Computing BY Martin Campbell-kelly

origin of The information age began

with the realization that
machines could emulate
the power of minds

I
n the standard story, the computers evolution has been brisk and
short. It starts with the giant machines warehoused in World War
IIera laboratories. Microchips shrink them onto desktops,
Moores Law predicts how powerful they will become, and Micro-
soft capitalizes on the software. Eventually small, inexpensive de-
vices appear that can trade stocks and beam video around the world.
That is one way to approach the history of computing the history
of solid-state electronics in the past 60 years.
But computing existed long before the transistor. Ancient astron-
omers developed ways to predict the motion of the heavenly bodies.
The Greeks deduced the shape and size of Earth. Taxes were
summed; distances mapped. Always, though, computing was a hu-
man pursuit. It was arithmetic, a skill like reading or writing that
helped a person make sense of the world.
The age of computing sprang from the abandonment of this lim-
Key Concepts itation. Adding machines and cash registers came first, but equally
critical was the quest to organize mathematical computations using
The first computers
what we now call programs. The idea of a program first arose in
were peopleindividuals
and teams who would the 1830s, a century before what we traditionally think of as the
tediously compute sums by birth of the computer. Later, the modern electronic computers that
hand to fill in artillery tables. came out of World War II gave rise to the notion of the universal
computera machine capable of any kind of information process-
holly lindeM (photoillustration); Gene Burkhardt (styling)

Inspired by the work of a com-

ing, even including the manipulation of its own programs. These are
puting team in revolutionary
France, Charles Babbage, a the computers that power our world today. Yet even as computer
British mathematician, created technology has matured to the point where it is omnipresent and
the first mechanical device that seemingly limitless, researchers are attempting to use fresh insights
could organize calculations. from the mind, biological systems and quantum physics to build
wholly new types of machines.
The first modern computers
arrived in the 1950s, as
researchers created machines The Difference Engine
that could use the result In 1790, shortly after the start of the French Revolution, Napoleon
of their calculations to alter Bonaparte decided that the republic required a new set of maps to
their operating instructions. establish a fair system of property taxation. He also ordered a switch

62 S c i e n t i f i c A m e r i c a n September 2009
20 09 SCIENTIFIC AMERIC AN, INC.
w w w. S c i e n t i f i c A m e r i c a n . c o m  S CIENTIFIC A MERIC A N 63
20 09 SCIENTIFIC AMERIC AN, INC.
 from the old imperial system of measurements ect. It was grunt work, demanding no special
to the new metric system. To aid the engineers skills above basic numeracy and literacy. In fact,
and mathematicians making the change, the most computers were hairdressers who had lost
French ordinance survey office commissioned a their jobs aristocratic hairstyles being the sort
fresh set of mathematical tables. of thing that could endanger ones neck in revo-
In the 18th century, however, computations lutionary France.
were done by hand. A factory floor of be- The project took about 10 years to complete,
tween 60 and 80 human computers added and but by then the war-torn republic did not have the
subtracted numbers to fill in line after line of the funds necessary to publish the work. The manu-
tables for the surveys Tables du Cadastre proj- script languished in the Acadmie des Sciences

The Difference
engine

GeArS OF CHAnGe: Charles Babbage produced

a functioning prototype of his Difference
engine (left and detail above) in 1832. Al-
though it demonstrated the feasibilty of his
idea, it was too small to be of practical use.
The first full version of a working Difference
engine would not be built until 1991, 159
years later, by the london Science Museum,
which was guided by Babbages detailed
design notes.

Courtesy of the sCienCe MuseuM (Difference Engine);

sCienCe MuseuM/ssPl (inset)

