The Architecture Machine

This is a portion of the eBook at doi:10.7551/mitpress/8269.001.
0001
000 Introduction
Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282286/9780262368063_f000100.pdf by Xin Sun on 09 November 2024

This is a portion of the eBook at doi:10.7551/mitpress/8269.001.0001
Humanism through Intelligent Machines context or regardless of context. It follows that

a mechanism must recognize and understand
. . . so much corn, so much cloth, so much the context before carrying out an operation.
everything, that things will be practically with- Therefore, a machine must be able to discern
out price. There will be no poverty. All work will changes in meaning brought about by changes
be done by living machines. Everybody will be in context, hence, be intelligent (A. Johnson,
free from worry and liberated from the degrada- 1969c). And to do this, it must have a sophisti-
tion of labor. Everybody will live only to perfect cated set of sensors, effectors, and processors
himself. to view the real world directly and indirectly.
Karel Capek, Rossum’s Universal Robots
Intelligence is a behavior. It implies the capac-
Computer-aided design cannot occur without ity to add to, delete from, and use stored infor-
machine intelligence— and would be dangerous mation. What makes this behavior unique and
without it. In our era, however, most people particularly difficult to emulate in machines is
have serious misgivings about the feasibility its extreme dependence on context: time,
and more importantly, the desirability of at- locality, culture, mood, and so forth. For ex-
tributing the actions of a machine to intelligent ample, the meaning of a literary metaphor is
behavior. These people generally distrust the conveyed through context; assessment of such
concept of machines that approach (and thus meaning is an intelligent act. A metaphor in a
why not pass?) our own human intelligence. In novel characterizes the time and culture in
our culture an intelligent machine is immediate- which it was written.
ly assumed to be a bad machine. As soon as
intelligence is ascribed to the artificial, some One test for machine intelligence, though not
people believe that the artifact will become necessarily machine maturity, wisdom, or
evil and strip us of our humanistic values. Or, knowledge, is the machine’s ability to appreci-
like the great gazelle and the water buffalo, we ate a joke. The punch line of a joke is an about
will be placed on reserves to be pampered by a face in context; as humans we exhibit an intel-
ruling class of automata. ligence by tracing back through the previous
metaphors, and we derive pleasure from the
Why ask a machine to learn, to understand, to new and surprising meanings brought on by the
associate courses with goals, to be self- shift in context. People of different cultures
improving, to be ethical— in short, to be have difficulty understanding each other’s
intelligent? jokes.
The answer is the underlying postulate of an Some architects might propose that machines
architecture machine. A design machine must cannot design unless they can think, cannot
have an artificial intelligence because any think unless they want, and cannot want unless
design procedure, set of rules, or truism is they have bodies; and, since they do not have
tenuous, if not subversive, when used out of bodies, they therefore cannot want; thus cannot
1
The Spanish colonials laid

out entire cities with enough
megalomania to accommo-
date expansion for many
centuries. These cities were
usually designed by small
bands of soldiers whose de-
sign skill was limited to a
book of rules. Accordingly,
irrespective of context,
giant grids were decreed as
a result of “global goals”
such as riot control and reli-
gious prominence.
The two illustrations are of

LaPaz, Bolivia. The top pho-
tograph shows the central
city, which still conforms to
the original scheme. The
bottom photograph shows
expansion to the north. It is
interesting to note that this
growth beyond the Spanish
colonial plan has forced a
“pebble-oriented” architec-
ture. This is caused by two
shifts in context: one of time
and one of terrain.

think, thus cannot design: quod erat demon- the unique and the exceptional. It would con-
strandum. This argument, however, is usually centrate on the particulars, “ for particulars, as
emotional rather than logical. Nonetheless, the everyone knows, make for virtue and happiness;
reader must recognize, if he is an “ artificial generalities are intellectually necessary evils”
intelligence” enthusiast, that intelligent ma- (Huxley, 1939). Human designers cannot do
chines do not exist today and that theories of this; they cannot accommodate the particular,
machine intelligence at this time can at best be instead they accommodate the general. “ He
substantiated with such an example as a com- (the architect) is forced to proceed in this way
puter playing a superb game of checkers because the effectuation of planning requires
(Samuel, 1967) and a good game of chess rules of general applicability and because
(Greenblatt, et al., 1967). Furthermore, archi- watching each sparrow is too troublesome for
tecture, unlike a game of checkers with fixed any but God” (Harris, 1967a).
rules and a fixed number of pieces, and much
like a joke, determined by context, is the cro- Consider a beach formed of millions of pebbles;
quet game in Alice in Wonderland, where the each has a specific color, shape, and texture.
Queen of Hearts (society, technology, econom- A discrete pebble could have characteristics,
ics) keeps changing the rules. for example, black, sharp, hard. At the same
time the beach might be generally described as
In the past when only humans were involved in beige, rolling, soft. Humans learn particulars
the design process, the absence of resolute and remember generalities, study the specific
rules was not critical. Being an adaptable and act on the general, and in this case the
species, we have been able to treat each prob- general conflicts with the particular. The prob-
lem as a new situation, a new context. But lem is therefore twofold: first, architects cannot
machines at this point in time are not very handle large-scale problems (the beaches) for
adaptable and are prone to encourage repeti- they are too complex; second, architects ignore
tion in process and repetition in product. The small-scale problems (the pebbles) for they are
result is often embodied in a simple procedure too particular and individual. Architects do not
that is computerized, used over and over, and appear to be well trained to look at the whole
then proves to be immaterial, irrelevant, and urban scene; nor are they apparently skilled
undesirable. at observing the needs of the particular, the
family, the individual. As a result “ less than 5
Ironically, though it is now difficult for a ma- percent of the housing built in the United States
chine to have adaptable methods, machines can and less than 1 percent of the urban environ-
be employed in a manner that treats pieces of ment is exposed to the skills of the design pro-
information individually and in detail. Imagine fessions” (Eberhard, 1968b).
a machine that can respond to local situations
(a family that moves, a residence that is ex- But architects do handle “ building-size” prob-
panded, an income that decreases). It could lems, a kind of concern that too often competes
report on and concern itself specifically with with general goals and at the same time couches
3
2 4

1The diagram is a meta- life, now that the serial, re personal needs in antihuman structures. The
phor. The many little forces pititious, and generalized result is an urban monumentalism that, through
are not summed or aver- aspects of the industrial
aged, rather they are con- revolution can be supersed- default, we have had foisted upon us by opulent,
stantly and individually af- ed. (Photograph courtesy of self-important institutions (that can at least
fecting a single body. It is Gabinetto Fotografico Na- control large chunks of the beach); our period
this multitude of forces, zionale, Rome, Italy)
causes, and effects that the is a period of neo-Hancockism and post-
machine can so readily han- Prudentialism. The cause is the distinct maneu-
dle as individual events in a verability gap that exists between the scale of
particular context.
the mass and the scale of the individual, the
2 Handling design problems scale of the city and the scale of the room.
solely at the building scale
can provide a monumental
ism by ignoring all the local Because of this, an environmental humanism
forces. Of course, Brasilia might only be attainable in cooperation with
works, but only as a symbol- machines that have been thought to be inhuman
ic statement of power and
not as a place to live and devices but in fact are devices that can respond
work. It is the result of glob- intelligently to the tiny, individual, constantly
al and general (and perhaps changing bits of information that reflect the
unethical) goals housed at
the wrong scale. identity of each urbanite as well as the coher-
ence of the city. These devices need the adapta-
3 Mojacar in the province of bility of humans and the specificity of present-
Almeria, Spain. This is an
example of local forces day machines. They must recognize general
shaping the environment. shifts in context as well as particular changes
The unity, which results in need and desire.
from more global causes,
comes from the limitation of
materials, resources, weath- The following chapters have a “ pebble-preju-
er, and so on. (The photo- dice.” Most computer-oriented tasks today are
graph first appeared in Ar-
chitecture without Archi- the opposite: the efficient transportation system,
tects [Rudofsky, 1964]. the public open space, the flow of goods and
Photograph courtesy of money. Our bias toward localized information
José Ortiz Echagüe)
implies two directions for the proposed rela-
4Italian hill towns. “The tionship between designer and machine. The
very thought that modern first is a “ do-it-yourselfism,” where, as in
man could live in anachron-
istic communities like these the Marshall McLuhan (1965) automation cir-
[Positano, Italy] would cuit, consumer becomes producer and dweller
seem absurd were it not that becomes designer. Machines located in homes
they are increasingly be-
coming refuges for city could permit each resident to project and
dwellers” (Rudofsky, overlay his architectural needs upon the chang-
1964). The unmentioned ing framework of the city. The same machine
amenities are in fact attain-
able in high-density urban might report the number of shopping days
5
1 Trick automaton feigning

to write, draw, and calcu-
late, made by Leon Joly (cir-
ca 1855). (Illustration cour-
tesy of Editions du Griffon,
Neuchâtel, Switzerland)
2 The computer at home is

not a fanciful concept. As
the cost of computation low-
ers, the computer utility will
become a consumer item,
and every child should have
one. (Cartoon from January
13, 1968, issue of Business
Week. Courtesy of George
Price)

before Christmas as well as alert the inhabitant done, what can be done, and what might
to potential transformations of his habitat. be done are all fuzzy. Our interest is simply to
preface and to encourage a machine intelli-
The second direction presupposes the architect gence that stimulates a design for the good life
to be the prime interpreter between physical and will allow for a full set of self-improving
form and human needs. The machine’s role in methods. We are talking about a symbiosis that
this case is to exhibit alternatives, discern in- is a cohabitation of two intelligent species.
compatibilities, make suggestions, and oversee
the urban rights of individuals. In the nature of
a public service the architect-machine partner-
ship would perform, to the utmost of each actor’s
respective design intelligence, the perpetual
iteration between form and criteria. The two
directions are not exclusive; their joint enter-
prise is actually one.
What needs to be articulated, regardless of the

format of the man-machine relationship, is the
goal of humanism through machines. The
question is not one of rationalism versus vitality
(Juenger, 1949), nor the degree of rationalism
(Ellen Berkeley, 1968), nor the castration of
spirit by technique (Mumford, 1967). The con-
cern is to avoid dehumanizing a process whose
aim is definitely humanization. It is simply untrue
that “ unpleasant as it may be to contemplate,
what probably will come to be valued is that
which the computer can cope with—that is,
only certain kinds of solutions to social prob-
lems” (Michael, 1963). We will attempt to dis-
prove the pessimism of such comments. To do
this, we will ask machines not only to problem
solve but also to problem-worry (S. Anderson,
1966).
In this book, there is no distinction between

hardware and software, between special-
purpose computers and general-purpose
computers. The lines between what has been
7
This is a section of doi:10.7551/mitpress/8269.001.0001
The Architecture Machine

Toward a More Human Environment
By: Nicholas Negroponte
Citation:
The Architecture Machine: Toward a More Human Environment
DOI: 10.7551/mitpress/8269.001.0001
ISBN (electronic): 9780262368063
Publisher: The MIT Press
Published: 1973
The open access edition of this book was made possible by generous funding and support from The National
Endowment for the Humanities/Andrew W. Mellon Foundation Humanities Open Book Program.

MIT Press Direct
Copyright© 1970 by
The Massachusetts Institute
of Technology
Printed and bound

in the United States
of America.
All rights reserved. No part

of this book may be repro-
duced in any form or by any
means, electronic or me-
chanical, including photo-
copying, recording, or by
any information storage and
retrieval system, without
permission in writing from
the publisher.
Open access edition funded by

the National Endowment for the
Humanities/Andrew W. Mellon
Foundation Humanities Open
Book Program.
The text of this book is licensed

under a Creative Commons Attri-
butionNonCommercialNoDeriva-
tives 4.0 International License:
https://creativecommons.org/li-
censes/byncnd/4.0/
ISBN 0 262 14008 X (hardcover)

ISBN 0 262 64010 4 (paperback)
Library of Congress catalog

card number: 75-95283

001 Architect-Machine
Symbiosis
Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282287/9780262368063_c000000.pdf by Xin Sun on 09 November 2024

Prelude to an Architect-Machine Dialogue is the symbiosis. It evolves through mutual

training, in this case, through the dialogue.
Something essential to man’s creativity, even
in science, may disappear when the defiantly Such man-machine dialogue has no historical
metaphoric language of poetry gives way precedent. The present antagonistic mismatch
completely to the denatured language of the between man and machine, however, has gen-
computer. erated a great deal of preoccupation for it.
Lewis Mumford, The Myth of the Machine In less than a decade the term “ man-machine
communication” has passed from concept to
You are in a foreign country, do not know the cliché to platitude. Nevertheless, the theory is
language, and are in desperate need of help. important and straightforward: in order to
At first your hand movements and facial expres- have a cooperative interaction between a
sions carry most of your meaning to the silent designer of a certain expertise and a machine
observer. Your behavior uses a language of of some scholarship, the two must be congenial
gestures and strange utterances to communi- and must share the labor of establishing a com-
cate your purpose. The puzzled listener mon language. A designer, when addressing a
searches for bits of content he can understand machine, must not be forced to resort to
and link to his own language. You react to his machine-oriented codes. And in spite of compu-
reactions, and a language of pantomime begins tational efficiency, a paradigm for fruitful con-
to unfold. This new language has evolved from versations must be machines that can speak
the mutual effort to communicate. Returning to and respond to a natural language.
the same person a second time, let us say with a
new need, the roots of a dialogue already exist. With direct, fluid, and natural man-machine
This second conversation might be gibberish to discourse, two former barriers between archi-
a third party brought into the exchange at this tects and computing machines would be re-
time. moved. First, the designers, using computer-
aided design hardware, would not have to be
A designer-to-machine introduction should have specialists. With natural communication, the
a similar linguistic evolution. Each should track “ this is what I want to do” and “ can you do it”
the other’s design maneuvers, evoking a rhetor- gap could be bridged. The design task would no
ic that cannot be anticipated. “ What was mere longer be described to a “ knobs and dials”
noise and disorder or distraction before, be- person to be executed in his secret vernacular.
comes pattern and sense; information has Instead, with simple negotiations, the job would
been metabolized out of noise” (Brodey and be formulated and executed in the designer’s
Lindgren, 1967). The event is circular inasmuch own idiom. As a result, a vibrant stream of
as the designer-machine unity provokes a ideas could be directly channeled from the
dialogue and the dialogue promotes a stronger designer to the machine and back.
designer-machine unity. This progressively inti-
mate association of the two dissimilar species The second obstruction overcome by such close
9

This photograph first ap- communion is the potential for reevaluating the
peared in Edward Steich- procedures themselves. In a direct dialogue
en's The Family of Man.
(Photograph courtesy of
the designer can exercise his proverbial capri-
Peter Moeschlin) ciousness. At first a designer may have only a
meager understanding of his specific problem
and thus require machine tolerance and com-
patibility in his search for the consistency
among criteria and form and method, between
intent and purpose. The progression from vis-
ceral to intellectual can be articulated in subse-
quent provisional statements of detail and
moment-to-moment réévaluations of the meth-
ods themselves.
But, the tête-à-tête must be even more direct

and fluid; it is gestures, smiles, and frowns that
turn a conversation into a dialogue. “ Most
Americans are only dimly aware of this silent
language even though they use it everyday.
They are not conscious of the elaborate pattern-
ing of behavior which prescribes our handling
of time, our spatial relationships, our attitudes
towards work, play, and learning” (Hall, 1959).
In an intimate human-to-human dialogue, hand
waving often carries as much meaning as text.
Manner carries cultural information: the Arabs
use their noses, the Japanese nod their heads.
Customarily, in man-machine communication
studies, such silent languages are ignored and
frequently are referred to as “ noise.” But such
silent languages are not noise; a dialogue is
composed of “ whole body involvement—with
hands, eyes, mouth, facial expressions— using
many channels simultaneously, but rhythmized
into a harmoniously simple exchange” (Brodey
and Lindgren, 1968).
Imagine a machine that can follow your design

methodology and at the same time discern and
11

The sequence of photo- assimilate your conversational idiosyncrasies.

