisely to interrupt the
lated condition was satisfied. normal succession. For example, Instruction 100
At the start of his talk, he picked up might be “Jump back to instruction 25.”
Hartree’s idea of the general-purpose machine,
which could perform a wide range of tasks Of course, one needs to be able to exit from
despite its relatively small repertoire of ele- the loop created by jumping back to an earlier
mentary operations: instruction:
It is the arrangement of these elementary opera- There must be some way of bringing the repeti-
tions, and the way they are interrelated, that tions of a cycle to an end when they have gone
causes us to call one series [that is, program] a on long enough. […] This is accomplished by
April–June 2004 7
Five 1951 BBC Broadcasts
introducing a conditional or branched type of is, programmer] cannot predict what course the
instruction, for example: “If line 27 is empty (i.e., calculations will take, and it is not obvious that
contains 0) step on to the next instruction in the anyone could discover a routine to obtain the
normal way; but if it is not empty jump to results of such a programme […]. In view of these
instruction 12.” facts it seems that the dictum of Lady Lovelace,
as quoted by Professor Hartree, that “the
Such accounts of conditional jumps have a machine can only do what we know how to
way of being couched in anthropomorphic order it to perform,” needs to be received with
terms, which Newman wished to counter: some reserve. However the end of my talk is not
the place to enter on these fascinating but con-
Now there is some danger here that the jargon of troversial topics.
“obeying instructions” and “choosing alterna-
tives” which has become the customary way of Alan Turing
describing the behaviour of these machines, may By the time of Turing’s broadcast, roughly a
evoke a picture of the machine “conning” [that year had passed since the publication of his
is, reading and memorizing] the branched now famous Mindarticle in which he discussed
instruction, looking to see if line 27 is empty, and the issue of whether computers could be said to
then faithfully choosing the appointed alterna- think.20At the outset of his broadcast, Turing
tive. In fact the machine “obeys” its instructions made it clear where he stood:
in exactly the same sense that a railway train
“obeys” the points [that is, switches], going to Digital computers have often been described as
Crewe if they are set one way and to Macclesfield mechanical brains. Most scientists probably
if the other. regard this description as a mere newspaper
stunt, but some do not. One mathematician has
Presumably, Newman intended his railway expressed the opposite point of view to me rather
analogy to suggest that a program can no more forcefully in the words “It is commonly said that
vary its route during a calculation than can a these machines are not brains, but you and I
train driver during a journey: Once a program know that they are.” […] I shall give most atten-
and its data are read into a computer, the future tion to the view which I hold myself, that it is
course of the data-processing operation is as not altogether unreasonable to describe digital
completely determined as is the route of a computers as brains.
train. However, the analogy is potentially mis-
leading. The course of a train is knowable in Much of the rest of the talk is a summary of
advance, but this is not necessarily true of a Turing’s justification for regarding computers
computing program’s calculations, as Newman potentially as brains, and the kinds of reason
himself said later (quoted below). that people put forward to oppose the sugges-
Like Hartree, Newman saw self-modifying tion that computers might one day be able to
programs as holding an intriguing possibility think. Of these objections, the principal one is
for something close to what we would now call Lady Lovelace’s argument that computers only
artificial intelligence: do what they have been programmed to do.
Turing was careful to make clear that the
The machine will add lines 2 and 3, if instructed computers of his day could not plausibly be
to do so, without the least regard to whether one called brains; his point is that digital comput-
or the other of these lines is to be used later on as ers had the potential for being plausibly regard-
an instruction. This means that we can modify ed as brains. His argument, familiar from the
not only the true numerical material, but also the Mind article, depends on the concept of the
instructions themselves, in the course of the com- universal machine, which he had conceived in
putation. […] This has, with some justification, connection with his celebrated 1936 paper,21
been described as the ability to learn from results. although he did not mention that paper here:
However, whereas Hartree was clear that A digital computer is a universal machine in the
such self-modifying programs could only “lit- sense that it can be made to replace any
erally and blindly” carry out the programmer’s machine of a certain very wide class. It will not
instructions, Newman was less certain: replace a bulldozer or a steam-engine or a tele-
scope, but it will replace any rival design of cal-
It is not difficult to make up programmes of culating machine, that is to say any machine
moderate length leading to networks of opera- into which one can feed data and which will
tions so complex that even the composer [that later print out results.
8 IEEE Annals of the History of Computing