64 Scientific AmericAn September 2009

20 09 SCIENTIFIC AMERIC AN, INC.
 for decades. Then, in 1819, a young British math- not until the 1970s, well into the computer age, TeamWork
ematican named Charles Babbage would view it that scholars studied these papers for the first
on a visit to Paris. Babbage was 28 at the time; time. The Analytical Engine was, as one of those
three years earlier he had been elected to the Roy- scholars remarked, almost like looking at a com-
al Society, the most prominent scientific organi- puter designed on another planet.
zation in Britain. He was also very knowledge-
able about the world of human computers at The Dark Ages
various times he personally supervised the con- Babbages vision, in essence, was digital com-
struction of astronomical and actuarial tables. puting. Like todays devices, such machines
On his return to England, Babbage decided manipulate numbers (or digits) according to a
he would replicate the French project not with set of instructions and produce a precise numer-
human computers but with machinery. England ical result.
The Harvard Observatorys human
at the time was in the throes of the Industrial Yet after Babbages failure, computation en-
computers, seen here circa 1890,
Revolution. Jobs that had been done by human tered what English mathematician L. J. Comrie examined hundreds of thousands
or animal labor were falling to the efficiency of called the Dark Age of digital computing a pe- of photographic plates between
the machine. Babbage saw the power of mecha- riod that lasted into World War II. During this the 1880s and the 1920s,
nization and realized that it could replace not time, machine computation was done primarily classifying stars based on color,
position and brightness.
just muscle but the work of minds. with so-called analog computers. These devices
He proposed the construction of his Calcu- model a system using a mechanical analog. Sup-
lating Engine in 1822 and secured government pose, for example, one wanted to predict the
funding in 1824. For the next decade he im- time of a solar eclipse. To do this digitally, one
mersed himself in the world of manufacturing, would numerically solve Keplers laws of mo-
seeking the best technologies with which to con- tion. Before digital computers, the only practi-
struct his engine. cal way to do this was hand computation by hu-
The year 1832 was Babbages annus mirabi- man computers. (From the 1890s to the 1940s
lis. That year he not only produced a function- the Harvard Observatory employed just such a
ing model of his calculating machine (which he group of all-female computers.) One could also
called the Difference Engine) but also published create an analog computer, a model solar system
his classic Economy of Machinery and Manu- made of gears and shafts that would run time
factures, establishing his reputation as the into the future [see box on next page].
worlds leading industrial economist. He held Before World War II, the most important an-
Saturday evening soirees at his home in Dorset alog computing instrument was the Differential
Street in London, which were attended by the Analyzer, developed by Vannevar Bush at the
front rank of society. At these gatherings the Massachusetts Institute of Technology in 1929.
Courtesy of Curator of Astronomical Photographs at Harvard College Observatory (human computers);

model Difference Engine was placed on display At that time, the U.S. was investing heavily in [The Author]
as a conversation piece. rural electrification, and Bush was investigating
A year later Babbage abandoned the Differ- electrical transmission. Such problems could be
ence Engine for a grander vision that he called encoded in ordinary differential equations, but
the Analytical Engine. Whereas the Difference these were very time-consuming to solve. The
Engine had been limited to the single task of ta- Differential Analyzer allowed for an approxi-
ble making, the Analytical Engine would be ca- mate solution without any numerical process-
pable of any mathematical calculation. Like a ing. The machine was physically quite large it
modern computer, it would have a processor filled a laboratory and was something of a
that performed arithmetic (the mill), memory Rube Goldberg construction of gears and rotat-
to hold numbers (the store), and the ability to ing shafts. To program the machine, research- Martin Campbell-Kelly is a
alter its function via user input, in this case by ers connected the various components of the professor in the department of
computer science at the University
Courtesy of Martin Campbell-Kelly (author)

punched cards. In short, it was a computer con- device using screwdrivers, spanners and lead
of Warwick in England, where he
ceived in Victorian technology. hammers. Though laborious to set up, once specializes in the history of com-
Babbages decision to abandon the unfinished done the apparatus could solve in minutes equa- puting. He is author of Computer:
Difference Engine was not well received, how- tions that would take several days by hand. A A History of the Information
ever, and the government demurred to supply dozen copies of the machine were built in the Machine (along with William
Aspray) and of From Airline Reser-
him with additional funds. Undeterred, he pro- U.S. and England.
vations to Sonic the Hedgehog:
duced thousands of pages of detailed notes and One of these copies belonged to the U.S. Ar- A History of the Software Industry.
machine drawings in the hope that the govern- mys Aberdeen Proving Ground in Maryland, He is editor of The Works of
ment would one day fund construction. It was the facility responsible for readying field weap- Charles Babbage.