graphs is taken from the
16mm film, Three Experi-
This same machine, after observing your be-
ments in Architecture Ma- havior, could build a predictive model of your
chines, first shown at the conversational performance. Such a machine
Environmental Design Re-
search Association Confer-
could then reinforce the dialogue by using the
ence, Chapel Hill, North predictive model to respond to you in a manner
Carolina, June 1969. The that is in rhythm with your personal behavior
prints are cropped from ev-
ery fourth frame of a four- and conversational idiosyncrasies.
second scene. In these few
seconds the user of this ter- What this means is that the dialogue we are
minal has said more to the
machine in hand-movement proposing would be so personal that you would
language than in any string not be able to use someone else’s machine, and
of text, but it is all unheard. he would not understand yours. In fact, neither
This particular person has
never used a machine be- machine would be able to talk directly to the
fore; he does not know what other. The dialogue would be so intimate— even
a language is without ges- exclusive—that only mutual persuasion and
tures.
compromise would bring about ideas, ideas
unrealizable by either conversant alone. No
doubt, in such a symbiosis it would not be
solely the human designer who would decide
when the machine is relevant.
The overlaying of a specific design character

upon a generalized machine is not fanciful;
subsequent chapters will illustrate some primi-
tive attempts. An anonymous machine, after
identifying a speaker, can transform itself into
an exclusive apparatus that indeed would re-
flect previous encounters with that speaker. The
extent of the metamorphosis depends on the
degree of acquaintance. At the onset of the
partnership, the machine gathers gross features;
later it avails itself of subtleties. The design
dialogue is one of mutual development.
One might argue that we are proposing the

creation of a design machine that is an exten-
sion of, and in the image of, a designer who, as
he stands, has already enough error and fault.
13

The three photographs However, we have indicated that the maturation

were taken by George De- would be a reciprocal ripening of ideas and
Vincent and first appeared
in the Institute of Electrical ways. At first, jobs where the man is particularly
and Electronics Engineers inept would stimulate a nontrivial need for
Spectrum, September 1967. cooperation. Subsequently each interlocutor
The illustrations describe a
succession of interfaces would avoid situations notably clumsy for his
from hard to soft, from inan- constitution, while prying into issues that were
imate to animate. In most originally outside the scope of concern (or the
machines the interface
where the flesh hits the concern of his profession). Eventually, a sep-
steel (as Warren Brodey aration of the parts could not happen: “ The en-
would say) is no more sub- tire ‘symbiotic’ system is an artificial intelli-
tle and no more advanced
than that of an old typewrit- gence that cannot be partitioned” (Pask, 1964).
er. (Photographs courtesy
of the Environmental Ecolo- In the prelude to an architect-machine dialogue
gy Laboratory, Boston, Mas-
sachusetts) the solidarity of the alliance will rely on the
ease of communication, the ability to ventilate
one’s concerns in a natural vernacular, and the
presence of modes of communication respon-
sive to the discipline at hand. A wine taster
would expect his partner to have taste buds and
an understanding of vintages. An architect
would expect his associate to have at least a
graphic ability capable of manipulating and
displaying a host of environmental data and, in
particular, physical form.
15
Writing machine made

by M. F. Weisendanger.
This device was actually
built in 1946. When the
mechanism worked, the
amateur mechanician add-
ed, “People would be aston-
ished to see a man of our
time sacrifice so much lei-
sure and so many hours to
such a useless piece
of work.”
The device was built after

studying the complete pa-
pers describing the Jaquet-
Droz Writer, built in 1774.
(Photograph courtesy of
Editions du Griffon, Neu-
châtel, Switzerland)

Natural and Not-So-Natural Computer Graphics There exist two families of graphic mecha-
Man’s prolific need for graphic expression can nisms: those devices used to “ input” information
be seen in telephone booths, subway stations, to the machine and those for the machine to
and public men’s rooms. More constructively, “ output” information to the designer. One par-
graphic media have been indigenous to archi- ticular output mechanism of prime importance
tects. Traditional applications range from the is the cathode-ray tube, a televisionlike display
thumbnail sketch to the rendering to the work- device. An electron beam, positioned by the
ing drawing. In general, the conveniences of computer, sweeps across the face of the scope
two-dimensional graphic representation have (in an “ on” or “ off” state) to draw a picture by
warranted overcoming the technical difficulties exciting tiny phosphors that glow for about a
of describing three-dimensional events; conse- twentieth of a second. Once traced, the image
quently, mechanical drawing has become the is regenerated and continually redrawn on the
“ Latin” of all architecture students. face of the screen until a change in content im-
poses a recalculation of the beam’s path. This
Now machines can do mechanical drawing too. regeneration is costly because, in order to de-
So-called computer graphics has popularized liver the illusion of a still image, it must occur
the architect-machine dialogue by affording a between twenty and forty times per second,
natural language—the picture—where the de- depending on the complexity of the picture.
signer can talk to the machine graphically and
the machine can graphically respond in turn. The cathode-ray tube’s most common input de-
This congenial technique is surely a natural vice is the light pen. Rather than squirt out
way for architects to express their thoughts light, this stylus is a sensing device that can
and is certainly in vogue. In the past few years, discern the light of the electron beam. With this
however, it has so dramatically overstated itself instrument the designer can either detect lines,
that the “ message” has indeed become domi- points, or characters, or he can drag about a
nated by the “ medium.” spot of light, a tracking cross, to draw lines. At
present it is not much like a pencil; it is a blunt
Computer graphics is not a synonym for pointer and to write with it is like applying a
computer-aided design. The significance of crayon to a postcard. The picture is small, the
graphic interaction can be no greater than the lines are thick, and the complexity of the dis-
meaningfulness of the content in the transac- played image is limited. Nonetheless, at present
tion. No matter how fancy and sophisticated it is one of the more acceptable vehicles for
the computer graphics system, it is only a glori- research and does allow the necessary, real-
fied blackboard or piece of paper (even though time graphic intercourse.
possibly three dimensional), that is, until it
overtly “ talks back” and actually participates The awkwardness of display devices such as
in the dialogue. Nonetheless, let us isolate the cathode-ray tube goes beyond clumsiness.
computer graphics for a moment and look at it For example, one original acclaim in computer
as a medium of communication. graphics was that “ crooked lines are automat-
17
1 3 4

1 Computervision’s 5 The IBM 2250. (Photo- ically turned into straight ones” (and if properly
INTERACT-G RAPHIC. graph courtesy of the IBM programmed, can even make them perfectly
Presently under develop- Corporation)
ment, this terminal com-
horizontal or vertical to the nearest millionth of
bines several low-cost facil- 6 The Adage display unit. an inch). Unfortunately, “ instant accuracy” is
ities into one configuration not always desirable. In a design dialogue the
that will allow a high level of
interaction. The unit is de- wobbliness of lines often expresses the degree
signed as a transition be- of clarity of architectural thought. The embodi-
tween present methods and ment of an idea should reveal and be congruous
future computer graphics.
With this device the opera- with the stage of the design. One does not
tor can even use his own sketch with a 6H pencil and a straightedge or
pencil. make working drawings freehand with a felt pen.
2 Computer Displays’ Ad-
The refinement of a project is a step-by-step
vanced Remote Display process of sharpening both the comprehension
Station (ARDS). This three- and representation of one’s image of the prob-
faced configuration was
designed for the Depart-
lem. A straight-line “ sketch” on a cathode-ray
ment of Architecture at tube could trigger an aura of completeness in-
M.l.T. Each screen is a jurious to the dialogue as well as antagonistic
storage tube, a device that
will retain an image on the
to the design.
face of the scope without
retracing with the electron The clumsiness of computer graphics hardware
beam. The scope does not
allow dynamic displays
is surrounded with technical difficulties, and,
(rotation, translation, etc.) even when tackled, its resolution will not yield
and does not allow erasing the same textural feeling as graphite on paper.
parts of a picture without
recreating the whole im-
Computer displays will force a new doodle ver-
age. However, the unit re- nacular if they are to capture those original
quires very little computing ideas that usually reside on the backs of enve-
in communication and
costs less than 10 percent
lopes. Displays will have to allow for hazy nego-
of an IBM 2250. tiations to be sloppily expressed. In the mean-
time the important work of Timothy Johnson
3 The Stanford Research In-
stitute terminal used in the
(1963) satifies the research need for a
Augmented Human Intellect “ sketchpad.”
Research Center. The scope
is a commercial (875 line)
television monitor. Beyond the antisketch nature of our present
computer sketch pads, there is a second awk-
4 A mouse, used on both the wardness. Traditionally, the architect has drawn
Stanford Research Institute
terminal and the ARDS. This
plans, sections, elevations— two-dimensional
mechanism is an input de- representations— to describe graphically to
vice, a cheap device ($400), himself and others his three-dimensional vision
and a clumsy device.
of an architectural solution. From the two-
19
2 3

1 The Rolls Royce of dis- but the operator is. (Draw- dimensional documents, a three-dimensional
plays, the IBM Cambridge ing courtesy of IBM Systems
Scientific Center’s 2250,
representation, a physical model or perspective
Journal)
model 4, with Sylvania tab- drawing, can be extrapolated. More recently the
let. This configuration has a design process has been inverted in that we
small computer (an IBM
1130) devoted to maintain-
sketch with study models of clay, cardboard,
ing the graphics. The Syl styrofoam, or little wooden blocks. (Unfortunate-
vania tablet has been added ly, the gestalt of the forms generated by these
to give both a smoother and
a more simple way of draw-
three-dimensional study models unconsciously
ing “into” the computer. implies the form of the final solution.) In the
The tablet is transparent as later stages of design, sections are derived from
well as sensitive to the third
dimension, in that it can
the model in order to study or represent aspects
recognize three discrete concealed by, or unrepresentable in, the physi-
pen distances away from its cal model.
surface (up to about one
inch). The tablet can be
used on the face of the In computer graphics, unlike the traditional
screen (thus coincident with trends and more like contemporary methods, a
the displayed lines) as well
as horizontally, off to the model always exists. Regardless of how it is
side. stored within the machine, a description of the
physical form must reside in the memory. From
2 Drawing by Morse
Payne of The Architects this internal description the machine can pro-
Collaborative made on the duce a section at any point, innumerable plans,
IBM Cambridge Scientific and unlimited perspectives. Though it affords
Center’s 2250 and subse-
quently plotted on a Cal- prolific two-dimensional output, this internal
comp plotter. This drawing model becomes an imposition on the dialogue.
displays a sketchiness that For example, when drawing a section every
is most often absent in com-
puter displays. It is com-
point must have a clearly identified depth, or
posed of tiny lines whose else the designer must draw in several or-
end points are stored in the thogonal views simultaneously. Furthermore,
1130’s memory. Note that, at
about the shoulder and foot,
the designer must explicitly tag surfaces and
the 1130 ran out of memory volumes. At their present stage of development
locations and was unable to computer graphics systems demand an a priori
display the complete draw-
ing. knowledge of whether the designer is working
with lines, planes, or volumes, because each
3 The typical mechani- requires a different reception.
cal engineering format of
top, front, and side view
used in Timothy Johnson’s In computer graphics systems the architect is
SKETCHPAD III. By drawing obliged to work in a predetermined mode (usu-
in several views the ma-
chine is never confused as ally volumetric) which employs predefined ele-
to where the lines belong, ments whose proportions and scale may be
21
manipulated. Such a system was developed by Computer-Aided versus Computerized

Lavette Teague (1968) when at M.l.T. Teague’s “ Computerized” operations are too often mis-
system— BUILD—allows the multiple juxtaposi- named “ computer-aided.” The computerized/
tion of parallelepipeds. Spaces are described computer-aided distinction is too often con-
by volumes and are attached to each other by fused with, or solely embodied in, the mode of
complete or partial surface-to-surface connec- machine usage.
tions. In this case the topology of the shapes is
kept constant, and the proportions are manipu- The traditional (for the past 20 years) mode of
lated. The systems try to offer comprehensive, computer usage, “ batch processing,” entails a
architectural computer graphics. It does not computation center to which a user delivers a
provide for a dialogue. It is computerized. “ program” (a deck of cards, magnetic tape,
paper tape) to be “ run.” Then several hours or
days later the user returns to receive his “ out-
put.” More recently, a new mode, “ time-
sharing,” allows terminals (usually teletypes) in
the office or at home. The terminals are con-
nected to a large central machine (and thus
interconnected with each other) by standard
telephone lines. This system of remote and
multiple machine access permits many physi-
cally separated users to share one large ma-
chine at the same time. The rapid swapping of
users’ programs in and out of the central ma-
chine provides each user with the illusion of a
dedicated machine and permits him continual
use of his terminal. This mode of operation is a
form of “ on-line” usage.
It is commonly suggested that by furnishing a

time-sharing system the on-line nature of the
interaction in itself is a dialogue and transforms
computerized procedures into computer-aided
ones. This is simply not true. For example, let
us suppose you desire the average apartment-
to-parking-space distance for some design
project. In a batch-processing mode (assuming
the program exists) you supply as data the
description of your design, and the average dis-
tance returns hours later, indeed a computer-

ized procedure. On the other hand, in a real- Unfortunately, the present time-sharing phil-
time environment you have a teletype terminal, osophy fosters a cause-and-effect conversation.
the project description resides in the machine, Time-sharing assumes that a designer’s explicit
and you simply type in the apartment-to- manipulations will occupy between one and ten
parking-distance command. But just because percent of any sitting; the remaining time rep-
the answer comes back in three seconds rather resents his deliberations and distractions. Each
than three days, computerized does not become user’s moments of contemplation are in effect
computer-aided. It simply becomes more con- another user’s instants of computation. A
venient “ computerizedness.” Computer-aided- designer can interrupt his own program, but a
ness demands a dialogue; events cannot be routine cannot easily interrupt its partner in
merely a fast-time manifestation of causes and thought. In order to leave the computational
effects. utility available for other users, each routine
resides in the machine only when explicitly
On-line communication therefore is not a suffi- called into service by its particular user. In
cient (though necessary) condition for a com- other words, the routine (the user’s machine)
puter-aided environment. Computer-aided can listen but cannot interrupt.
design requires at least three additional fea-
tures: (1) mutual interruptability for man and for To retain the assets of time-sharing, avoid the
machine, (2) local and dedicated computing anathema of batch-processing, and acquire
power within the terminal, and (3) a machine mutual interruptability, we adjust the allocation
intelligence. of computing power. We transfer some of the
information-processing power and transfer a
Interruptability gives a dimension of interaction certain manipulative and storage capacity to
that allows the process, as well as the product, the terminal that was originally a teletype trans-
to be manipulated. In a computer-aided system, mission and reception device. This semiautono
the machine may interrupt the user and present mous terminal (possibly portable) is a small
the unsolicited information, for example, that computer that would be a “ machine in resi-
the cost of his low-income housing project is dence.” An architecture machine would be such
fifty-eight dollars per square foot. The architect a machine. The designer would speak directly
might welcome the remark, ignore it, or take to this satellite machine. In turn, this small,
offense and request that such interludes of remote computer would interactively converse
finance be restricted. However, regardless of with larger parent machines. (Sending work out
the designer’s response, the apparent high cost to a central mechanism would be automatic
might have overlooked substantial indirect and exclusive of the designer; the recourse
savings not accounted for in the original es- would be for reasons of speed or memory or
timating routine. In this case the designer could information or all three.)
tamper with the estimating procedure and in-
corporate hitherto neglected parameters. The machine at the location of the designer
would undergo the personalization. It would be
23
1 3

