Technology|Architecture + Design

ISSN: 2475-1448 (Print) 2475-143X (Online) Journal homepage: http://www.tandfonline.com/loi/utad20

Drawing Disruptions: Representing Automated

Distortions of Multi-Perspectival Form

Joshua M. Taron & Matthew Parker

To cite this article: Joshua M. Taron & Matthew Parker (2017) Drawing Disruptions: Representing
Automated Distortions of Multi-Perspectival Form, Technology|Architecture + Design, 1:2, 219-230,
DOI: 10.1080/24751448.2017.1354624

To link to this article: https://doi.org/10.1080/24751448.2017.1354624

Published online: 28 Nov 2017.

Submit your article to this journal

Article views: 24

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=utad20
 219

Joshua M. Taron
University of Calgary

Matthew Parker
University of Calgary

Drawing Disruptions:
Representing Automated
Distortions of Multi-
Perspectival Form

PEER REVIEW / SIMUL ATIONS

SIFT descriptors exist within the lineage of flat-
tened architectural representation but pose new
disciplinary ambiguities for architecture given
their subscription to particular algorithmic sub-
routines that unflatten images for the purpose of
producing a singular image of the environment. A
selection of contemporary architectural projects
is used to identify different modes of flatness
in architectural representation that SIFTs both
subscribe to and complicate given the disrup-
tions they induce within the objects they image.
Particular attention is given to the algorithmic
components within SIFTs that track, agitate, and
exploit differences between multiple images of
individual objects. A small architectural project is
subjected to SIFTs in an effort to experiment with
their otherwise hidden aesthetics of disruption.
 Drawing Distruptions 220

v Figure 1 (Previous …algorithms are the steps of a process of abstracting mathematical forms and individual
page). Vector-flow-field. facts, and that they are also conceptual prehensions of eternal objects, or potentialities,
w Figure 2. Difference- that determine the arrival of changing conditions in the process of calculation.
of-Gaussian (GoG) +
keypoint identification
diagram. —Luciana Parisi, Contagious Architecture: Computation, Aesthetics, and Space1

Introduction

Drawings as Algorithms
This paper begins with the proposition that architectural drawings are algorithms and always
have been. Architects have historically used them as a kind of cognitive prosthetic2 for measur-
ing, calculating, thinking, projecting, and imagining architecture and its effects. Terminology
surrounding architectural drawings has provided a way of describing and addressing algorith-
mic behaviors and, by extension, frame the questions architecture might be able to address.
The plan, for example, is one kind of algorithm that lends itself well to circulation, program-
ming and partition diagrams by distorting the form such that it can be read orthographically.
Perspective drawing enables individual point of view to be incorporated thus integrating that
distortion into the drawing itself. Architecture has also adopted digital modeling and auto-
mated processes into its set of representational techniques. As an algorithmic arms race moves
forward attempting to contend with the worlds complexity while simultaneously contributing
to it, new distortions have undoubtedly manifested whether we know it or not. But how and by
what means might architecture address its own representation in automated environments?
And is there evidence that architectural drawing itself may have already become automated?
And if so, where are such processes taking place and can they be expressed through the repre-
TAD 1 : 2

sentation of architectural form?

