V
computers might be able to learn [90, p.208]. It is in this setting that, for
Oettinger, “Alan Turing and others became familiars at meetings of the Ra-
tio Club” [90, p.208]. Oettinger’s work on the EDSAC led up to his 1952 paper
‘Programming a Digital Computer to Learn’ [86, my emphasis], in which he tar-
geted an audience of psychologists, neuro–physiologists, and everyone concerned
with non-numerical applications of digital computers. He connected the McCul-
loch & Pitts’s paper to automatic digital computers and wrote that “digital
computers can be made to serve as models in the study of the functions and of
the structures of animal nervous systems” [86, p.1243].
9
Due to Turing’s direct
influence, Oettinger used the words “universal digital computer” in his paper [86,
p.1244].
10
But, although Oettinger included both Turing’s 1936 and 1950 papers
in his bibliography, it was Turing’s 1950 “imitation game” that mattered most
to him. Oettinger was, after all, concerned with programming a real computer,
the EDSAC, so that it could “learn”.
After returning to Harvard, Aiken disapproved of his student’s turn towards
learning organs and forced him to return to his automatic language translation
project [90], which would lead up to his 1954 Ph.D. dissertation, A Study for the
Design of an Automatic Dictionary. In the immediately following years, he dived
into the needs of the milk and banking industries [7, p.6]. Oettinger’s papers on
data processing [51, 87] and his comprehensive 1960 book Automatic Language
Translation [88] — which was partially about programming the Univac I for the
purpose of machine translation — did not contain any references to Turing and
the like.
By 1961, Oettinger was referring — through Paul Rosenbloom’s 1950 book
The Elements of Mathematical Logic — to the works of Alonzo Church, Turing,
and especially Emil Post [100, Ch.IV]. As professor of Mathematical Linguis-
tics, Oettinger was giving the bigger picture of the converging developments
that had taken place during the 1950s. “Syntactic analysis”, he said, had re-
ceived considerable attention, not only from the “mathematical linguists” (such
as Noam Chomsky and himself), but also from “applied mathematicians” (such
as Carr and Perlis) and “mathematical logicians”. The mathematical linguists
were seeking algorithms for automatic translation among natural languages. The
applied mathematicians (i.e. the computer programmers) were concerned with
the design and translation of languages suitable for programming machines. The
mathematical logicians, in turn, had been exploring the structure of formal ar-
tificial languages [89, p.104].
By 1963, the theoretical work of Post, Turing and several others had be-
come common currency among some influential academics in both mathemati-
cal linguistics and in computing, particularly in automata theory and automatic
programming. The professional linguist Bar-Hillel had already expressed his ap-
preciation for “recursive function theory, Post canonical systems, and the like”
in 1960 [14, p.84].
11
And in the Classification System of the December 1963 issue
of the Computing Reviews, “Turing Machines” was explicitly mentioned next to
“Automata”.
VI
3
Automatic Programming — a Worm’s Eye View
Besides the machines housed at MIT, Harvard University, and Cambridge Uni-
versity, also the ENIAC, EDVAC, and IAS machines should not go unmentioned.
The ENIAC and EDVAC were constructed at the Moore School of Engineering,
Philadelphia, and were later housed in Maryland at Aberdeen Proving Ground
— where John von Neumann was a consultant for several years [10][24, p.886].
During the second half of the 1940s, von Neumann and his team built the IAS
computer at the Institute for Advanced Study, Princeton [48].
During the early 1950s, Andrew Booth was, together with his wife Kathleen,
writing a book entitled Automatic Digital Calculators (cf [20]). In the preface of
their first edition, published in 1953, the authors explicitly acknowledged “their
indebtedness to John von Neumann and his staff” at Princeton for “the stimu-
lating period spent as the guests of their Project” [20, p.vii]. Reflecting on the
history of computer building, the authors referred to several people, including
Leibniz, Pascal, Babbage, Jacquard, and Hollerith. No mention was made of Tur-
ing throughout the whole book, except for a brief reference to the ACE computer
which had been “under the direction of Womersley, Turing and Colebrook” [20,
p.16].
12
Moreover, in the section “The universal machine”, the authors referred to
“the Analytical Engine” of Babbage and essentially described him as the father
of the universal computer.
13
In this connection, they mentioned the machines
built by Howard Aiken (1937–1944) and at Bell labs (1938–1940, 1944), describ-
ing them as “universal” and “general purpose”. More specifically, the authors
distinguished between a special purpose and a general purpose machine in the
following manner. A “special purpose machine” was constructed to perform one
set of operations, to solve one particular problem. For example, a computer that
can only compute Income Tax (for different sets of input) was special purpose.
General purpose computers, by contrast, were “capable of being set up to solve
any problem amenable to treatment by the rules of arithmetic” [20, p.1]. The
authors also clarified their notion of “universality” by noting that multiplication
and addition can be defined in terms of subtraction; therefore:
Whatever type of computing machine is projected, so long as it is to have
the attribute of universality , it must necessarily have in its structure some
component capable of performing at least the most elementary operation of
arithmetic — subtraction. [20, p.22, my emphasis]
It was due to the “expense” and “difficulty” of incorporating several electrical or
electronic components that “most modern general purpose computers” did not
include square root units, dividers, and — in some cases — even multipliers [20,
p.3].
14
Some of the first computers, like the ENIAC, were — at least initially — pro-
vided with a plugging system “so that the various units [could] be connected
together and sequenced to suit the particular problem to be solved.” [20, p.14].
Some of the later machines of the 1940s and early 1950s, like the EDSAC and the
EDVAC, were based on the principle of a large store containing both numbers and