1 Leon Groisser at home in composed of additive and subtractive pieces of

his garden.
hardware as determined by the discipline of its
2 The author at home. partner. This local aggregation of parts would
perform the dialoguing, the evolving, and the
3 Computers at home are al
ready being used in an in-
interrupting. Observe that the interrupting and
formal manner. the reinterrupting would depend on the nature
of the designer’s activities, on the context of
4 Architecture students us-
ing the time-sharing system
his efforts. Through familiarity with a specific
CP/CMS. Since 1965, all designer’s idiosyncrasies, the appropriateness
M.l.T. architecture students of the machine’s interruptions would be suitably
have been required to take
at least one semester of
reinforced by context—the inception of an
computer programming as a intelligent act.
prerequisite to the Bachelor
of Architecture degree.
Most of them have had the
A mechanical partner, as we have suggested,
good fortune to learn on a must have intelligence. Customarily, computer-
time-sharing system. The aided design studies and intelligent automata
advantage is obvious: on a
console, a student can take
studies have been antipodal efforts “ between
high risks and can play. mechanically extended man and artificial intel-
This is what learning is all ligence” (Licklider, 1960). On the one hand, in
about.
the context of computer-aided design we are
told to render unto each their respective design
functions and talents: man thinks and the ma-
chine calculates. On the other hand, in the
context of automata studies we are told that
“ Anything you can do, a machine can do
better.”
The two outlooks are not necessarily contradic-

tory. For the present discussion there is a real
issue whether machine intelligence can be in-
dependent of human intelligence. In computer-
aided design only the combination of mechani-
cal amplification and mechanical imitation will
validate the dialogue. The dialogue will evolve
an intelligence, this intelligence will stimulate
a more profound dialogue, which in turn will
promote further intelligence, and so on. Further-
more, the concurrence of “ extended designer”
and “ artificial designer” will force a design
25
redundancy and an overlapping of tasks that Adaptable Machines, Sensory Machines, and
are necessary for the understanding of intricate Parent Machines
design couplings. Perpetual cross-examination A computer-aided design system is too often
of ideas by both the man and the machine will characterized or glorified by its size and its
encourage creative thought that would other- repertoire of operations. A zoo of design ser-
wise be extinguished by the lack of an antagon- vices frequently provides the designer with the
istic (and thus challenging) environment. illusion of generality through sheer quantity of
Computer-aided design concerns an ecology specific routines. In Steven Coons’s original
of mutual design complementation, augmenta- Outline of the Requirements for a Computer-
tion, and substitution. Aided Design System (1963), the danger of ex-
hibiting a false generality has been well
marked. As long as the designer never calls for
a capacity that is not rigidly embedded in the
machine, the system will be successful. How-
ever, since it is not feasible to predefine and to
pinpoint all plausible operations and design
activities, it follows that a successful design
partner might be composed of one intelligent
and adaptable service rather than a group of
special-purpose services.
The principle is simple and, in computer-aided

design history, old. A well-nourished platoon of
specific design operations expects a status quo
and excludes a methodological evolution and
self-improvement. As a consequence, so-called
problem-oriented languages have been devel-
oped in an attempt to avoid this stagnation by
providing each user, after a brief learning peri-
od, with the potential of creating his own tailor
made utilities.
A problem-oriented language is a high-level

computer language whose formulation and
implementation assume a specific discipline or
set of disciplines. Such a language provides
the equivalent of a set of nouns, verbs, and
phrases. A user can easily learn them because
of their simplicity and relevance to his profes-

sion. For a civil-engineer user, basic operations, at any instant can transform itself (in response
like calculating bending moments or shear to a change in context) to appear as a special-
forces, might appear as verbs and be combined purpose machine. By sampling its environment,
with declarations; for example, t y p e p l a n e an adaptable machine could freely move from a
TRUSS YZ/LOADING LIST ‘TRUS-UNI’/DETERMINATE state of universality to a state of single-
(Logcher, 1967). With such commands,
a n a l y s is mindedness.
the user can implement his own algorithm for
determining the behavior of a structure. No adaptable machine exists today. However,
we can (and should) discuss the environment
But two things are wrong. First, we have a con- that such machines might sample in order to
dition where each designer is creating his own transform themselves. So far we have presented
library of services out of the problem-oriented a duet— designer and machine— in which the
language. Once created, note that these opera- machine’s “ image” of the real world is solely
tions are no less rigid and specific than the through one human partner. The designer’s
predefined package of design commodities. personal prejudices and distortions of the real
Even though the routines are user chosen and world would be planted, consciously or sub-
user made, they might be less effective than if consciously, in the machine. In such a closed
created by someone (or a machine) versed in system the machine could easily develop into a
the computer sciences, with the full potential of “ design patsy” or “ yes-man.” The machine
lower-level languages available to him. Second, would not challenge goals; it would only be pre-
when using a problem-oriented language, the pared to mimic the communicative manner and
user-made repertoire of operations is largely methodology of its one user. In this situation
determined by the language itself and the user’s the designer could embed his preconceived an-
understanding of it rather than by the nature of swers, and, accordingly, a noncreative, compla-
the design problem itself. The appearance of cent partnership would be formed through the
particular commands in the language and the lack of a challenging environment.
absence of others completely prejudices both
the choice of problem and the method of imple- Beyond the one-architect-one-machine dia-
mentation. In other words, a problem-oriented logue, the milieu of an adaptable machine must
language gives the same illusion of generality embody two further contacts with the real world.
as the rigid regiment of services. The common First, an adaptable machine (and thus an archi-
failure is a misunderstanding of the difference tecture machine) must receive direct sensory
(which is not a semantic difference) between information from the real world. It must see,
flexibility and adaptability. hear, and read, and it must take walks in the
garden. Information should pass into the ma-
The omission is evolution. A dialogue must be chine through observation channels that are
evolutionary; a mechanical partner must be direct rather than undergo the mutations of
evolutionary and hence adaptable. An adapt- transfer from the real world to designer’s sen-
able machine is a generalized mechanism that sors to designer’s brain to designer’s effectors
27
1 George Moore’s
steam man (circa 1893). A
gas-fired boiler operated an
engine of 1/2horsepower.
Exhaust was let out through
the helmet, steam through
the cigar. Its walking speed
was between seven and
nine miles per hour. (Photo-
graphs courtesy of Ronan
Picture Library, Newmarket,
England)
2 The Stanford Research In-

stitute robot.

to machine’s sensors. The designer does not parent machine, not by machines in residence.
completely control data that are collected in a The same central machine that provides the big
manner that avoids the consequent losses of bursts of computational power would be host to
information at each transfer point. Such data all the local issues resulting from many separate
bolster, once again, the machine’s capacity to architecture machines. The parent machine can
challenge. be a referee, an information source, a communi-
cation medium, a historian, as well as simply a
A great deal of research is being conducted in giant calculating mechanism. The parent
the quest for mechanical sensors. Probably the machine would store all building codes, Sweets
first to be incorporated with an architecture Catalog, Graphics Standards, and all the geo-
machine will be a seeing machine; this will be graphic and demographic data of the world.
briefly described in a subsequent chapter. At (The reader should note that the constitution
first, machines with eyes will observe simple and design of such a machine is not the prime
physical models; eventually, they will observe concern of this book. Our main concern here is
real environments. the machine in residence.)
A second tie to the real world would be the An architecture machine that could observe
capability of overviewing designers, their existing environments in the real world and
definitions, activities, and methods. Surveil- design behaviors from the parent would furnish
lance of other designers’ procedures again the architect with both unsolicited knowledge
nourishes the machine’s ability to challenge. and unsolicited problems. Someday machines
For example, the machine could alert its part- will go to libraries to read and learn and laugh
ner to the practices of other designers. (They and will drive about cities to experience and to
provide greater direct outdoor access in high- observe the world. Such mechanical partners
density residential areas.) This response would must badger us to respond to relevant informa-
allow the designer either to correct himself or tion, as defined by evolution and by context,
to reinforce his own and his machine’s convic- that would otherwise be overlooked.
tions further. At the same time, each designer
would be able to tune into controversy by dial- Machines that poll information from many
ing an anonymity or an opposing view that designers and inhabitants, directly view the
would discount his own whims and pet details. real world, and have a congenial dialogue with
A designer would be able to subject himself one specific designer are architecture ma-
and his machine to an objective scrutiny that chines. They hint at being intelligent machines.
would consist of (1) an evolutionary mapping of
popular desires, (2) a statistical overlay of
solution patterns, and (3) the images of archi-
tects he esteems.
This overviewing would be achieved through a

29

Citation:
DOI: 10.7551/mitpress/8269.001.0001
Published: 1973

MIT Press Direct
Copyright© 1970 by
of Technology
Printed and bound

of America.

the publisher.

Book Program.

censes/byncnd/4.0/



010 Aspects of
Design Procedures

From Perspectives to Holography replaced with the cathode-ray tube in anticipa-

Drawing a perspective is a procedure for visu- tion of creating perspective drawings (in their
alization, a procedure that has suffered from appropriate transformations) at a rate of sixteen
faddism in computer graphics, a procedure that to twenty frames per second, for providing the
can unfairly influence the process of creating illusion of traveling through an environment at
physical form. Alberti’s formalization of per- any speed and in a flicker-free manner
spective in 1432 helped foster the Renaissance (Negroponte, 1966). Assume that on the screen
preoccupation for the observer and the viewing of a cathode-ray tube we have a perspective
position (and possibly symmetry). Later, when drawing, derived from the machine’s model of
man regularly built above six stories, the “ bird’s some project. The rendered perspective is a
eye” and the “ worm’s eye” view made the crude jungle of lines describing a wire frame
stationary viewer even more manifest. Photog- structure. Larry Roberts (1963) has taken out
raphy even further reinforced this syndrome. the hidden lines, David Evans et al. (1967) have
put in halftones, and everybody is trying to
Finally, the movie camera relieved the perform the perspective transformations in
stationary-observer obsession by allowing the real time.
consideration of a path of movement and rota-
tions of a field of vision. Cinematographic Meanwhile, General Electric’s Electronics
methods, however, were cumbersome, and the Laboratory, Syracuse, New York, under NASA
film processing time made movie-making a contract, has developed a special-purpose
presentation procedure (off-line) rather than a computer that permits a viewer to voyage
study medium (on-line). Then came the instan- through an environment with hidden lines re-
taneous images of closed-circuit television. moved, with halftones, in real time, and in color.
Coupled with a model scope or fiber-optics Furthermore, the user of this system commands
cord (optical devices for visually placing one- the movement with an aircraft-type control stick
self within scale models), a designer could push that delivers him a motor involvement with the
his way through a model to simulate roughly the visual simulation. P. Kamnitzer of U.C.L.A. is
visual experience. Unhappily, television tech- presently applying the NASA-General Electric
niques are unwieldy. system to urban visual simulation problems
(Kamnitzer, 1969).
The computer is a natural medium for the mass
production of perspective images. At first, The history is long; the list of participants is long
numerically controlled plotters were employed (M. Milne, 1969). Why the great concern with
to draw perspectives at hundreds of small in- perspectives? First, the problem is intrinsically
crements along a path. These drawings were natural for computer graphics studies, its formu-
then filmed with animation procedures to pro- lation is technically difficult (thus stimulating),
duce a cartoon of moving figures (Fetter, 1964), and it requires no examination of design
a general procedure more cumbersome than philosophy.
any previous method. Then the plotter was
31

1 da Vinci: “ Le Prospecto
graphe,” drawing circa
1488. (Courtesy of Bibliote
ca Ambrosiana, Milan)
2 Durer: “ Le Dessinateur de
la femme couchée,” en-
graving, 1525.
3 Durer: “Le Portraitiste,”

engraving circa 1525.
(Courtesy of the Biblio-
thèque nationale, Paris)
4 Six frames of a computer

graphics film on visibility
studies of an aircraft carrier
landing. The film was made
by William Fetter for the
Boeing Company. The last
ten seconds of the carrier
landing required two
hundred and forty comput-
er drawn perspectives to be
plotted, touched up by an
artist, and then filmed. (Il-
3
lustrations courtesy of the
Boeing Company)
5 Three black and white

photographs of Peter Kam
nitzer’s color display at the
Visual Simulations Labora-
tory of General Electric. The
images are from the CITY-
SCAPE program, and they
are presently restricted to
240 edges per frame. (Pho-
tographs courtesy of Peter
Kamnitzer)
6 A rendering made to
study the effects of increas-
ing to 1,500 edges in the
above system. (Courtesy
of Peter Kamnitzer)
3
3
2 3

1 Larry Roberts’ Wand. Perspective is a natural procedure for repre-

(Courtesy of Lincoln Labo-
ratories) senting in two dimensions the illustration of a
three-dimensional event. On a picture plane a
2 An electromechanical de- trace of points defines the intersection of
vice used for input of three-
dimensional data. The de-
imaginary lines between a monocular observer
vice is much like an aircraft and the real or unreal world. When the picture
joy stick and is coupled with plane is removed from this world and viewed
the adjacent stereoscope.
(Photograph courtesy of Mi-
from the same vantage point, the image is an
chael Noll) accurate representation with no distortion. The
mode thus affords an appropriate visual rep-
3 A stereoscopic viewing resentation of the visual aspects of an architec-
attachment on a large
cathode-ray tube. This tural real world. But, with future three-
attachment was designed by dimensional displays and input mechanisms,
C. F. Mattke for use by Mi- the virtuous role of the perspective drawing
chael Noll in his investiga-
tion of three-dimensional surely will be diluted. As Coons states, “ In a few
man-machine communica- years from now (April 1968) you (a group of
tion, performed at the Bell architects) will be able to walk into a room and
Telephone Laboratories.
(Photograph courtesy of Mi- move your hand and have a plane or surface
chael Noll) appear before you in light. You will be able to
build a building in light so that you can walk
around it and change it” (Herzberg, 1968).
The dramatics of such dazzling statements

stem from the age-old desire of the architect
to be able to lift his pencil, gesticulate in mid-
air, and have the stylus ooze out lines that float'
in space. Part of this desire has already been
fulfilled by an ultrasonic position-sensing de-
vice. Larry Roberts’ Lincoln Wand allows the
computer (within a work space of ninety-six
cubic feet) to track the x-y-z position (to the
nearest one-fifth of an inch) of a hand-held,
pen-size device (Roberts, 1966). Four ultra-
sonic transmitters recurrently pulse bursts of
energy, and the Wand reports to the computer
the time at which it heard each signal. The
computer uses three time lengths to determine
trigonometrically the pen’s position; the fourth
transmitter provides a geometric check on the
35
1 2