Flattened Representations and the Viral Distortion

Within algorithmic environments, architectural images function as a medium through which
architectural objects exist and experience transformation. This demands a new mode of criti-
cism from ones that have preceded it. Historically, architectural representation has focused
on the problems of formal accuracy and human perception. This is perhaps best documented
by Alberto Perez-Gomez and Louise Pelletier who identify through an exhaustive lineage of
drawing “an invisible perspective hinge [that] is always at work between these common forms
of representation and the world to which they refer.”3 This gap between the real and the per-
ceived has certainly animated the aforementioned representational technologies in their pur-
suit of representing more complex and information-laden environments, or what we might
describe as a desire for high-fidelity abstraction. Echoing Walter Benjamin and a host of crit-
ics4 that would follow, this proposition runs in parallel to the fact that these increasingly com-
plex architectural representations and the nature of their legibility and accessibility replace
and supplant the reality they are meant to reproduce. Citing the gridded globe in Ptolemy’s
Geographia, Perez-Gomez and Pelletier go on to warn against the “fallacy of a progressive
history of the image,” doing so in the name of preserving a kind of space that “allow[s] us to
recognize the purposefulness of human existence in the plot of the tragedy of our uncertain
destiny.”5 This seems a rather romantic and naïve claim that on one hand assumes that any
representational calculation that excludes the human perspective (literally) is necessarily
invalid while simultaneously failing to acknowledge the magnitude and intensity of force that
algorithmic images (and image-models) of the world exert when reified.6 They also fail to posit
that non-human entities might be capable of producing perspectives themselves which in turn
exerts a force that might orient the world toward them.
The abstract pattern most identifiable with high-fidelity abstraction has been that of flat-
tening information in order to make it accessible. This is as much the case for Ptolemy’s globe
as it is for cinema7 as it is for architecture; that is, flattening multi-dimensional problems onto
a sheet of paper or computer screen. This reading is significant in that architectural represen-
tation has historically presumed itself to be flat and is instead discussed in terms of the appar-
ent dimensionality a particular type of drawing or image is capable of expressing. For example,
a single point perspective drawing might conventionally be understood as being capable of
 TARON & PARKER 221

scale n+1
(2nd octave)

scale

scale n
x - n - (n+1) ... - (n+7) = magnitude
(1st octave)

Gaussian scale Difference of Gaussian (DoG) keypoint selected if pixel DoG keypoint orientation average vector orientation
is < or > than 26 nearest neighbors based off nearest and magnitude at the
neighbors keypoint of SIFT

expressing three dimensional spaces and objects. Alternatively a address all questions at once but in combination form a kind of
flattened reading argues that the space or object is subjected to complex image of a project. These drawings were born out of
a specific set of instructions, in this case a single-point perspec- an evolution of methods that overcame limitations of the ones
tive, that calculate it into a 2-dimensional form; or in other words, that preceded them. Plan and section were synthesized into
an algorithm that solves the problem of flattening space as per- the oblique drawing so both can accurately be seen at once.
ceived from a specific point of view. Therefore, instead of evaluat- Perspective emerged as a way of incorporating spatial depth into
ing a particular technique of representation based on the thing it two dimensions. Physical models resulted from the translation of
is attempting to represent, we posit that modes of representation 2-dimensional drawings into 3-dimensional space, and so forth.