w w w. S c i e n t i f i c A m e r i c a n . c o m  S CIENTIFIC A MERIC A N 65
20 09 SCIENTIFIC AMERIC AN, INC.
 ons for deployment. To aim artillery at a target tained about 3,000 entries. Even with the Dif-
of known range, soldiers had to set the vertical ferential Analyzer, the backlog of calculations
and horizontal angles (the elevation and azi- at Aberdeen was mounting.
muth) of the barrel so that the fi red shell would Eighty miles up the road from Aberdeen, the
follow the desired parabolic trajectory soaring Moore School of Electrical Engineering at the
skyward before dropping onto the target. They University of Pennsylvania had its own differ-
selected the angles out of a fi ring table that con- ential analyzer. In the spring of 1942 a 35-year-
tained numerous entries for various target dis- old instructor at the school named John W.
tances and operational conditions. Mauchly had an idea for how to speed up calcu-
Every entry in the firing table required the in- lations: construct an electronic computor
tegration of an ordinary differential equation. [sic] that would use vacuum tubes in place of the
A human computer would take two to three mechanical components. Mauchly, a theoreti-
days to do each calculation by hand. The Dif- cally-minded individual, found his complement
ferential Analyzer, in contrast, would need only in an energetic young researcher at the school
about 20 minutes. named J. Presper (Pres) Eckert, who had al-
ready shown sparks of engineering genius.
Everything Is Change
On December 7, 1941, Japanese forces attacked
the U.S. Navy base at Pearl Harbor. The U.S.
was at war. Mobilization meant the army need-
ed ever more fi ring tables, each of which con-

The Analog Computer

SeeInG STArS: An example of the analog computer is

the planetarium projector, which is designed to
produce a physical analog of the motion of the stars
and planets. The Adler planetarium in Chicago
installed the first American example in 1930 (left).
Although the projectors are not accurate enough
for practical computing, the planetarium still
thrives. The latest projectors can be seen at new
york Citys Hayden planetarium (above).

Courtesy of d. finnin American Museum of Natural History (inset)

Courtesy of the adler PlanetariuM (Zeiss projector);

66 Scientific AmericAn
20 09 SCIENTIFIC AMERIC AN, INC.
September 2009
 The Digital Computer

A year after Mauchly made his original pro-

posal, following various accidental and bureau-
cratic delays, it found its way to Lieutenant Her-
man Goldstine, a 30-year-old Ph.D. in mathe-
matics from the University of Chicago who was
the technical liaison officer between Aberdeen
and the Moore School. Within days Goldstine
got the go-ahead for the project. Construction
of the ENIAC for Electronic Numerical Inte-
grator and Computer began on April 9, 1943.
It was Eckerts 23rd birthday.
Many engineers had serious doubts about
whether the ENIAC would ever be successful.
Conventional wisdom held that the life of a vac-
uum tube was about 3,000 hours, and the
ENIACs initial design called for 5,000 tubes.
At that failure rate, the machine would not
function for more than a few minutes before a
broken tube put it out of action. Eckert, howev- matical prodigy who tore through his educa- power on: Computing entered
er, understood that the tubes tended to fail un- tion. By 23 he had become the youngest ever the electronic age with the
der the stress of being turned on or off. He knew privatdozent (the approximate equivalent of an ENIAC, invented by J. Presper
it was for that reason radio stations never turned associate professor) at the University of Berlin. Eckert and John W. Mauchly of
the Moore School of Electrical
off their transmission tubes. If tubes were oper- In 1930 he emigrated to the U.S., where he
Engineering at the University of
ated significantly below their rated voltage, they joined Albert Einstein and Kurt Gdel as one of
Pennsylvania. The ENIAC used
would last longer still. (The total number of first faculty members of the Institute for Ad- vacuum tubes to hold numbers
tubes would grow to 18,000 by the time the ma- vanced Study in Princeton, N.J. He became a in storage and consumed 150
chine was complete.) naturalized U.S. citizen in 1937. kilowatts of power, equivalent
Eckert and his team completed the ENIAC in Von Neumann quickly recognized the power to more than 1,000 modern PCs.
two and a half years. The finished machine was of electronic computation, and in the several
an engineering tour de force, a 30-ton behe- months after his visit to Aberdeen, he joined in
moth that consumed 150 kilowatts of power. meetings with Eckert, Mauchly, Goldstine and
The machine could perform 5,000 additions per Arthur Burks another Moore School instruc-
second and compute a trajectory in less time tor to hammer out the design of a successor
than a shell took to reach its target. It was also machine, the Electronic Discrete Variable Auto-
a prime example of the role that serendipity of- matic Computer, or EDVAC.
ten plays in invention: although the Moore The EDVAC was a huge improvement over
School was not then a leading computing re- the ENIAC. Von Neumann introduced the ideas
search facility, it happened to be in the right lo- and nomenclature of Warren McCullough and
cation at the right time with the right people. Walter Pitts, neuroscientists who had developed
Yet the ENIAC was finished in 1945, too late a theory of the logical operations of the brain
to help in the war effort. It was also limited in (this is where we get the term computer mem-
its capabilities. It could store only up to 20 num- ory). Like von Neumann, McCullough and
bers at a time. Programming the machine took Pitts had been influenced by theoretical studies
days and required manipulating a patchwork of in the late 1930s by British mathematician Alan
cables that resembled the inside of a busy tele- Turing, who established that a simple machine
phone exchange. Moreover, the ENIAC was de- can be used to execute a huge variety of complex The Analytical
signed to solve ordinary differential equations.
Some challenges notably, the calculations re-
tasks. There was a collective shift in perception
around this time from the computer as a math-
Engine was
quired for the Manhattan Project required the ematical instrument to a universal information- almost like
solution of partial differential equations.
John von Neumann was a consultant to the
processing machine.
Von Neumann thought of the machine as
looking at
Manhattan Project when he learned of the having five core parts: Memory held not just nu- a computer
Bettmann/CORBIS