1 Ivan Sutherland’s helmet. measurements. Unfortunately, the Wand does

(Courtesy of Ivan Suther-
land) not leave traces in midair upon which to build
consecutive lines (which aggravates the prob-
2 The Direct-View Three- lem of hand trembling). Though a three-dimen-
Dimensional Display Tube
developed at Hughes Re- sional model is being constructed within the
search Laboratories by R. machine, the output associated with the Wand
D. Ketchpel (1963). In con- is at present a perspective or axonometric
trast to other presentations
in which the third dimension display.
is simulated stereoscopi-
cally, this device displays Further efforts will eventually allow three-
the information in actual
space. The three-dimen- dimensional displays to be joined with wandlike
sional display tube utilizes a devices. Ivan Sutherland is creating a machine
phosphor-coated disc spin- that gives the illusion of actually walking
ning at 900 rpm. Upon exci-
tation by a cathode-ray around and within visual models (I. Sutherland,
beam at selected times, any 1968). The device is a helmet mounted with
point in the volume “swept two eyeglass-size cathode-ray tubes (with
out’’ may be illuminated
thirty times a second. (Pho- prisms) that permit stereoscopic images to be
tograph courtesy of R. D. transformed in accordance with the head posi-
Ketchpel) tion of the wearer. In this case three antennas
3 This illustration is a di- report the user’s position, but the movement
rect positive print of a holo- could also be monitored with the user driving
gram of the model to the a simulated car and actually driving through a
left. If viewed as a trans-
parency with a coherent city that does not exist in the real world. With
light source behind, this halftones, color, and real time, this technique
hologram would display the would afford an excellent simulation of the
model in three dimensions.
visual world. Sutherland’s device even allows
for a split image, through the prisms, that per-
mits the designer to view his project overlaid
upon the visual real world.
Another three-dimensional display technique

is holography (Gabor, 1948). TV Guide peri-
odically tempts its readers with three-
dimensional television: ballet dancers in your
living room and the Tonight Show in your bed.
In hologram television, “ the pictures have a
realism unattainable by any other means. The
three-dimensional effect is obtained without the
need for a stereo pair of pictures, and without
37
the need for any devices such as Polaroid vices that have been discussed do not delve
glasses. In addition, all the visual properties of into the crucial problem of machine response to
the original scene, such as parallax between nonvisual involvement with the environment:
near and far objects in the scene, and a change auditory simulation, tactile presence, feeling of
in perspective as a function of the observer’s a breeze in a lonely space.
viewing position, are present” (Leith and
Upatnieks, 1965). This apparition is achieved
by recording the interference patterns of two
sources of coherent light (usually lasers), one
reflected directly from the object and the other
by a mirror.
At present, efforts are being conducted to

construct through computation synthetic
holograms for simple geometric configurations
(Lesem et al., 1968). One method calculates the
interference patterns and plots the result on a
transparency. Another method positions a small
mirror in three-dimensional space and traces
the configuration in the presence of the neces-
sary light sources, in effect, taking a time-lapse
hologram (Stroke and Zech, 1966). So far,
neither method is in real time.
When computers can simulate holograms in

real time (using some flat-screen television
technique), views of the machine’s mathe-
matical model could be selected in a general
manner, and the designer’s head movements
could supply specific vantage points. Soon, on
a display device, architects will have glimpses
of physical environments that do not exist.
These witnessings will be in full color, with
halftones, and in three dimensions.
The reader must remember that these appar-

ently ghostlike images are only visual simula-
tions. Though the better ones will furnish a
motor involvement with the designer, the de-

Generation of Solutions posed certain factors and courses of action on

To some of the more hidebound architects the the generated solutions.
concept of a machine generating three-
dimensional solutions is immoral, impossible, There are two distinct types of generated
or endorses unemployment for threatened solution: one accommodates underconstrained
architects. The premise is that a human archi- problems; the other works within overcon-
tect’s experience gives him license to be the strained situations. The underconstrained
exclusive translator of human requirements situation (rare in architecture) has a large set
into physical form. of possible solutions. The criteria are satisfied
by many alternatives. These alternatives must
Physical form, according to D’Arcy Thompson then be evaluated by the architect using “ in-
(1917), is the resolution at one instant of time tuitive” means, selection criteria he either does
of many forces that are governed by rates of not understand or has never presented to the
change. In the urban context the complexity of machine.
these forces often surpasses human compre-
hension. A machine, meanwhile, could pro- In the overconstrained problem, the generating
create forms that respond to many hereto un- mechanism is presented with great amounts
manageable dynamics. Such a colleague would of factional criteria that no form can completely
not be an omen of professional retirement but satisfy. The generating mechanism searches
rather a tickler of the architect’s imagination, for a solution that best relaxes the constraints,
presenting alternatives of form possibly not a point of greatest “ happiness” and least “ fric-
visualized or not visualizable by the human tion.” The resulting form is a status of criteria
designer. compromise where the constraints least an-
tagonize one another.
An architect would not and should not confront
a “ criteria machine” to decrease visual privacy, Both problem types involve trial-and-error
increase public access, and watch contortions procedures, tasks well suited for self-improving
of form on a television screen. Instead, in the machines. In many cases random numbers are
rhythm of the dialogue, a solution-generating employed; they deserve mention, as their use is
capacity would be an evolutionary enterprise often misunderstood. In solution generation, a
where the machine would act in “ interrupt” or random number is a substitute for missing
“ reply” to its partner’s activity. The architect information or unpredictable information. Rath-
might search for a configuration by observing er than just cast a Monte Carlo atmosphere of
the machine’s attempts at satisfying a state- surprise, random numbers simulate non-
ment of the problem, or the machine might learn deterministic events such as family displace-
by observing the architect. In such a system ments, employment changes, physical expan-
both the architect and the machine could inter- sion. Usually, as a system grows, events be-
rogate each other in order to locate those char- come more and more deterministic, and the
acteristics of the site and the criteria that impossible alternatives diminish. Generating pro-
39
2 3

1 GRASP, Generation of bathroom (a debatable

Random Access Site Plans. functional relationship), a
This computer program gen- TV room, a washroom, and a
erates solutions only within sewing/laundry room. (Illus-
an underconstrained situa- trations courtesy of the Har-
tion where the operator vard Laboratory for Comput-
specifies dimensions, “ no- er Graphics)
build areas,” density, cost,
and aspects of privacy. 4 A preliminary output from
“Good” solutions can be the Children’s Hospital Proj-
plotted in perspective or or- ect of the Leo A. Daly Com-
thographic modes. (Work by pany, Architects. The 134
Eric Teicholz. Illustration activities are given minimal
courtesy of the Harvard interrelationships. While
Laboratory for Computer talking with a particular de-
Graphics) signer, the program implicit-
ly develops functional rela-
2 Two outputs of COMPRO- tionships through trial and
GRAPH 3 by Eric Teicholz error, punishment and re-
and Thomas Follett ward. Over time the system
for architects Perry, Dean, should improve. It is now
and Stewart. The user pre- under research by Stephen
pares a three-dimensional Flanders and Lee Wind
matrix as input, specifying heim, using the Service
size and functional relation- Bureau’s CALL 360. (Illus-
ships. After specifying the trations courtesy of the Leo
envelope, radial or linear, A. Daly Company)
the routine will generate
schematic plans on a floor-
by-floor basis. This particu-
lar computer program pro-
motes a present-day meth-
od that in itself is debatable
and is certainly question-
able in the light of emerging
computer techniques. (Illus-
trations courtesy of the Har-
vard Laboratory for Comput-
er Graphics)
3 RUMOR, The random gen-

eration and evaluation of
plans. A matrix of relation-
ships is established by the
operator for each criterion.
“No effort has been made to
generate only ‘good’ plans”
(Bernholz, 1969). The two il-
lustrations represent a
house plan composed of a
living room, dining room,
kitchen, four bedrooms, one
4
41

The three small illustrations cedures can appropriately acknowledge this

are models of three of the
ten inputs to LEARN. The
sort of growth by changing the distribution of
remaining illustrations are “ randomness” in response to the present state
representative of the out- of the form, as described by previous actions,
puts at different time inter-
vals. The work was per-
external information, and stage of growth.
formed by Anthony Platt,
Peter Bailey, Gary Ridgdill, As one example of solution generation, a stu-
and William Hurst.
dent project— LEARN— was developed by a
group of M.l.T. Master’s of Architecture stu-
dents who had no previous computer program-
ming experience. LEARN was a computer
mannerist. It watched the designers’ activities
by observing ten simple solutions. (In this case
they were “ sugar-cube” models transcribed to
punch cards describing x-y-z centroid locations
of solids and voids). Following these ten arche-
types, the machine was asked to generate a
solution of its own. The appeal of this simple
experiment is that the criteria were first de-
termined from the form and then used in the
generation of the alternatives. The students
observed the variations within the given “ style”
of the solution. The mannerism was derived
from the original ten solutions and was then
updated by the eleventh. The machine pro-
ceeded to generate a twelfth solution, updated
its “ manner,” generated a thirteenth, and so on.
After a denouement of five thousand separate
solutions to the same problem, the mannerist
machine did not generate or embark on wild
tangents. In fact, the conviction of the machine
was so intense that the last thousand solutions
had little distinguishing variety.
A second example is GROWTH, also a student

project. This system operated within a larger
work space (approximately a square mile) than
LEARN and did not observe a specific design-
er’s methods. The generated solutions were
43
GROWTH. The final run of

this program used two
hours of dedicated IBM
360/65 computer time to
simulate 266 stages of
growth. The experiment was
conducted by Judd Knoll,
John Maugh, and Chin Pai.
The eight illustrations, from

top to bottom, represent the
following stages with the as-
sociated number of solids:
stage-solids
11-11
26-69
35-103
59-205
131-555
179-801
235-1082
266-1251

periodic glimpses at stages of growth. The solution is a function of weighted proximities,

computer employed the principle of “ influ- orientations (site and exposure), visual access,
ences,” where each element’s status (solid or acoustical access, circulation, and others to be
void) was determined by its “ conviction” (to be implemented. It is displayed to the user for his
what it was or to be what it was not). As soon consent and rearrangement. Subsequently, the
as a void became solid or a solid became void, machine regenerates a solution more specified
ripples of influence would disperse, locally than the last but in the same tenor as the last.
disturbing the convictions of adjacent elements Because the machine does not explore diver-
(in proportion to proximity and activity relation- gent tacks, it could channel the unwary user in
ships). A solid might become more convinced the wrong direction.
of its solidity or else an adjacent void might
tend toward a state of solidity, being now un-
convinced of its status. In effect, the rules of
conviction were the generating force. For
example, a lone ten-foot cube in the middle of
a large field might influence its void neighbors
under one set of rules to be less convinced of
their voidness and accordingly raise their
probability of changing state in the next stage of
growth. Meanwhile, another set of rules might
make the edge members of a large complex
thoroughly convinced voids or thoroughly
convinced solids. The same rules might tend to
lower the conviction of deeply embedded solids
(in order to avail the form of interior open
spaces in response to size).
A third example is the ongoing research of

Timothy Johnson and Richard Krauss at M.I.T.,
under NSF contract (T. Johnson et al., 1969).
Under the direction of Albert Dietz, this space-
allocation work employs sophisticated mathe-
matics and sophisticated graphics to optimize
cross-coupled constraints and display the re-
sults. The generated solutions are composed of
“ use-surfaces” and “ use-volumes.” They are
the result of optimization techniques that
assume missing information (as opposed to
replacing it with random numbers). A generated
45
A photographic record of a
circulation conflict. In this
case the simulation is the
real world, the best model
but the most expensive.
Similar displays will soon
be manageable by comput-
ers. (Photographs by Tom
Payne)

Simulation of Events assumes the rules are correct. Whether empir

When enough empirical or experimental rules ical or experimental, simulations are no better
are known about a process, machines can be than their underlying rules, whether the rules
made to take on the character of the event are provided by the man or by the machine. If
and undergo a make-believe happening of that the simulation model is correct, a designer or a
process, a simulation. Given reasonable proto- machine can observe the performance of an
cols and maxims, this form of mechanical mas- environment, a specific context. Someday,
querade is a powerful method for refining an designers will be able to subject their projects
original set of rules or pretesting designs and to the simulations of an entire day or week or
procedures. year of such events as use patterns and fast
time changes in activity allocations. On display
The simulation of events can benefit the archi- devices, designers will be able to see the
tect in two ways. If the designer does not fully incidence of traffic jams, the occurrence of
understand the behavioral aspects of an event, sprawl, or sweltering inhabitants searching for
he can play with rules and regulations, search- shade. For the present discussion, the most
ing for recognizable activity patterns. In other easily reproducible event is circulation, a
words from empirical knowledge of a set of perplexing and important urban situation in
actions and reactions for specific environments, itself.
a designer could inductively compose postu-
lates or algorithms applicable in other contexts. Many sophisticated organizations have spent
For example, if he understands from on-pre time and money in programs that simulate
mise measurements the vertical circulation circulation, primarily vehicular circulation.
patterns for several different environments, Rather than observe these elaborate simulation
he could describe these environments to the techniques, let us observe two very simple cir-
machine and hypothesize seemingly appro- culation models that have been devised to simu-
priate rules. Then when the machine, using late pedestrian movement. The models result
these rules, displays the vertical circulation from two M.l.T. student projects involving
patterns for the known environments, it re- architecture students, once again with almost
veals the divergencies between the empirical no previous programming experience.
data and the designer’s rules— between
what he knows he should see and what he The first simulation model describes three
does see— giving him information by which parameters: spaces (function, capacity, desir-
to modify the rules, always observing whether ability), circulation interfaces (direction, ca-
the change has a positive or negative effect. pacity, demand) and people (arrivals, de-
Eventually, using a dynamic on-line system, partures, frustrations). The model assumes the
he will be able to converge on rules of simula- chosen environment to be a discrete chunk of
tion that can be applied to other environments. the real world, with a certain number of pedes-
trians leaving the system and others arriving at
The second design application, pretesting, each time interval. At each instant, the circula-
47
CARS — computer-
automated routing and
scheduling. This system is
designed to provide door-
to-door transportation in
low-density suburban areas.
The aim of CARS is to pro
vide service approximating
that of the taxicab, but at a
price approximating mass
transit bus systems.
1 The six illustrations repre

sent a simulation of CARS
in operation with twelve ve
hicles on an area of nine
square miles with about
ninety demands per hour.
Two particular criteria are
enforced: no one should
wait more than fifteen min
utes, and no one should
travel more than 1.8 times
the direct driving distance.
2 3 The illustrations have been

photographed from an
ARDS tube which runs off
the M.l.T. time-sharing sys
tem. The work is being per
formed at M.l.T’s newly cre
ated Urban Systems Labora
tory under the direction of
Daniel Roos.
1 All waiting demands at

time 45
2 The projected tour of vehi

4 cle 11 at time 45
3 A history of vehicle 11 at
time 45
4 All waiting demands at

time 60
5 The projected tour of vehi

cle 11 at time 60
6 A history of vehicle 11 at
time 60
5 6

tion activity and the space populations are borhood immediately responds to a change,
determined by random numbers controlled by generates new demands, and the supposedly
parameters of frustration and desire. Although beneficial event is too often invalidated. Such
this work was not implemented on a graphic negations can be avoided. Direct interplay
display device, the authors (with some effort) between event and effect, desire and result can
can observe jammed doorways, vacant com- be observed and can be enveloped in simulation
mercial spaces, and periodic peaks in major procedures.
circulation routes. Their physical model can be
changed and manipulated in search for less
antagonistic circulation patterns, iterating
toward a design solution that would display
ambulatory ease and facility.
In this example, simulation techniques describe

agglomerates of people, whole groups moving
from space to space in one cycle. The second
student model applies variable parameters to
each individual pedestrian. Characteristics of
desire and destination control the simulated
movement of each individual pedestrian in ac-
cordance with his local environment. The stu-
dent can observe frustrations and localized
frictions that are not only a function of the
physical form but responses to the individual
personalities of the other pedestrians in the
same space. The student can observe a dashing
blonde unsettle corridors or a precipitate
fleet-footed latecomer disrupt a reception area.
Both student projects, even in their infancy,

exhibit viable methods for prediction. When
simulation techniques improve and are part of
architecture machines, physical structures
can be tested within environments that acknowl-
edge their presence. In other words, when a
change is contemplated for some neighbor-
hood, it can be tested by observing its effect
over time, but in fast time, unreal time. Usually,
in the nonpretend world, the real world, a neigh-
49