PEER REVIEW / SIMUL ATIONS

be addressed based on the specificity of the distortions produced Given that the drawing or model is produced in advance of con-
through algorithmic flattening. As such, the utility of flattening is structing the building itself, both drawing and modeling have held
that it either prepares information for algorithmic consumption or a privileged position within the discipline of architectural design
is a transformative result of having been subjected to an algorith- as they constitute the things that architects actually make (and
mic process. In particular, the ubiquity of machine vision (in this sometimes own) while the making of buildings is done at arm’s
case, Scale-Invariant Feature Transform [SIFTs] and their insatia- length. Thus, drawings as a representational tool operate like a
ble appetite for latent information embedded in gathered image kind of cognitive prosthetic that connect architects to the modes
sets) begins to pair with architectural form that migrates freely of thought that make buildings possible.
through and across a variety of information platforms. When As digital technologies have evolved, the architectural model
looking inside SIFT code, one discovers a disruptive relationship has emerged as the dominant design document given its ability
between the already-flattened image of architectural form and to synthesize project specifications with auto-generated ortho-
the algorithmic processes of SIFTs where the pair autonomously graphic drawings while mediating processes of communication
produces variants of the form-image through subroutines of vec- between parties into a single centralized building information
toral mapping at the scale of individual pixels. model. Consequently, architectural drawings are being relegat-
While SIFTs do not care and are not necessarily aware that ed to the status of byproduct—generated in order to serve as a
architectural objects are explicitly within their purview, we will body of information from which a project’s form can be explic-
look at the way in which architecture might behave in SIFTed itly communicated, analyzed for code compliance, used for cost
environments as well as exploring the way in which SIFTs might estimation, and referenced as a set of instructions for construc-
be simultaneously producing and resisting a limit condition within tion—revealing a productive albeit sanitized oscillation between
flattened fields of representation. This paper is interested in how drawing and model as active agents within the design process. As
representational distortion produced through multi-perspectiv- a result, while many might view the model as a tool or extension
al environments agitates the architectural image and how actual of human thought, in many ways it has functioned to displace it.
instances of buildings might already be contributing to that condi- In addition to the influence of the digital model upon architec-
tion through their automated consumption by SIFTs. Toward that tural drawings, conventions of architectural representation have
end, the paper explores how the algorithmic rendering of archi- become constrained by the trades and disciplines related to but
tecture in flattened/pixelated environments can be both framed outside of the discipline of architecture. This perhaps accounts
as a process of and critically engaged through drawing, functioning for the primary agenda behind BIM development, which is an
at one scale to disrupt and agitate for the purpose of flattening attempt to recentralize the architectural project into a single, full-
and smoothing at another. resolution, information-rich model chiefly managed by the archi-
tect containing descriptions and positions for every component
Drawing Territories for the purposes of addressing, simulating, and managing the
complexity of the design and its construction. This has produced
On the Marginalization of Drawing a highly constrained environment whereby architectural drawing
Architectural drawing has historically been broken down into has all but eliminated representational distortions, instead func-
a series of isometric orientations that alone do not attempt to tioning as a way of maximizing fidelity between the model and the
 Drawing Distruptions 222

built form. Where the architectural drawing has historically been

used as a shorthand for the calculation of complex forms, the
model has become equally, if not more complex than the actual
building it is meant to describe. As a result, if architects are occu-
pied making models of buildings, who or what if anything is mak-
ing architectural drawings? Is there a way to reanimate drawing
such that it regains its status as a medium of critical authority; if
not through the architect, then perhaps through the architectural
object itself?

SIFT Algorithms: Pixel-Vector Agitations

SIFT algorithms see the world by computationally deconstruct-
ing images into collections of unique features that can be iden-
tified, organized and matched across multi-perspectival image
sets. First developed by David Lowe, 8 they seek to mitigate dif-
ferences between pairs of images based on variant spatial points
of perspective. This is achieved by identifying specific image fea-
tures invariant to scaling, rotation, changes in illumination, and 3D
camera viewpoints.9 SIFTs assign descriptors to each pixel of an
input image, encoding them with contextual information through
processes that reduce images to a finite set of distinctive key
points (Figure 2). These key points facilitate the filtration of visual
clutter/noise within the image while providing a high probability
of feature matching and correlation across images (Figure 3). The
strong matching capabilities and computational stability of SIFTs
TAD 1 : 2