ENIAC on a visit to Aberdeen in the summer of

1944. Born in 1903 into a wealthy Hungarian
merical data but also the instructions for opera-
tion. An arithmetic unit performed calcula-
designed on
banking family, von Neumann was a mathe- tions. An input organ enabled the transfer of another planet.
w w w. S c i e n t i f i c A m e r i c a n . c o m  S CIENTIFIC A MERIC A N 67
20 09 SCIENTIFIC AMERIC AN, INC.
THe FUTUre programs and data into memory, and an output er originally designed for mathematical calcu-
organ recorded the results of computation. Fi- lations turned out to be infinitely adaptable to
oF CompUTer nally, a control unit coordinated operations. different uses, from business data processing to
arCHITeCTUre This layout, or architecture, makes it possi- personal computing to the construction of a
The stored-program computer ble to change the computers program without global information network.
has formed the basis of comput- altering the physical structure of the machine. We can think of computer development as
ing technology since the 1950s.
What may come next?
Moreover, a program could manipulate its own
instructions. This feature would not only enable
QuAnTuM: The much touted quan- von Neumann to solve his partial differential
tum computer exploits the ability of a equations, it would confer a powerful flexibility
particle to be in many states at once. that forms the very heart of computer science.
Quantum computations operate on In June 1945 von Neumann wrote his classic
all these states simultaneously.
First Draft of a Report on the EDVAC on behalf
neurAl neT: These systems are
of the group. In spite of its unfinished status, it
formed from many simple processing was rapidly circulated among the computing co-
nodes that connect to one another in gnoscenti with two consequences. First, there
unique ways. The system as a whole never was a second draft. Second, von Neu-
exhibits complex global behavior. mann ended up with most of the credit.

lIVInG: Computers based on strands

of DNA or RNA process data encoded Machine Evolution
in genetic material. The subsequent 60-year diffusion of the com-
puter within society is a long story that has to
be told in another place. Perhaps the single most
remarkable development was that the comput-

CHAnGInG prOGrAMS: The first practical stored-

The Stored-program Computer program computer was the eDSAC, built at the
university of Cambridge by Maurice Wilkes and
William renwick in 1949 (below). early at-
tempts to make a symbolic programming sys-
tem (above) were a breakthrough in simplifying
programming.