SYMAP. This computer sys- Bits of Design Information

tem is primarily concerned Census data, site descriptions, transportation
with the display of spatially
distributed data rather than statistics, activity constraints, economic criteria,
with its manipulation. Devel- and material specifications are all part of the
oped by Howard Fisher, it
employs an overprinting
bulky dossier of design information necessary
technique on a high-speed for any urban design project. The information
printer, which in the days burden is fantastic. What usually happens in
before computer graphics
was fine, but is quite obso-
most design procedures is that a handful of
lete today. The four maps criteria are chosen and thoroughly developed;
are based on the 1960 all the remaining information relationships are
census and are at the scale
of the census tracks. From
expected to fall into place, or else residual
left to right, they display issues are crammed into unsuspecting recep-
density per acre of total tacles. Or in a gesture of design fatalism, ac-
population, whites, blacks,
cepting the unfeasibility of it all, a group of
Puerto Ricans. The work
was performed by Peter parallelepipeds are contrived and refined to
Rogers and Isao Oishi. (Il- accommodate as well as possible the internal
lustrations courtesy of the demands of some institution. The problem and
Harvard Laboratory for
Computer Graphics) the result are commonplace— look at your own
city.
An architect’s role in urban design requires a

complex information supply with characteristics
of retrieval, labeling, and interassociation. But
machines are good at this. Though there are
technical problems and real computer-program-
ming issues, machines can respond to and have
access to millions of billions of bits of informa-
tion. It is estimated (Servan-Schreiber, 1967)
that the number of all letters in all words in all
books in all libraries in the world exceeds one
thousand trillion (1,000,000,000,000,000). J. W.
Senders (1963) estimates that the current
growth rate of this store is about four hundred
thousand letters per second. Even a modest
architect might assume that he needs some of
this store.
In the human nervous system, information genu-

inely constitutes authority (McCulloch, 1965). In
51
DISCOURSE
In the five maps, the follow-

ing symbols apply:
1 = residence
2 = industry
3 = centers
B = residence and centers
C = industry and centers
+ = river
—= channel
1 2 * = channel and river
1 Ciudad Guayana 1961,

showing residence and in-
dustry
2 Ciudad Guayana 1969,

showing residence and
industry
3 The proposed design for

Ciudad Guayana generated
by designers from the Vene-
zuelan Development Corpo-
3
ration for the Guayana Re-
gion
4 Two patterns generated

by DISCOURSE decision
rules (Illustrations courtesy
of William Porter)

design, however, abundant data can confer Information search, by either designer or ma-
prestige on mediocre designs, especially when chine, would occur for the most part in a
facts arrive from the unequivocating computer. localized fashion, investigating by proximity
Data can be prepared to support any design if (by neighborhood, by street, or by immediate
the selection of evidence is limited to that which adjacency). The thrust of this sort of data-
favors the cause. “ Poor data and good reasoning structure argument is that information is treated
give poor results. Good data and poor reasoning locally, by positions, and less globally, by
give poor results, poor data and poor reason- attributes. Thus, design information is retrieved
ing give rotten results” (Edmund Berkeley, by geometric (topical) search rather than by
1967). intersecting generalities.
A machine could store relevant information in Such a position-oriented storage vehicle may
many ways. Relational and associative data be unique in the physical design problems of
structures, for example, store classes of items the urban environment. In a library reference
by properties of similarity and retrieve them by system, this type of information structure is
querying for that which has “ this and that, but ridiculous; books are not categorized by their
not those.” Another structure uses lists of at- position on the shelf, books are redundantly
tributes that point “ fingers” at members that classified by name, author, subject, publisher,
have the same attributes, thus tying threads and so forth. Unfortunately, good library data
among the various members of the data struc- structures are all too often foisted onto design
ture. Still another (and simpler) method is a problems.
matrix organization where rows and columns of
entries are entered and looked up by addressing One particular design information system—
a two-, three-, four-, or n-dimensional table. DISCOURSE— warrants mention, as it exempli-
fies a flexible data structure that combines the
But in architecture, most information has a assets of associative and matrix organizations,
natural disposition— the positional relationship attribute and geometric searches. This research
— which can help to organize the proliferation team (Fleischer et al., 1969, and Porter et al.,
of data. Design manipulations invariably wield 1969) uses the M.l.T. time-sharing facilities to
locational data expressed in terms of position, interact (no dialogue) with data files and print
distance, area, or volume. This natural geo- the results in tabular or map format. It is a prob-
metrical referencing suggests a data structure lem-oriented language that derives flexibility
where each physical location (solid or void, from (1) providing multiple data structures for
building or open space), to as small a grain as both local and global interrogation, and (2) pro-
possible, would describe itself in an autono- viding a “ meta-language” that allows the de-
mous fashion (even the voids!). This has strong signer to create his own search techniques.
implications, especially the Euclidean and re- The reader must understand, however, that
dundant nature of geometrically related data. DISCOURSE is not computer-aided design
within our definition. It is an excellent computer
53
system that manipulates bits of design informa- Machines in Residence

tion, that is, information that has been explicitly
given to the machine by the user. Modern decision theory, economics, psychology,
and game theory recognize, as a basic case,
Another example is MEMORY, an information clearly motivated individual choice under con-
storage and retrieval system that is being ditions of complete information. It is also recog-
studied within M.I.T.’s Urban Systems Labora- nized that two unfortunate facts of life remove
tory. MEMORY’S dominant feature is its “ for- us from the relative simplicity of this basic case.
getting convenience.” It is a way of storing The first concerns man as an information pro-
events in neural nets that are highly redundant cessor and the second the conflict of individual
and, at first, rather random. Over time and with group preferences.
through repetition or the lack of it, events be- Martin Shubik, “ Information, Rationality and
come, by the strength of traces left in memory, Free Choice in a Future Democratic Society”
either stronger remembrances or fainter recol-
lections. At the onset of such a system, for any Lower-class people need big kitchens; middle-
given input the output will be mostly garbage. class people need big bedrooms; corridors are
Overtime, the responses should gain meaning for the poor, and so forth. Design universals en-
with respect to both the input and the relevancy able federal housing authorities to set minimum
(defined by time) of the input. The reason that standards, they enable architects to disregard
this experimental work is important to an archi- specifics, they delight lovers of empirical gen-
tecture machine is that the design process is an eralizations. In short, empirical generalizations
evolution of (1) the product, the form; (2) the of life styles are for the comfort and conven-
process, the algorithms; (3) the criteria, the ience of the decision makers’ tools, not neces-
information. MEMORY addresses itself to item sarily for the well-being of the people.
number three.
Today we have “ advocacy planning,” a design
procedure that tries to overcome the lumping
of life styles, that tries to satisfy particular re-
quirements. Attempts to procure individual
needs and desires have embodied several for-
mats: the questionnaire (fill in the missing
spaces), the neighborhood meeting (we are
here to listen to your problems), the personal
interview (tell me what you want). Note that in
each of these communications media it is as-
sumed that the asker knows what to ask, the
answerer knows what to answer, and that minds
will not change rapidly. Furthermore, advocacy
planning is conducted in such unreal time that

the fancies of the individual householder dissonance that exists in today’s housing
change in the lapse of time. problem.
Before suggesting procedures that are more Even today, the touch-tone telephone gives rise
appropriate to the articulation and satisfaction to a home computer terminal whose ten-button
of local desires, let us first assume two future dialect humors a potentially ubiquitous man-
technological advances: versatile building sys- machine conversation. Coupled with audio
tems capable of responding to changing (per response units, such telephones can converse
month, season, year) human needs and the di- with button-pushing as an input and spoken
rect concern of this book, home computer English as an output. Frank Westervelt (and
terminals capable of talking in a graphic and Smith, 1968) has incorporated such a system at
auditory fashion— “ but I don’t see any computers the University of Michigan’s Computation
getting into my house” (A. Milne, 1963). Center.
You need not look too far, maybe ten years: Richard Hessdorfer is expanding Westervelt’s
“ . . . computer consoles installed in every home system by constructing a machine conver-
. . . everybody will have access to the Library sationalist. Hessdorfer’s work is aimed at initiat-
of Congress .. . the system will shut the windows ing conversation with an English-speaking user.
when it rains” (McCarthy, 1966). Such omni- His problem is primarily linguistic. The machine
present machines, through cable television tries to build a model of the user’s English and
(potentially a two-way device), or through pic- through this model build another model, one
ture phones, could act as twenty-four-hour of his needs and desires. It is a consumer item
social workers that would be available to ask (as opposed to an industrial or professional
when asked, receive when given. Imminent tool) that might someday be able to talk to
changes in family size could be overlaid upon a citizens via touch-tone picture phone, or inter-
local habitat in an effort to pursue growth that active cable television.
would not curtail the amenities children need.
As a part of the Hessdorfer experiment, a tele-
Granting machines in the home, each urbanite typewriting device was brought into the South
could intimately involve himself with the design End, Boston’s ghetto area. Three inhabitants
of his own physical environment by (in effect) of the neighborhood were asked to converse
conversing with his own needs. Or, another with this machine about their local environment.
way of thinking of the interaction is that every- Though the conversation was hampered by the
body would be talking to the architect, not ex- necessity of typing English sentences, the chat
plicitly but implicitly, via a machine-to-machine was smooth enough to reveal two important
interchange. Architects would respond to partic- results. First, the three residents had no qualms
ular patterns of a neighborhood and submit or suspicions about talking with a machine in
alternatives to be played with and in such a English, about personal desires; they did not
manner possibly penetrate the designer-dweller type uncalled-for remarks; instead, they im-
55
1 2

1 The three protagonists of mediately entered a discourse about slum land-

the Hessdorfer experiment,
Maurice Jones (top right),
lords, highways, schools, and the like. Second,
Barry Adams (top left), and the three user-inhabitants said things to this
Robert Quarles (bottom machine they would probably not have said to
left). It is interesting to note another human, particularly a white planner or
the button Robert Quarles
happened to be wearing politician: to them the machine was not black,
that day: “Tenant Power.” was not white, and surely had no prejudices.
2 Picturephone. Copyright
(The reader should know, as the three users did
1969 Bell Telephone, Inc., not, that this experiment was conducted over
Murray Hill, New Jersey. Re- telephone lines with teletypes, with a human at
printed by permission of the the other end, not a machine. The same experi-
Editor, Bell Laboratory REC-
ORD. ment will be rerun shortly, this time with a ma-
chine at the other end of the telephone line.)
"Reused with permission of
Nokia Corporation"
With these domestic (domesticated) machines,
the design task becomes one of blending the
preferences of the individual with those of the
group. Machines would monitor the propensity
for change of the body politic. Large central
processors, parent machines of some sort,
could interpolate and extrapolate the local
commonalities by overviewing a large popula-
tion of “consumer machines.”
What will remove these machines from a “Brave

New World” is that they will be able to (and
must) search for the exception (in desire or
need), the one in a million. In other words, when
the generalization matches the local desire, our
omnipresent machines will not be excited. It is
when the particular varies from the group pref-
erences that our machine will react, not to
thwart it but to service it.
57

Citation:
DOI: 10.7551/mitpress/8269.001.0001
Published: 1973

MIT Press Direct
Copyright© 1970 by
of Technology
Printed and bound

of America.

the publisher.

Book Program.

censes/byncnd/4.0/



011 Aspects of
Design Processes

Sequential and Temporal Events Your ozalids are ready. Your wife has just
A process is a progressive course, a series of called. . . .
procedures. A procedure is replicable (if you
understand it) in an algorithm; its parts have a The example describes a participation where
chronological cause-and-effect relationship each party is interjecting and superpositioning
that can be anticipated. A procedure can be events directed toward a common goal.
replicated with the appropriate combination of
commands. In short, a procedure is determin- Each event is either a temporal or sequential
istic and can be computerized within a given occurrence; together they constitute part of a
context. process. A sequential response of one protag-
onist is generated by the previous event in the
Conversely, a process cannot be computerized, dialogue, usually on the behalf of the other. A
but, as we have said, it can be computer-aided. sequential event is a reply. It can be the reply
Particularly in the design process, respective to a facial expression or the answer to a ques-
events are not chronologically ordered. The tion. What is important, however, is that not
following scenario, without the enrichment of only is one actor responding but he can assume
graphics, intonations, bodily involvement, that the other is listening and probably is aware
crudely illustrates an architect-machine of the context. In other words, a sequential epi-
dialogue: sode assumes the reply of one (intelligent) sys-
tem and the attention of the other system— a
Machine: chain of chronologically ordered incidents.
George, what do you think about the children’s
activities in this project? This well-known command-and-reply relation-
Architect: ship between man and machine does not in
How far must a child walk to nursery school? itself constitute a dialogue, as it ignores all
Machine: events except those ordered by time sequence.
The average distance is 310 feet. The Soviet Union’s A. P. Yershov (1965) has a
Architect: diagram illustrating this proverbial man-
Each dwelling unit must have direct outdoor machine interaction, as he calls it, “ director-
access and at least three hours of direct agent” interaction. Note that in the diagram,
sunlight. Professor Yershov has drawn three arrows
Machine: within the man’s head and only one arrow within
Of the children we were just discussing, 92 per- the machine. The three arrows imply an ever
cent must cross a road to get to school. continuing act particular to the role or constitu-
Architect: tion of the man and not the machine. Let us call
We will look at that later. With respect to dwell- this act deliberation.
ing units, we must assume at least two vehicles
per family. The act of perpetual cogitation can be equally
Machine: accorded to machines, especially since we have
59