are typically used for image retrieval, image stitching, machine

vision, object and gesture recognition, match moving, and, not
coincidentally, the digital (re)construction of Google Earth’s
3-dimensional virtual environments. This produces an indetermi-
nate exchange of information where the object codifies the SIFT
algorithm while the algorithm inhabits and recodifies the architec-
tural image-object. Paradoxically, the SIFT agenda is one of flat-
tening the real world into a seamless virtual environment, but in
r Figure 3. Gaussian order to do so it must draw and thereby unflatten pixel fields via
Scale Space (GSS) diagram. dynamic vector fields (Figure 4).
GSS of an image along
with the associated SIFT
descriptors identified at Architectural Precedents: Drawing Flat
each level. As the Gaussian
blur increases within the
scale space, fewer SIFT Flattening Complexity
descriptors are deployed. As the world has become more complex, technologies of abstrac-
tion-through-flattening have become commonplace. Our desktop
w Figure 4. Keypoint computers abstract the content of our actual desks into a screen;
localization sample region. architectural complexity has been abstracted and compressed
into the surface of the building envelope.10 Urban spaces are con-
structed and mediated by screens and spaces themselves flat-
tened into programmed “event urbanisms,” expressing the rather
sterile if not entirely cynical claim “that people come together
around planned activities: celebrations, victory parades, concerts,
corporate cocktail parties and the like – [a] decidedly a top-down
notion to program and market the spaces of the city.”11 Reactions
to, symptoms of, or even complicity with event urbanism (inten-
tional or not) have been well documented through the conceptual
representation of form through the territory of the architec-
tural surface and its sensational effects.12 Lately a kind of critical
authority toward flatness has begun to emerge whereby architec-
ture attempts to translate flattened surfaces back into something
volumetric, massive, and/or material. Within these projects, we
TARON & PARKER 223

image

identified SIFT descriptor spatial histogram of gradients at pixel level 3D vector average at SIFT

might begin to define specific distortional aspects that contextu- produced by image making in architecture.”15 Both projects
alize SIFTs within architectural discourse. mobilize drawings of the component parts of renderings against
one another to produce images of abject, seemingly impossible
Tracing Extractions: Flat is More forms using visualization tools meant to accurately represent real
Kristy Balliet’s Possible Volumes13 subtracts the material surfaces objects (Figure 6) while also approaching rendering as a kind of

PEER REVIEW / SIMUL ATIONS

of Wright’s Larkin building in order to excavate its spatial vol- drawing subroutine aggregator.
umes through purposely and [perhaps] ironically flat axonometric The affective potency of these render-drawings is acute, imme-
drawings arguing that the volumes “shift the center,” “animates” diately expressing a distorted and somewhat unreal strangeness
and “distributes detail” within the spaces of the project. The axo- about them. They demonstrate the critical capability of draw-
nometric drawing serves as a kind of double agent. On one hand, ings to exhibit agency when confined within models, in this case
making the visualization of the volumes implies solid material the model being a particular software environment. The weird-
assemblies; whereas, at the same time, tracing the spatial logic of ness that the render-drawings produce is not at all an error to be
the project through its material edges would otherwise prevent culled but rather an effect made possible through an exploitation
such a reading (Figure 5). and exaggeration of the algorithmic subroutines within the model.
The possibility of the volumes critically lies first in the built form Second, the distortions are more specific than the general meth-
with the drawing only afterward employed as an analytical tool od of drawing or the software environment alone and instead
as opposed to positioning the drawing as a description in advance use geometry to reveal the uniqueness of the model’s algorith-
of the project. The architectural projection takes place not in mic components in communication with one another. The effects
advance of, but rather as a result of, the Larkin building itself. of the drawings are generated by compositing multiple (typically
The use of the axonometric drawing is particularly significant in isometric) orientations into the projection of a single object as is
that it replaces perspectival orientation with a directional orienta- the case in Possible Table. Because of the mathematical accura-
tion, thus extending observation from a point to a general direc- cy of isometric drawing as opposed to the distortions produced
tion for the purpose of analysis. The axonometric as an algorithm through perspective drawing, the flatness of the drawing itself
could then be described by this directional distortion. Possible elicits not-quite-flat readings and results.
Volumes positions the architect as one who enables buildings to Perhaps the most interesting aspect of the work is the
express themselves by committing to a single directional orien- approach toward architecture that it necessitates, calling for an
tation and then exploiting its precise distortions in order to gen- “examination of the construction of the image as careful as that
erate a critical analysis.14 The building is already doing a form of which has already been applied to the object.”16 What we can see
work: the axonometric drawing, therefore, functions as a process here is a shift of attention once trained upon objects to the way in
that mobilizes flatness against itself yielding a difference of distor- which its image is constructed, which implies that these process-
tion through which the project can then be articulated. es of construction already exist and are in action. What remains
for the architect is the task of discovering through experimenta-
Render-Drawings: Flat on Flat tion what parts might be and drawing out their effects. Similar to
While the axonometric drawing’s familiarity and legibility makes Possible Volumes, the processes that mobilize the work are already
it somewhat difficult to frame in terms of an obvious distortion, at play and, yet, the task of the architect is to catch up to an archi-
there is perhaps a more recognizable instance of representational tecture that is already at play.
distortion in a series of explorations undertaken by First Office. In
Couch on Couch (2014) and Possible Table (2013) Andrew Atwood Models that Draw: Flat is the New Grid
and Anna Neimark unpack technical images by “exploding them The decoupling of representation from objects forms a strange
into [their] various components, by visualizing the abstractions problem for architects but one that may be best suited to the
 Drawing Distruptions 224