PoPPerfoto/Getty iMaGes (EDSAC); sourCe: the edsaC siMulator ProGraM doCuMentation,

By the dePartMent of CoMPuter sCienCe, university of WarWiCk (inset)

68 Scientific AmericAn
20 09 SCIENTIFIC AMERIC AN, INC.
September 2009
 having taken place along three vectors hard-
ware, software and architecture. The improve-
ments in hardware over the past 60 years are
legendary. Bulky electronic tubes gave way in
the late 1950s to discrete transistors that is,
single transistors individually soldered into
place. In the mid-1960s microcircuits contained
several transistors then hundreds of transis-
tors, then thousands of transistors on a silicon
chip. The microprocessor, developed in the
early 1970s, held a complete computer process-
ing unit on a chip. The microprocessor gave rise
to the PC and now controls devices ranging
from sprinkler systems to ballistic missiles.
The challenges of software were more subtle.
In 1947 and 1948 von Neumann and Goldstine
produced a series of reports called Planning and
Childs play: Simple program-
Coding Problems for an Electronic Computing 1945. The situation mirrors that of the gasoline-
ming languages such as Basic
Instrument. In these reports they set down doz- powered automobile the years have seen many
allowed the power of program-
ens of routines for mathematical computation technical refinements and efficiency improve-
ming to spread to the masses.
with the expectation that some lowly coder ments in both, but the basic design is largely the A young Paul Allen (seated) and
would be able to convert them into working same. And although it might be possible to de- his friend Bill Gates worked on
programs. It was not to be. The process of writ- sign a radically better device, both have achieved a Teletype terminal attached by
ing programs and getting them to work was ex- what historians of technology call closure. In- a phone line to a mainframe
cruciatingly difficult. The first to make this dis- vestments over the decades have produced such computer that filled a room.
covery was Maurice Wilkes, the University of excellent gains that no one has had a compelling
Cambridge computer scientist who had created reason to invest in an alternative [see Internal-
EDSAC, the first practical stored-program com- Combustion Engine, on page 97].
puter [see box on opposite page]. In his Mem- Yet there are multiple possibilities for radical
oirs, Wilkes ruefully recalled the moment in evolution. In the 1980s interest ran high in so-
1949 when the realization came over me with called massively parallel machines, which con-
full force that a good part of the remainder of tained thousands of computing elements oper-
my life was going to be spent in finding errors in ating simultaneously. This basic architecture is
my own programs. still used for computationally intensive tasks
He and others at Cambridge developed a such as weather forecasting and atomic weap-
method of writing computer instructions in a ons research. Computer scientists have also
symbolic form that made the whole job easier looked to the human brain for inspiration. We
and less error prone. The computer would take now know that the brain contains specialized
this symbolic language and then convert it into processing centers for different tasks, such as
binary. IBM introduced the programming lan- face recognition or speech understanding. Sci-
guage Fortran in 1957, which greatly simplified entists are harnessing some of these ideas in
the writing of scientific and mathematical pro- neural networks for applications such as li- More To
grams. At Dartmouth College in 1964, educa- cense plate identification and iris recognition. Explore
tor John G. Kemeny and computer scientist More blue sky research is focused on build- The Difference Engine: Charles
Thomas E. Kurtz invented Basic, a simple but ing computers from living matter such as DNA Babbage and the Quest to Build
mighty programming language intended to de- [see Bringing DNA Computers to Life, by the First Computer. Doron Swade.
mocratize computing and bring it to the entire Ehud Shapiro and Yaakov Benenson; Scientif- Penguin, 2002.
undergraduate population. With Basic even ic American, May 2006] and computers that
Computer: A History of the Infor-
Courtesy of the Lakeside School

schoolkids the young Bill Gates among them harness the weirdness of the quantum world mation Machine. Martin Campbell-
could begin to write their own programs. [see The Limits of Quantum Computers, by Kelly and William Aspray. Westview
In contrast, computer architecture that is, Scott Aaronson; Scientific American, March Press, 2004.
the logical arrangement of subsystems that 2008]. No one knows what the computers of 50
The Modern History of Comput-
make up a computer has barely evolved. Near- years hence will look like. Perhaps their abilities
ing. Stanford Encyclopedia of Philoso-
ly every machine in use today shares its basic ar- will surpass even the powers of the minds that phy. http://plato.stanford.edu/
chitecture with the stored-program computer of created them. entries/computing-history

w w w. S c i e n t i f i c A m e r i c a n . c o m  S CIENTIFIC A MERIC A N 69
20 09 SCIENTIFIC AMERIC AN, INC.