1 The Yershov diagram. previously insisted on a dedicated small ma-

This first appeared in a chine in residence, devoting its full computa-
paper presented at a Sem-
inar on Automation of the tional ability full time. We will call machine
Thinking Process held at deliberation “ temporal” work. It resides in the
the Kiev Center for Scien-
background and surfaces as an interrupt. The
tific and Technical Informa-
tion, Kiev, USSR, 1963. interrupt (though not necessarily the delibera-
tion) is context-dependent; thus we can
2 Design History Chart
(Myer and Krauss, 1967).
probably assume that the temporal zone re-
The diagram represents a quires an intelligence. Furthermore, note that
series of procedures rather it is this zone of temporal events that the de-
than a process. The chro-
nology of left to right, the
signer interrupts when presenting a fact or a
two-dimensional apsects of task.
the printed page, and the
start/finish overtones all
contradict the nature of the
In the foregoing sketch, the machine addresses
design process. If it were the architect, presumably interrupts him. Fol-
possible to diagram this lowing, the architect addresses the machine (in
process, then such diagram-
ming would occur out of
fact interrupts) with a specific question that is
context, and present-day not a reply but is within the same context. The
machines could handle it machine’s reply is sequential: “ . . . 310 feet.”
without an artificial intelli-
gence. (Diagram courtesy of
While the architect thinks about the response,
the M.l.T. Center for Build- the machine further investigates the children-
ing Research) nursery relationship (we assume here a pre-
vious experience by the machine with such
issues). Within three seconds of user delibera-
tion, a machine could devote between three
hundred thousand and three million operations
to the children-nursery relationship.
Meanwhile, during the machine’s activities, the

architect reinterrupts the machine and states
criteria with reference to a new context: “ Each
dwelling unit must have direct outdoor access
and at least three hours of direct sunlight.”
After the machine has listened (and heard), it
interrupts the architect and lets surface from
the temporal zone the unsolicited information
about children’s circulation. The architect
postpones consideration of his oversight and
proceeds to supply further design constraints.
61
Following, the machine interrupts again with a The Geometry of Qualities

time-dependent occurrence. In order to make adept, temporal comments,
an architecture machine must have a certain
It may now be more evident why an evolutionary basic understanding of qualities. Though at first
machine must have the capacity for context primitive, this qualitative appreciation itself
recognition. A complete mishmash of irrelevant would evolve within a value system that is very
comments from a nonintelligent, nonevolution personal, between a man and a machine.
ary machine would confuse the designer and
thus stifle the design process. While at the The handling of qualitative information is too
onset of any partnership the machine’s inter- often presumed hopeless for the constitution of
ruptions might appear random or disorderly, machines. Or it is granted feasibility only
they would gain relevance through evolution. through the abortive techniques of quantifica-
The sophistication of these temporal actions is tion. No doubt, characteristics of identity, op-
essential for machines to mature into intelli- pression, and fulfillment are hard for our present
gent partners. machines to comprehend. Nevertheless, even
with existing machines, properties of privacy or
accessibility or the natural environment furnish
qualitative features that can be readily ex-
pressed in terms that are understandable to
machines, machines that for the time being
have not experienced these qualities. This is
because we already have a model whose base
is geometry. This geometric structure, resulting
from the form base of urban design and the
pictorial structure of computer graphics, hap-
pens to suit the topology of many environmental
qualities.
For example, within some context, visual priva-

cy has an explicit geometry. The presence of a
transparent surface, while providing light and
view, might not yield visual privacy. A machine
can check this without disturbing the architect,
by weighing the activity (sleeping, eating, bath-
ing), the actor (housewife, bachelor, exhibition-
ist), the external uses (commercial, recreation-
al, residential), and the geometry of the two
spaces that abut the surface. With an evolution-
ary designer-machine agreement of the defini-

tion of visual privacy, the four ingredients can An architect attempting to provide natural
determine either an unequivocal absence or amenities, a resident trying to overlay his own
presence of visual privacy, or a graded value needs, and a machine endeavoring to tran-
of it. scribe these qualities through some geometry
all together comprise a system that must al-
Unfortunately, visual privacy has psychological ways be in equilibrium. The maintenance of
and personal ramifications not expressible in this equilibrium is the design process. Within
the four parameters. These subjective and this definition, the urban environment is a multi-
personal parameters are important; however, tude of quantitative and qualitative, local and
they are more appropriately manipulated by the global, individual and group forces that push
inhabitant (and his machine) rather than the and pull on a membrane. The shape of this
designer. A prospective lessor or buyer, in adaptable membrane at any instant of time is
conversation with his terminal (less elaborate urban form.
than an architecture machine), can placate his
need for privacy by manipulating surfaces and In effect, the graphic manipulations from many
volumes in a given framework. Thus we have a remote terminals would manipulate the urban
situation where a general scaffolding is locally form. Each action, by designer, by resident, or
nourished by residents managing their own by satellite machine, would generate repercus-
insular needs. The concept of an architect (a sions throughout the system. In most cases,
professional) handling topical qualities and effects of a change would have local impact
each urbanite interjecting personal standards and lose force within several hundred feet of
is particularly compatible with the notion of the modification. Effects of a highway or the
“ plug-in” environments. Machines are the equivalent of the year 2000 would have more
architects for a range of qualities (using hu- global effects than a family adding a bedroom.
man or nonhuman values) that structure the But, given a machine that can interpolate quali-
environment, the architects are architects for a ties, design by-products would no longer be
piano nobile of qualities, and the householders unforeseen civil disasters.
are architects for local qualities.
As Peter Cook (1967) asks, “ Does consumer

choice of pre-fabricated living units and the like
imply that every man might become his own
architect?” Or, as another author suggests,
“ The housing modules can be bought and sold
much like cars, new or second hand.. . . The
individual units can be combined vertically and
horizontally . . . residents will buy and own their
housing modules, but rent the space they
occupy” (Hosken, 1968).
63
About Unsolicited Notes and Comments chine implicitly applies this maxim to its ob-
servation of your soap-tray. In this case the
You never actually told the machine that you machine’s notice is simple and unsolicited
were interested in lepidoptera, but the machine only in time, not content.
is finding out—from experience. It contains,
that is, a “ learning model” which stores, meas- A second way, at the other extreme of com-
ures, sorts and computes the probabilities of plexity, is through direct experience and
your interests, reactions and ways of thinking. real-world observation. For example, a
It is learning about you all right, and will soon robot might have seen bathrooms, observed
be giving you extra information about butter- soap being used, or fumbled with soap trays
flies. on its own. Such a machine might witness
Stafford Beer, “ Cybernetic Thrills and Threats” soap melting in water and from that make
the necessary chain of observations to as-
For a machine to present uninvited comments sume that. . . and so on. Even though this
upon the qualities of a design may seem pre- type of learning exceeds the scope in time
sumptuous. Yet consider that these observa- of our interests, it is important that learning
tions might well fall into the category of “ If I through groping not be underplayed or
had only thought o f . . . , ” and so forth. Further- ignored; in the distant future that is how
more, in an evolutionary system any continual machines will probably do their learning.
and machine-initiated surveillance would be
guided by a joint maturing of the architect’s A third method, more realizable in the near
ideas along with machine observation of his future, is through deduction. For example,
methods, problems, and intents. in describing the function and the environment
of the soap dish, you might have stated that
You are designing a soap tray, for example. soap melts in water and water runs downward.
Sitting at your graphic terminal with your ma- The machine, with the knowledge of the tray’s
chine, you draw an open rectangular box and geometry and the surrounding activities, could
specify that it is to be formed from a continuous deduce that water would indeed collect in
sheet of moldable plastic. All of a sudden a bell the same place as the soap. And, since soap
rings or a voice speaks or some text appears on melts, a conflict would exist; either the soap
the television screen, bringing to your atten- or the water must go elsewhere.
tion the lack of any drainage facility. How did
the machine know enough to make the observa- Such machine scrutiny is particularly interest-
tion? ing. The facts used to deduce that the collection
of water was a conflict are not necessarily
There are three sources for such unsolicited unique to the design of soap trays. Water col-
comments. First, you could previously have lection is a problem with roofs, sills, pavements,
stated very specifically that all soap trays must and so forth. In other words, after a few years
drain water. The criterion is specific. The ma- of evolutionary dialoguing, a designer and a

machine can establish a large repertoire of chronology of design development and design
low-level axioms from which the machine procedures. For example, the level of detail is
can temporally deduce high-level conflicts. time-contextual. A comment on bending mo-
ments is probably inopportune at the early
But now the question arises: Why must each stages of design. Similarly, in another time
architect struggle with indisputable facts? He context, it might be more appropriate for the
should not. Simple events— water runs down- machine to withhold a disastrous conflict until
ward, the sun rises in the east— would be built after a weekend.
into the machine’s design pedigree. Their com-
bination and association, however, must be A rate context is a fine-grain time scale. It may
unformed at the onset and must mature through be the most important of the three. Observation
deducing conflicts in the course of a partner- and recognition of work rates could attempt
ship. In other words, a built-in knowledge may to rhythmize the dialogue. The machine would
exist that, for example, children do not always try to enter a time phasing personal to and
look where they are going, and cars can kill. compatible with the designer. Some people, in
However, the constraint that children must not moments of deliberation, might enjoy a barrage
cross roads alone to get to nursery school of compliments and comments; others might
would not be an embedded maxim. demand complete silence. Moreover, this at-
titude may change with mood. Machines must
Given a set of axioms and a set of deductive discern such moods. A temporal, unsolicited
procedures, how does a machine establish comment, deduced and timely, could be antag-
the timeliness of an observation? Through con- onistic. Such prodding, however, dispels com-
text. Three types of context are particularly placency and begins to transform machine
important: an activity context, a time context, servants into machine partners.
and a rate context. Each involves ubiquitous
monitoring and observing. We must assume
that the machine continually tracks what the
designer has been and is doing.
An activity context is the easiest to implement.

Here the machine must balance between
commenting on apples when the designer is
working with pears. Only when the circulation
pattern has been ignored by the heating sys-
tem, for example, would the machine comment,
directing the designer’s attention from a con-
text of environmental to circulatory problems.
A time context is a chronology of events, a

65
Baron von Kempelen’s

Chess Player, sometimes
called “ Maelzel’s Chess
player,” after its third own-
er. The top picture comes
from a pamphlet published
in 1783 by Chrétien de
Mechel whose preface in-
cluded, “The most daring
idea that a mechanician has
ever ventured to conceive
was that of a machine which
would imitate, in some way
more than the face and
movement, the master work
of Creation. Von Kempelen
has not only had the idea,
but he has carried it out and
his chessplayer is, indisput-
ably, the most astonishing
automaton that has ever ex-
isted.” (Chapuis and Droz,
1958)The bottom picture
was published much later. It
shows the accomplice hid-
den within the automaton.
(Photographs courtesy of
Editions du Griffon, Neu-
châtel, Switzerland)

Games: Local Moves and Global Goals The Maelzel Automaton. The hoax was
achieved by the labors of a concealed dwarf
Games provide a happy vehicle for studying who observed the moves from beneath and
methods of simulating certain aspects of intel- manipulated a mechanical dummy. The need
lectual behavior; happy because they are fun, for such fraudulence has since been overcome
and happy because they reduce the problem to with computing machinery. The pioneering
one of manageable proportions. works of Claude Shannon (1956) and the later
Arthur L. Samuel, “ Programming Computers efforts of Herbert Simon (and Baylor, 1966)
to Play Games” and his colleagues have led to the develop-
ment of chess-playing machines that demon-
Games have fixed rules; gaming involves de- strate sophisticated techniques for intelligent
ception; gamers have opponents. The general decision making by strategically looking
game fabric, therefore, is not necessarily con- ahead. The approximately 1,000,000,000,000,
sonant with design. Architecture is not Mo- 000 ,000 ,000 ,000 ,000 ,000 ,000 ,000,000 ,000 ,000 ,
nopoly, Parcheesi, or checkers. Such games 000 ,000 ,000 ,000 ,000 ,000 ,000 ,000,000 ,000 ,000 ,
assume perfect information, winning is explicit, 000 ,000 ,000 ,000 ,000 ,000 ,000 ,000 ,000 ,000 ,000 ,
and the process is composed purely of sequen- 000,000,000 possible chess positions render it
tial acts— moves— governed by immutable, improbable that a calculating device can ex-
fathomable, and predefined rules. Design does haustively search all possible courses of
not have a clear-cut format; so why is “ design action. As a result, a chess-playing machine
gaming” considered avant garde and fashion- looks at local situations, looks ahead some
able? What good are games? small number of moves, and makes a specula-
tion. Such techniques are indeed relevant to
Games are a learning device for both people the construction of an architecture machine.
and machines. “ Play and learning are intimately
intertwined, and it is not too difficult to demon- However, rather than map intelligent chess
strate a relationship between intelligence and techniques into design tactics, let us concen-
play” (McLuhan, 1965). Games are models by trate on one key issue in gaming that is particu-
which or with which learning takes place. They larly relevant to design machines, that is, the
eliminate worrisome complications and per- relation of local actions to global intents. In
plexities by using artificially contrived situ- architecture the local moves are embodied in
ations. They involve the amalgamation of physical construction and destruction pro-
strategies, tactics, and goal-seeking, proc- cedures (whether explicitly executed by a de-
esses that are useful outside of the abstraction signer or implicitly by zoning laws or the like),
of gaming, certainly in design. and the global goal is quite simply “ the good
life.” In chess, the consensus is that the global
Historically, chess has been the machine’s goal to win, by taking the opponent’s king, has
baccalaureate. In 1769, Baron Kempelen con- little bearing on the local actions and the skill-
structed a fraudulent chess-playing machine, fulness of making these moves, particularly in
67
1 Cartoon that appeared in

the Manchester Evening
News on May 10, 1957.
(Courtesy of North News
Ltd., Manchester, England,
copyright Copenhagen,
Denmark)
2CLUG, Cornell Land Use

Game. CLUG is a game to
help humans learn about
planning (Wolin, 1968).
1 Each player starts with a
fixed amount of cash. The
game board is gridded
with secondary roads, utility
plants are marked, and top-
ographic features can be
added. Players risk such
real-world disappointments
as depreciation, uncontroll-
able disaster, transportation
costs, and so on. The com-
puter in this case, however,
is used only as a bookkeep-
er, keeping participants
from losing their interest
and making the game move
faster when highly paid re-
searchers or officials are
playing. (Photograph cour-
tesy of Alan Feldt, develop-
er of CLUG)

the opening and middle game. The loser can in- knows how to play, maybe everybody applauds
deed have played the better game. at the wrong time, and maybe the good life is
the wrong goal. But the thrust of the game
In architecture, the losers are rarely the play- analogy is that we do not have to answer these
ers. This is historically true, but let us assume questions in order to proceed.
that it changes and each resident can play the
game with the global goal being the good life.
The rules for achieving this goal are certainly
unclear; they vary for each person, and, as in
our Alice in Wonderland croquet game, they
are ever-changing. Furthermore, in this game
there is no coup de grace or checkmate; the
global goal has no “ utility function,” no
cost-effectiveness, no parameters to optimize.
But the chess analogy suggests that a machine

could learn to play architecture from local de-
sign pursuits and that these actions would be
draftable without an absolute definition of the
good life. A machine’s adroitness in design
could evolve from local strategies that would
self-improve by the machine testing for local
successes and failures. In other words, we are
suggesting that a machine, as well as any stu-
dent of architecture, can learn about design
by sampling the environment for cheers and
boos. For example, in a tennis match a human
spectator who is ignorant of the rules, scoring
procedures, or criteria to win can begin to dis-
tinguish good from bad play merely by observ-
ing the applause of the other spectators.
Such learning by inference can apply to the

breeding of intelligent design partners able to
discriminate between plausible patterns and
dubious forms. With a history of local punish-
ments and rewards, an adaptable machine can
evolve without a global set of values and adapt-
able rules to achieve them. Maybe nobody
69

Citation:
DOI: 10.7551/mitpress/8269.001.0001
Published: 1973

MIT Press Direct
Copyright© 1970 by
of Technology
Printed and bound

of America.

the publisher.