r Figure 5. Axonometric drawings of the Larkin Building’s possible volumes. r Figure 6. Render-drawings of couch on couch and possible table.

discipline at a moment where buildings seem to be divesting them- and ultimately adds or discovers latent dimensionality to represent-
selves from architecture. If architecture’s techniques of drawing ed things. Second, isolated subroutines can be mobilized against
are not going to be effectively used to produce buildings, then per- one another to undermine the opaque and inaccessible by sub-
haps they can be deployed upon the images of buildings instead. jecting them to the complexities that flatten them in the first place.
Besler and Sons’ On the Resolution Frontier (2016)17 begins to Third, when examining automated or autonomous drawing pro-
outline a design brief for such a project (Figure 7). By tracing an cesses, flattening is both something to resist as well as a technique
emergent boundary only evident within the Google Earth model worth employing at various scales and degrees of resolution. They
where 3-dimensional mapping abruptly ends for no apparent rea- may, in fact, necessitate one another the same way that figure-
son, Along the Resolution Frontier speculates upon an existing edge ground or solid-void relationships articulate one another. Fourth,
“at which algorithmic geomodeling ends and the handmade model flatness functions as a new kind of field, territory or grid that can
is allowed a tentative stay of execution, until, inevitably, Google’s then be inhabited, agitated, and stimulated, even recursively by the
scanning efforts envelope the entire surface of the Earth.”18 very algorithms that are bent on making the world as flat and seam-
What this project begins to reveal is the degree to which draw- less as possible. Lastly, distortions not only qualitatively describe
TAD 1 : 2