Book Program.

censes/byncnd/4.0/



101 Toward
The Evolution of
Architecture
Machines

URBAN5: A Postmortem or any of the other abstractions; rather they

are failings engendered either by a lack of
“ Yes. But not one of those antiquated adding knowledge or lack of forethought.
machines. It will be a superb, super-hyper-
adding machine, as far from this old piece of The first problem is due to the original over-
junk as you are from God. It will be something sight of evolution. URBAN2, the baby brother
to make you sit up and take notice, that adding and core of URBAN5, presupposed a rigid sys-
machine.. . . It will be the culmination of human tem, concluding that all the embedded assump-
effort—the final triumph of the evolutionary tions about the design process, where true
process.” (because many designers agreed), were fixed
Elmer L. Rice, The Adding Machine (because computer programs are that way,
so we thought) and were universal (because
Too often a research proposal has to establish that would be nice). After certain enlighten-
the project’s worth so completely that the ac- ments, particularly that machines can grow and
quired budget is used for the development of self-improve, some maturation processes were
an already worked out but hastily assembled appended to the system. Parts of URBAN5
idea. However, through the generous support actually do change and develop over time. In
of I.B.M. and M.I.T., URBAN5 did not suffer a patchwork manner the system can transform
from any symptoms inflicted by proposal writ- some of its internal workings. However for the
ing. There was no proposal. At first, not only most part, the authors’ underlying presupposi-
were the authors unaware of how to get there, tions about the design process exhibit no evo-
they were ignorant of where they were going. lution. URBAN5, as it stands, can never be
Work on Wednesday resulted from an achieve- denuded of the original biases that are deeply,
ment on Tuesday which appeared to be a good sometimes unconsciously, rooted in its skeletal
idea on Monday and might well be discarded structure: that architecture is additive, labels
on Thursday. The spontaneous nature of the are symbols, design is nondeterministic.
project did generate unexpected and fascinat- URBAN5 can not display an attitude that contra-
ing results. URBAN5 is not a tool, it is a toy. Its dicts these preconceptions.
impetuous nature contributed, however, to
some major shortcomings. The general structure of URBAN5 has a second
critical failure. The system feigns generality by
Of its many deficiencies, URBAN5 has four providing a multitude of specific, predetermined
notably severe shortcomings that have been design services. It has over one thousand oper-
the primary cause for abandoning it, are the ations that in combination with one another
underlying reasons for writing this book, and support a good chance for providing a desired
will be the germinal concerns of our new sys- service. But URBAN5 is not a general-purpose
tem, the architecture machine. It should be architecture machine; instead it is a barrage of
noted, however, that none of the drawbacks special-purpose (little) architecture machines.
stems from the selection of the ten-foot cube Each routine does a particular job and only
95
that job. In computer-aided design, we have The hardware sensors and effectors of URBAN5
seen that this is not appropriate. cramp those styles of conversation that are
necessary for a dialogue. The hardware has no
The third problem is context. Even though a contact with the real world except through the
contextual cross-referencing does occur within designer. URBAN5 cannot hear the designer,
URBAN5, cues are explicit statements on the it cannot see the designer, it cannot see the
designer’s behalf. The underlying modal organ- designer’s world. The designer, in turn, can
ization imposes the categorical testimony that hear only a penetrating buzz or irritating hum
“ Now I am going to do this . . . and now I am from the machine. A future system must have
going to do that.” This unequivocal demarca- overlapping modalities and a full range of sen-
tion by the designer of design context is com- sors and effectors.
pletely unacceptable. It does not admit the
necessary ambiguity and the subtle intermin- Any postmortem statement should do some
gling of contexts that are required in order to eulogizing. Even though URBAN5 was a bit
respond to a real-world medley of events. talkative and was a sloppy problem solver, it
URBAN5’s operational structure demands a was a friendly system.
repartee that relies completely and at all times
on the good judgment of the human designer.
Again, this is not acceptable. Can we assume
that he always knows what he is doing or what
he will do next? Professor Licklider’s (1965a)
solution is that “the console of the procogni
tive system will have two special buttons, a
silver one labeled ‘Where am I?’ and a gold
one labeled ‘What should I do next?’ ” Even
this solution is only partial. The machine
should answer those questions implicitly, using
context as the prime operator. Context must be
articulated through many channels, rather than
the simple depression of one or two buttons.
Problem four: URBAN5 holds hands with only

one designer and not even enough hands with
that single user. The designer has a light pen,
a keyboard, and a few buttons— a meager selec-
tion of communication artifacts. The machine,
in turn, has only a monotonous buzzer and the
cathode-ray tube upon which it can trace
monochromatic characters, lines, and points.

Languages for Architecture Machines to that inquiry. With these assumptions,

“ The world view of a culture Is limited by the URBAN5 breaks down a sentence using dic-
structure of the language which that culture tionaries that contain both words and phrases.
uses.” (Whorf, 1956) The world view of a ma- Each context (mode) has separate dictionaries.
chine is similarly marked by linguistic structure.
At the present time, however, machines have In the case of criteria specification, the inter-
denatured languages— codes. Codes are in- pretation mechanism looks for a dyadic rela-
vented for specific purposes and they follow tionship and a desired answer. A mathematical
explicit rules, whereas languages develop and summation or ratio houses the constraint. The
they evolve. But language presupposes a cul- interpretation routine passes to the conflict
ture and presumes an understanding, two mechanism one or two operands (quality, sym-
features we are not about to ascribe unequiv- bol, solid or void, generic, topographical term),
ocally to machines at this time. an operation (sum or ratio), and a desired result
(number and units). For example, from the cri-
If you are in conversation with a machine and terion, “ 50 percent of all residential units must
using a machine-oriented code, when the mech- have outdoor access,” the transformation is
anism replies, you report a “ reaction.” How- Ratio =
ever, employing a man-oriented code— a
pseudolanguage—you might attribute to the residential units with access
machine an apparent “ understanding.” total number of residential units
There are many man-oriented languages. There generic quality = 0.5

are languages of gestures and smiles, a lan- generic
guage of posture, a language of touch. The
reader should be referred to the important on- In a simpler case, when a direct question is
going work of Warren Brodey and Avery John- asked, like “ Does there exist any previous ma-
son (1969); this section is concerned with only terial to read from tape?” , the designer’s re-
one subset, a formal language that architecture sponse can be recognized with as little as five
machines must have at the very beginning or ten dictionary words. In some cases, extra
English. URBAN5 does display an apparent words might be stored in the translation mech-
understanding of English. It does use context as anism because the author’s bad spelling re-
the prime operator in translation. It has the as- quires categorization of words under proper
sumed context of architecture. Modes further and improper forms.
define context. However, throughout any English
conversation with URBAN5, the overshadowing English responses by URBAN5 are all prepro-
assumption is that the designer will talk about grammed sentences. The machine has a reper-
that which is at hand when he pushes the toire of about five hundred phrases which pro-
SPEAK button. Or if he asked a question, the vide a source of replies that can be combined
assumption is that his answer is indeed a reply with quotes from the designer. URBAN5 did not
97
achieve the interesting capability of creating its routine was developed at first as a joke. It took
own error messages from words and small a string of text, looked at a history of previous
phrases. And the reader should not suppose sentences, and randomly pieced together a
that a group of architects working on computer- response. Though the syntax was poor, STUPID
aided urban design have solved or even seri- was able to make clever replies using only
ously tackled linguistic problems. The reader statistical operations based on the kinds of
should refer to the well-documented projects words it was receiving. Later, STUPID was
of Green (et al., 1963), Bobrow (1964), Weizen given Robert Venturi’s book, Complexity and
-baum (1967), Raphael (1964), and Kellogg Contradiction in Architecture (1966), and its
(1967). replies were little more than cute. The experi-
ment does prove, however, that if you are going
URBAN5’s diversion into linguistics animates to ignore semantics, you might as well ignore
the desire for natural interaction that underlies syntax. STUPID ignored both. A dialogue be-
the entire system. Also, the crucial and not-so- tween Warren Brodey and STUPID went as
obvious role of context once again manifests follows:
itself. Linguistic studies by professional lin-
guists, like mechanical translation, have often Brodey:
ignored context because it is difficult. A machine My name is Warren Brodey
translation from Russian to English came up STUPID:
with a now classical error: “ Time flies like an My name. What is Warren Brodey
arrow” became “ Time flies enjoy eating ar- Brodey:
rows.” It is not surprising that context is so What are evolutionary environments?
important. Even in human understanding, a STUPID:
knowledge of context, usually a priori, is What are the silliest machine that ignores con-
necessary. text?
Brodey:
One of the reasons that machines do not speak Tell me, how do butterflies make love?
English is that linguists have spent so much STUPID:
time on syntax and so little time on semantics. Tell me how do butterflies will be little concern
Only after fifteen years of research does there for you.
now appear to be a general consensus that Brodey:
semantics is the key to machine understanding Why do you insist?
of natural languages. Of course, nobody knows STUPID:
how to study linguistic associations and mean- Why do butterflies make love?
ings without becoming thoroughly involved in Brodey:
syntax, noun-verb structures, and so forth. Fink
STUPID:
One exercise in languages for architecture ma- Fink, why do butterflies make love
chines is the computer program STUPID. This Brodey:

As a lover you are a torture become cumbersome to write out words like
STUPID: “ residential units” after the second or third
As a problem for you have ever seen spelling. One aspect, probably the simplest one,
Brodey: of evolutionary linguistics would permit each
Popapolartuil designer to select some anagram to refer to
STUPID: residential units if he so chooses. In effect,
Popapolartuil yes, that is Warren Brodey each designer should be able not merely to
converse in English but simultaneously to con-
Why should a machine talk English? Is it a false struct his own private shorthand or telegraphese
anthropomorphization or a true facilitation? To that might, in fact, be gobbledygook to another
assume that it is indeed a true facilitation, it is architect or another machine.
necessary first to distinguish language from
style. For example, the error message, “ Boo This all implies a congenial idiom, but it is stilt
boo, you goofed again,” as a style of com- a narrow channel of communication that ig-
munication is surely unnecessary. Conversely, nores, as we have said, the language of ges-
a number appearing on a cathode-ray tube tures and the intonations available in human
referring the designer to a manual, is equally face-to-face contact. The informal sensory and
unnecessary. With almost no effort an explana- motor augmentation of understanding is verily
tory paragraph can be displayed on a scope or “ unavailable to readers of telegrams— be they
a hard copy retrieved on a printer. A string of computers or humans” (Weizenbaum, 1967).
characters can be effortlessly stored on a disk But who designs environments by telegram?
and retrieved and displayed in less than a
twentieth of a second.
The argument, however, should not be con-

fused with the reverse case, numerical answers
unnecessarily clothed in words. An architect,
in a cost estimation procedure, for example,
probably expects the cost per square foot
rather than the comment, “ cheap,” “ okay,”
“ expensive,” or “ forget it.”
The main issue is not only English versus pidgin

English versus codes. The question is one of
language that is not only “ human discourse” but
evolutionary discourse. Learning the rigors of a
computer language should be unnecessary
except as a mental exercise or training. On the
other hand, when using written English, it might
99
1 An early speaking ma-

chine built by Sir Charles
Wheatstone and demon-
strated in Dublin in 1835
(Holmes, 1968). Even
though human speech
sounds were understood
poorly at that time, the out-
put was often passable.
However, this relied greatly
upon the skill of the opera-
tor. (Photograph courtesy of
J. N. Holmes, appearing in
Science Journal, October
1968. Redrawn from the
Journal of the Acoustical
Society of America)
2 Cazeneuve’s magic hand

appearing to write. This
fraudulent writing mecha-
nism would write answers to
questions through the skill
of the demonstrator’s ability
to substitute his own written
answer while feigning to
1 blot the wet ink. (Photo-
graph from Chapuis and
Droz, 1958. Courtesy of Edi-
tions du Griffon, Neuchâtel,
Switzerland)

Interfaces for Architecture Machines In an architect-machine relationship, perhaps

Communication is the discriminatory response the most interesting sensory interfaces are
of an organism to a stimulus (Cherry, 1957). If auditory and visual. Machines that are capable
we are to reckon with communication beyond of visual perception and speech recognition
formal rhetoric or syntax, whether English or are two of the prime targets of researchers in
computer graphics, we must address ourselves artificial intelligence. Someday, machines
to the versatility of the discriminating mechan- that can see and hear will be commonplace
ism— the interface. In this case the interface machines. Setting aside the phantasmagoria
is the point of contact and interaction between of robot designers, consider speaking to a ma-
a machine and the “ information environment,” chine that sees you— a machine with eyes and
most often the physical environment itself. ears, a machine that walks and talks. In our
present culture the thought is either frightening,
We have looked at graphic interfaces for one, foolish, or, to some, quite realistic. To our chil-
and teletypes for another, but a dialogue de- dren it will be an ordinary daily occurrence. To
mands a redundant and multichanneled con- Mortimer Taube (1961) it is offensive. To
coction of sensory and motor devices far be- Marvin Minsky (1966) it is obvious.
yond these two mechanisms. We are talking
about a total observation channel for an archi- In the meantime, extrapolations into the future
tecture machine. should recognize current problems of imple-
mentation. In the January 1967 issue of Data-
For a machine to have an image of a designer, mation, Leslie Mezie portrayed a conversation
of a problem, or of a physical environment, with a machine that could listen, and talk. A
three properties are inherently necessary: an fragment went as follows:
event, a manifestation, a representation. The
event can be visual, auditory, olfactory, tactile, Professor:
extrasensory, or a motor command. The mani- What time is it?
festation measures the event with the appro- Computer:
priate parameters: luminance, frequency, It is 8:30 p.m., Thursday, December 5, 1985.
brain wavelength, angle of rotation, and so forth. Professor:
The representation is the act of mapping the I think I would like to start with some music
information into a receptacle that is compatible today, let’s have some chamber music.
with the organism’s processing characteristics. Computer:
These three properties— event, manifestation, You listen to Telemann most.
representation— form the interface between any Professor:
two organisms. The aspect of this interface with No, something earlier.
which we are primarily concerned is the mani- Computer:
festation, encompassed primarily by a piece of What about___
hardware. Six months later, in the August issue of Data-
mation, B. W. Boehm parodied the sketch im-
101
2 4

1 The hand of “ Butterfin- planting aspects of present-day dilemmas:

gers,” the Stanford Hand-
Eye Project. (J. A. Feldman
Professor:
et al., 1969) What time is it?
Computer:
2 The M.l.T. Robot Project’s
hand, Project MAC.
I did not catch the last word. Or was it two
words?
3 The eye and the hand of Professor:
the Stanford project. This
particular eye is a scanning
What is the time?
device, a vidicon. Computer:
The dime is a copper coin worth one-tenth of a
4 Voice input to the Stan-
ford robot. Pierre Vicens of
dollar. The word derives from the Latin decern,
that project giving simple meaning.. . .
voice instructions (with a Professor:
French accent), instructions
like “ pick up the big block.”
No. No. What is the time? The time?
Computer:
It is 8:30 p.m., Thursday, December 5, 1985. We
have been having some trouble with your lin
guals recently. Sometimes I can’t tell your d’s
from your t’s. Let’s practice them. Watch the
display screen for the intonation pattern, and
repeat after me: Teddy’s daddy toted two dead
toads to Detroit.
Professor:
Teddy’s daddy toted. . . .
Nilo Lindgren’s (1965a and b) comprehensive

survey describes a host of intriguing research
efforts in speech recognition, all of which fall
into one of three catagories: the auditory sen-
sation, the acoustical disturbance freely prop-
agating through air, and a sequence of articu-
latory events in a psychological structure. The
reader should also refer to the recent works of
Bobrow and Klatt (1968), Reddy and Vicens
(1968), and Rabiner (1968).
Beyond giving a machine ears, giving a machine

eyes is extremely critical to architecture ma-
chines. Just on the hunch that a blind machine
103
1 2

1 The two diagrams repre- will have shortcomings similar to those of a

sent an interface, in this
case between man and ma-
blind architect, the relevance of a seeing ma-
chine. The left one is re- chine warrants research. Outside of the design
drawn from Nilo Lindgren’s professions, giving machines eyes is of immi-
“Human Factors in Engi-
neering” (1966b). The im-
nent importance. For instance, space explora-
portant feature is that the tion will eventually require machines that can
“human factors” thinking both see and process the seen information.
treats the entire man-ma-
chine assemblage as a sin-
This is because the remote monitoring of a
gle entity. This implies that space robot’s movements by earthlings re-
the interface is so smooth quires too much transmission time (to Mars
and so adaptable that in ef-
fect it does not exist.
and back, for example), and a machine would
crash into that which it is told to avoid only be-
2 The illustration is redrawn cause the message to stop might arrive too late.
from a reinterpretation of
the above by Avery John-
More domestic applications involve visual dis-
son. Still considered as a crimination of simple objects. Eventually, ma-
single entity, the man-machines will package your purchased goods at
chine assemblage has a
more active interface. In
the counter of your neighborhood supermarket.
this interpretation, the inter-
face has local computing Oliver Self ridge (and Neisser, 1963) is credited
power and can thus exhibit
a behavior. This implies a
with the founding works in pattern recognition.
continuous sensing and ef- His mechanism, PANDEMONIUM, would ob-
fecting mechanism, and it is serve many localized visual characteristics.
the behavior of this device
that is observed by both
Each local verdict as to what was seen would
higher-order processors. be voiced by “ demons” (thus, pandemonium),
and with enough pieces of local evidence the
3 SEEK. This device is a
homemade sensor/effector
pattern could be recognized. The more recent
built by architecture stu- work of Marvin Minsky and Seymour Papert
dents. The device has multi- (1969) has extensively shown that solely local
ple attachments (magnets,
photocells, markers, etc.)
information is not enough; certain general ob-
which it can position in servations are necessary in order to achieve
three dimensions under complete visual discrimination.
computer control. It is antic-
ipated that the mechanism
will pile blocks, carry TV At present, these works are being applied to
cameras, observe colors, architectural problems as an exercise
and generally act as a peri-
pheral device for student
preliminary to the construction of an archi-
experiments in sensors and tecture machine. Anthony Platt and Mark
effectors that interact with Drazen are applying the Minsky-Papert eye to
the physical environment.
the problem of looking at physical models
(Negroponte, 1969d). The interim goal of this
105
5 6 7 8

1 The M.l.T. Minsky/Papert parallelogram indicates a

eye. In this case the eye is complete surface that can
an image dissector, a ran- be used as “strong evi-
dom-access device that dence” to place others in
does not scan back and space.
forth but rather goes to dis-
crete positions under com-
puter control. This was the
eye used for the Platt/Dra
zen vision experiment under
the supervision of Seymour
Papert.
2 3
2 Some vision problems —
reflections and tone
changes. Note that the top
surfaces of the front lower
cubes are a light gray, while
the rear upper one has a
black top surface. In other
words, in such lighting the
orientation of a surface can-
not be assumed from its
gray tone.
3 Other vision problems —

4 depth of field and shadows.
4 More problems— discon-

nected bodies.
5 A typical model presented

to the eye.
6 Printer output of the light

intensities.
7 Contours of similar inten-

sities.
8 A cathode-ray tube dis-

play of the discovered con-
tours. The “ noise” is due
to both bad lighting and a
poor choice of contours.
9 The seen lines. These

would be the lines seen un-
der ideal, noise-free condi-
tions.
10 Minimal surfaces. A
9 10
107
2 3

1 GROPE groping on the Ur- exercise is to observe, recognize, and de-

ban Atlas map of New
York’s residential popula-
termine the “ intents” of several models built
tion density. from plastic blocks. Combined with Platt’s
previously described LEARN, this experiment
2 The old GROPE.
is an attempt at machine learning through ma-
3 The new GROPE. The chine seeing. In contrast to describing criteria
slight glow beneath GROPE and asking the machine to generate physical
is from three little lights that
illuminate the area for the
form, this exercise focuses on generating cri-
fifteen photocells. It is inter- teria from physical form.
esting to note that, like most
Architecture Machine proj-
ects, GROPE started as a
A second example of interfacing with the real
toy costing $15. Even world is Steven Gregory’s GROPE (Negroponte
though it has evolved into a et al., 1969b). GROPE is a small mobile unit that
major experiment, its circui-
try and hardware have cost
crawls over maps, in this case Passonneau and
less than $80. Wurman’s (1966) Urban Atlas maps. It employs
a low-resolution seeing mechanism constructed
with simple photocells that register only states
of on or off, “ I see light” or “ I don’t see light.”
In contrast to the Platt experiment, GROPE
knows nothing about images; it deploys a con-
troller that must be furnished with a context and
a role (as opposed to a goal: play chess as
opposed to winning at chess). GROPE’s role is
to seek out “ interesting things.” To determine
future moves, the little robot compares where
he has been to where he is, compares the past
to the present, and occasionally employs ran-
dom numbers to avoid ruts. The onlooking
human or architecture machine observes what
is “ interesting” by observing GROPE’s behav-
ior rather than by receiving the testimony that
this or that is “ interesting.” At present, some
aspects of GROPE are simulated and other
aspects use the local computing power on
GROPE’s plastic back. GROPE will be one
of the first appendages to an architecture ma-
chine, because it is an interface that explores
the real world. An architecture machine must
watch devices such as GROPE and observe
109
1 Before the Architecture

Machine Project had its own
dedicated computing pow-
er, aspects of GROPE were
simulated on the ARDS dis-
play. The four illustrations
represent a sequence that
traces GROPE’s path
through an internal machine
representation of Urban At-
las data for Boston. Note
that by the fourth frame
2 GROPE has “scrubbed
out” two areas of the upper
right. It turns out that this is
Boston’s downtown water-
front, indeed an “ interest-
ing” area of the map.
2 A photographic overlay of
GROPE’s path with a road
map of Boston.
3 An overlay with “ personal

income” data.
3
4 An actual numerical dis-
play of the “personal in-
come” data.
5 An overlay with “ land

use” and “residential densi-
ty.”
1 5

their behavior rather than listen to their Architecture Machines Learning Architecture
comments.
There is no security against the ultimate de-
But why not supply the machine with a coordi- velopment of mechanical consciousness, in
nate description of the form on punch cards the fact of machines possessing little con-
and proceed with the same experiment? Why sciousness now . . . reflect upon the extra-
must a machine actually see it? The answer is ordinary advance which machines have made
twofold. First, if the machine were supplied a in the last few hundred years, and note how
nonvisual input, the machine could not learn slowly the animal and the vegetable kingdoms
to solicit such information without depending are advancing.
on humans. Second, it turns out that the com- Samuel Butler, Erewhon
putational task of simply seeing, the physi-
ology of vision (as opposed to the psychology When a designer supplies a machine with
of perception) involves a set of heuristics that step-by-step instructions for solving a specific
are apparently those very rules of thumb that problem, the resulting solution is unquestion-
were missing from LEARN, that made LEARN ably attributed to the designer’s ingenuity and
a mannerist rather than a student. labors. As soon as the designer furnishes the
machine with instructions for finding a method
It seems natural that architecture machines of solution, the authorship of the results be-
would be superb clients for sophisticated comes ambiguous. Whenever a mechanism is
sensors. Architecture itself demands a sensory equipped with a processor capable of finding
involvement. Cardboard models and line a method “ of finding a method of solution,”
drawings describe some of the physical and the authorship of the answer probably belongs
some of the visual worlds, but who has ever to the machine. If we extrapolate this argu-
smelt a model, heard a model, lived in a mod- ment, eventually the machine’s creativity will
el? Most surely, computer-aided architecture be as separable from the designer’s initiative
is the best client for “ full interfacing.” De- as our designs and actions are from the peda-
signers need an involvement with the sensory gogy of our grandparents.
aspects of our physical environments, and it
is not difficult to imagine that their machine For a machine to learn, it must have the im-
partners need a similar involvement. petus to make self-improving changes, to
associate courses with goals, to be able to
sample for success and failure, and to be
ethical. We do not have such machine capa-
bilities; the problem is still theoretical, still of
interest primarily to mathematicians and
cyberneticians.
A 1943 theorem of McCulloch and Pitts states

111
1 2
3 4
5 6

1 The Interdata computer, at that a machine constructed with regenerative

present the nucleus of the
Architecture Machine Proj- loops of a certain formal character is capable
ect. (Photograph courtesy of of deducing any legitimate conclusion from a
the Interdata Corporation) finite set of premises. One approach to such a
2 Interdata’s mother-boards faculty is to increase the probability of mean-
and daughter-boards. These ingfulness of the output (the design) generated
panels hold the circuitry from random or disorderly input (the criteria).
and plug into a chassis. This
simplicity permits rapid in- Ross Ashby (1956) states, “ It has been often
terfacing of peripheral sen- remarked that any random sequence, if long
sors and effectors. (Photo- enough, will contain all answers; nothing pre-
graph courtesy of the Inter-
data Corporation) vents a child from doodling: cos2X +
sin2X = 1.” In the same spirit, to paraphrase
3 The Architecture Ma- the British Museum/chimpanzee argument, a
chine’s first technician.
group of monkeys, while randomly doodling,
4 SEEK and its controller. can draw plans, sections, and elevations of
all the great works of architecture and do this
5 The Architecture Machine
configuration including the
in a finite period of time. As the limiting case,
three ARDS. (September 1, we would have a tabula rasa, realized as a
1969) network of uncommitted design components
6 Processor, expansion
or uncommitted primates. Unfortunately, in
chassis, and various con- this process our protagonists will have built
trollers. Levittown, Lincoln Center, and the New York
Port Authority Towers.
Surely some constraint and discrimination is

necessary if components are to converge on
solutions within “ reasonable” time. Compo-
nents must assume some original commitment.
As examples of such commitment, five particu-
lar subassemblies should be part of an archi-
tecture machine: (1) a heuristic mechanism,
(2) a rote apparatus, (3) a conditioning device,
(4) a reward selector, and (5) a forgetting
convenience.
A heuristic is a method based on rules of

thumb or strategies that drastically limit the
search for a solution. A heuristic method does
not guarantee a solution, let alone an optimal
113

1 The Architecture Ma- one. The payoff is in time and in the reduction
chine’s punish/reward and
BLAB, a rudimentary audio of the search for alternatives. Heuristic learning
output device. is particularly relevant to evolutionary ma-
chines because it lends itself to personaliza-
tion and change by talking to one specific
designer, overviewing many designers, or
viewing the real world. In an architecture ma-
chine, this heuristic element would probably
be void of specific commitment when the
package arrives at an office. Through architect
sponsored maturation, a resident mechanism
would acquire broad rules to handle excep-
tional information. The first time a problem is
encountered, the machine would attempt to
apply procedures relevant to similar prob-
lems or contexts. Heuristics gained from
analogous situations would be the machine’s
first source of contribution to the solution of a
new problem.
After repeated encounters, a rote apparatus

would take charge. Rote learning is the ele-
mentary storing of an event or a basic part of
an event and associating it with a response.
When a situation is repeatedly encountered, a
rote mechanism can retain the circumstance
for usage when similar events are next en-
countered. In architecture, this repetition of
subproblems is extremely frequent: parking,
elevators, plumbing, and so forth. And again a
rote mechanism lends itself to evolutionary ex-
pansion. But, unlike a heuristic mechanism,
this device would probably come with a small
original repertoire of situations it can readily
handle.
Eventually, simple repetitous responses be-

come habits, some good and some bad. More
specifically acclimatized than a rote apparatus,
115
1 Diagram taken from Mar-

vin Minsky and Seymour
Papert’s Perceptrons
(1969).
2 Projections of the pres-

ent Architecture Machine
configuration.

a conditioning mechanism is an enforcement fallacious. Information can assume less signifi-

device that handles all the nonexceptional cance over time and eventually disappear—
information. Habits, not thought, assist hu- exponential forgetting. Obsolescence can occur
mans to surmount daily obstacles. Similarly, through time or pertinence. A technological in-
in a machine, beyond rote learning, design novation in the construction industry, for exam-
habitudes can respond to standard events ple, can make entire bodies of knowledge
while the designer, the heuristic mechanism, obsolete (which, as humans, we tend to hate
and the rote apparatus engage in the problem- surrendering). Or past procedures might not
solving and problem worrying (Anderson, satisfy environmental conditions that have
1966) aspects of design. Each robot would changed over time, thus invalidating a heuristic,
develop its own conditioned reflexes (Uttley, rote response, or conditioned reflex.
1956). Like Pavlov’s dog, the presence of
habitual events will trigger predefined re- These five items are only pieces of an archi-
sponses with little effort until the prediction tecture machine; the entire body will be an
fails; whereupon, the response is faded out by ever-changing group of mechanisms that will
frustration (evolution) and is handled else- undergo structural mutations, bear offspring
where in the system. (Fogel et al., 1965), and evolve, all under the
direction of a cybernetic device.
A reward selector initiates no activities. In a
Skinnerian fashion (B. F. Skinner, 1953), the
reward mechanism selects from any action
that which the “ teacher” likes. The teachers
(the designer, the overviewing apparatus, the
inhabitants) must exhibit happiness or disap-
pointment for the reward mechanism to oper-
ate. Or, to furnish this mechanism with direc-
tion, simulation techniques must evolve that
implicitly pretest any environment. The design
of this device is crucial; bad architecture
could escalate as easily as good design. A
reward selector must not make a machine the
minion or bootlicker of bad architecture. It
must evaluate, or at least observe, goals as
well as results.
Finally, unlearning is as important as learning

(Brodey, 1969c). The idea of “ its [the compu-
ter’s] inability to forget anything that has been
put into i t . . . . ” (A. Miller, 1967) is simply
117

Citation:
DOI: 10.7551/mitpress/8269.001.0001
Published: 1973

MIT Press Direct
Copyright© 1970 by
of Technology
Printed and bound

of America.

the publisher.

Book Program.

censes/byncnd/4.0/



The Architecture Machine - MIT

Uploaded by

Copyright:

Available Formats

The Architecture Machine - MIT

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Architecture Machine - MIT

Uploaded by

Copyright:

Available Formats

This is a portion of the eBook at doi:10.7551/mitpress/8269.001.

Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282286/9780262368063_f000100.pdf by Xin Sun on 09 November 2024

Humanism through Intelligent Machines context or regardless of context. It follows that

The Spanish colonials laid

The two illustrations are of

Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282286/9780262368063_f000100.pdf by Xin Sun on 09 November 2024

Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282286/9780262368063_f000100.pdf by Xin Sun on 09 November 2024

1 Trick automaton feigning

2 The computer at home is

Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282286/9780262368063_f000100.pdf by Xin Sun on 09 November 2024

What needs to be articulated, regardless of the

In this book, there is no distinction between

By: Nicholas Negroponte

Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282286/9780262368063_f000100.pdf by Xin Sun on 09 November 2024

Printed and bound

All rights reserved. No part

Open access edition funded by

The text of this book is licensed

ISBN 0 262 14008 X (hardcover)

Library of Congress catalog

Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282286/9780262368063_f000100.pdf by Xin Sun on 09 November 2024

Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282287/9780262368063_c000000.pdf by Xin Sun on 09 November 2024

Prelude to an Architect-Machine Dialogue is the symbiosis. It evolves through mutual

Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282287/9780262368063_c000000.pdf by Xin Sun on 09 November 2024

But, the tête-à-tête must be even more direct

Imagine a machine that can follow your design

Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282287/9780262368063_c000000.pdf by Xin Sun on 09 November 2024

The sequence of photo- assimilate your conversational idiosyncrasies.

The overlaying of a specific design character

One might argue that we are proposing the

Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282287/9780262368063_c000000.pdf by Xin Sun on 09 November 2024

The three photographs However, we have indicated that the maturation

Writing machine made

The device was built after

Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282287/9780262368063_c000000.pdf by Xin Sun on 09 November 2024

Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282287/9780262368063_c000000.pdf by Xin Sun on 09 November 2024

Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282287/9780262368063_c000000.pdf by Xin Sun on 09 November 2024

manipulated. Such a system was developed by Computer-Aided versus Computerized

It is commonly suggested that by furnishing a

Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282287/9780262368063_c000000.pdf by Xin Sun on 09 November 2024

Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282287/9780262368063_c000000.pdf by Xin Sun on 09 November 2024

1 Leon Groisser at home in composed of additive and subtractive pieces of

The two outlooks are not necessarily contradic-

The principle is simple and, in computer-aided

A problem-oriented language is a high-level

Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282287/9780262368063_c000000.pdf by Xin Sun on 09 November 2024

2 The Stanford Research In-

Downloaded from http://direct.mit.edu/books/oa-monograph/chapter-pdf/2282287/9780262368063_c000000.pdf by Xin Sun on 09 November 2024

This overviewing would be achieved through a