ing might be embedded within models and how drawing itself a technique of representation but also locate distinct and recog-
might exert its agency once again from within the algorithms that nizable moments within otherwise flat environments. This may
constitute those models. Besler and Besler go on to claim that “As produce cognitive maps capable of addressing, visualizing, or aes-
digital representations become the predominate imagery shaping thetics beginning at seams generated through distortion.
our routine experiences and understanding of the environments
around us, our capacity to influence, edit, reject, or undermine the Drawing with the Architectural Object
objectivity and apparent immutability of the depiction becomes
critical.”19 Yes, undermining immutability should be a paramount Models as Incomplete Drawings
task; no disagreement here. But is naming semi-observable condi- Sukkah was designed and built by the Loboratory for Integrative
tions for the purpose of discussion enough? And if not, to answer Design during the Summer of 2016. This object was subjected
the call made by Besler and Besler, what would constitute an to the SIFT algorithm for the purpose of experimenting with the
action substantial enough to “appropriate the humorless sobriety drawings it would yield. (Figure 8 through Figure 10). Because the
of Google Earth’s visual renderings”?20 project was designed using stereotomic projection and has a dis-
The primary condition within both Resolution Frontier and tinctive elevational legibility, the SIFT was expected to handle the
Google earth is an insistence on supporting and even glorifying geometry of the object with reasonable accuracy. The form was
the perspectival distortion through a singular immersion within generated by projecting the three Hebrew letters for "sukkah"23
its virtual space. Here everything that is present at the surface is across three elevations of the project - their intersections carving a
drawn together into a single navigable image-space, becoming so subtraction from the otherwise solid initial cube (Figure 11). Given
common place that users rarely pause to interpret its presence the deliberate symbolism of the design, subjecting it to SIFTs would
or nature as a governing algorithm. However, as Felicity Scott provide the possibility of visualizing the distortions created by mul-
points out, “Google Earth is much more than just a virtual globe tiple points of simultaneous observation. To simulate this effect,
comprised from multiple layers of satellite imagery, aerial photo- the sukkah was rotated around a single point in spacewhile a fixed
graphs, GIS/GPS data and government documents.”21 And, going camera snapped images at regular intervals (Figure 12). These
even further, Google Earth is more than simply a collection of images were then paired at various frequencies in order to explore
snapshots from outer space but a collection of pixels or “pieces of the limits of the SIFT to stitch them back together.
data linked within digital networks to other pieces of data in other
information systems.”22 Vector-flow-fields
The SIFT flow algorithm manufactures a unique displacement
Attributes of Flat Distortions vector (UV coordinate space, that defines the direction of the
What can we learn from interpolating these three instances in 3D vector) for each pixel of an input image relative to its corre-
combination? First, flattened representation produces degrees of late position within its adjacent image – an explicit measure of
access, legibility and clarity but flattening as a process is complex the disruption induced within the image object. These vectors
TARON & PARKER 225

r Figure 7. Still from on the resolution frontier.

are extracted as a vector flow range spreadsheet that includes the

UV values for each pixel’s vector, and the maximum and minimum
vector difference between correlate images (max-flow). The vec-
tor values contained in these spreadsheets can be mapped in 3D
model space by extracting the UV directional values of a pixel and
its associated vector magnitude to produce dynamic vector-flow-
fields (Figure 1).

Distortion as Disruption r Figure 8. Project still.

As discussed in the previous section, vector-flow-fields are rep-

PEER REVIEW / SIMUL ATIONS

resentative of the pixel movement between correlate images with from the research? To begin with the fate of SIFTs likely rests
the effect of displacing pixels into a disrupted form-image. By con- in their fidelity to the forms that they represent. Given that we
verting images into soft data it allows information within an image know that closer image spacing translates into better fidelity
to include, exclude and transform its native response to new data (and less distortion), the environment may simply be imaged at
sources without losing fidelity. These bodies of soft data interact increasingly higher frequencies in order to agitate different sets
with each other exerting forces that produce data-rich territories of information. It also means that higher resolution images would
that retain the integrity of each original data point. By extension, eliminate confused associations within the algorithm. By exten-
the algorithmic components of the initial drawing inhere while sion, buildings might be able to anticipate or play with those fre-
additional distortions are layered into the resulting image. As two quencies and resolutions in much the same way that films can be
bodies of soft data confront each other under the constraints of seen in 3D where the gap between frequencies sends different
SIFT algorithms, pixels and their associative vectors autonomously image sets to each eye thereby creating the experience of depth
engage each other transforming in response to each other while through two different channels of information projected onto the
retaining their initial form in layers otherwise dormant or con- same flat screen. In the case of SIFTs, architecture might be able
cealed at the image surface. Disrupted form-images can then be to produce a multi-user image of a building to any single user of
understood as the visualization of two or more images drawn into a the space where the visual experience of a space is represented
single data-rich image containing the information of all input images as a SIFTed image. Buildings themselves could also anticipate the
within a specified resolution Figure 13 through Figure 15. 24 technological differences between various users. For example,
building envelopes could be encoded to express base building
Projecting Representational Disruptions performance information for use by municipalities at a relatively
As a result, how can we begin to make sense of SIFTs and address low frequency while at the same time and at a different level of
the way they are affecting change in architectural objects? resolution, it could communicate information to delivery drones
Shiftiness, glitchiness, and hyperactive edges might describe that engage the building at a much higher frequency.
some of the effects experienced through SIFT-generated form- What we can understand about the technological challenge
images that may begin to function as an aesthetic byproduct of of SIFT distortions is that they can both be seen as a thing to be
multi-user or multi-perspectival environments (Figure 16). But by eliminated for the purpose of instrumentalizing communication
framing these effects as specific distortions rather than anoma- between buildings and the things that “see” them as well as under-
lous glitches, we may also be getting a more precise read on the standing distortions that are fundamentally unique the same way
way in which architectural objects are perceived by SIFT algo- that a fingerprint or iris scan is unique to each individual. A project
rithms. Toward this end, it may be useful to divide an answer to may very well emerge where a building appears to certain users as
this question into categorical parts: the technological, the disci- completely generic while to others it may be highly singular—and
plinary, and the social. that the more difference between this understanding among vari-
ous user groups, the more viable an architectural project is likely
The Technological to be—especially if those user groups can be deliberately inter-
Now that SIFTed distortions have been initially identified, what faced with one another.
kinds of trends or tendencies might we be able to interpolate
 Drawing Distruptions 226
TAD 1 : 2

r Figure 9. Automated SIFT drawing-diagram. r Figure 10. Automated SIFT drawing-diagram.

TARON & PARKER 227

r Figure 11. Stereotomic operations. Disciplinary Challenges

The technological discussion immediately raises the obvious disciplinary questions of form, func-
s Figure 12. Camera orbit diagram. tion, and use. To what end, by what means, and for whom might a multi-perspectival representa-
tion of architectural form be deployed? Here, architectural design is tasked with coupling users
with application in the making of the city very much in the spirit of Corbusier. However, in the
context of architectural representation, the disciplinary implication is to shift away from a solely

PEER REVIEW / SIMUL ATIONS

anthropomorphic discussion of dimensionality in terms of what a drawing is trying to represent.
In its place, multi-perspectival representation must be framed through the distortions that are
produced upon an object when something is flattened into a new mode of representation. Per-
spective drawing, for example, then becomes not a matter of representing three dimensional
space, but rather a mechanism for flattening the edges of spatial extents into a 2-dimensional
plane. This also provides a new mechanism for addressing algorithmic modes of representation
whereby the object, its image and the algorithm that processes their translation into one another
replaces a discourse that would otherwise focus on the dimensions that appear to us when we
confront an architectural drawing or image. This shift from “what appears to us” to “how a thing
is drawn and how it appears to that which sees it” fundamentally impacts the kinds of questions
architecture is tasked with addressing and expands its constituency. The discussion of archi-
tectural visualization expands beyond a human sense of space and materiality but becomes a
medium for communication with algorithms that are tasked with “seeing” the built environment
and making use of the information that is embedded or encoded in it. Therefore, a technical
understanding may be absolutely necessary but at the same time insufficient. The discipline must
be able to anticipate technological patterns of behavior through a familiarity of its use in order
to exert any kind of agency within the environments it reproduces.
By integrating algorithmically induced flatness into the discipline, we might also be able to
establish a new kind of field condition against which architecture can operate. If the twentieth
century defined architecture first in relation to and then subsumed by urbanism, 25 then perhaps
architecture may be able to reclaim a critical authority by acknowledging that algorithmic flat-
ness is a new kind of urbanism. Where any single building is imaged, now a whole city consisting
of pixel neighborhoods now resides. Singular wholes are subverted by locally apparent features
at the scale of these pixel neighborhoods. Simultaneously, similar forms in geographically differ-
ent locations may also produce new communities. The homogeneity of built forms (think exurban
residential developments) may be able to use SIFTs to amplify otherwise undetectable differ-
ences to establish uniqueness. The same could be used to exploit blatant similarities in order to
establish the typology as a kind of constructed platform.
We might also be able to define architectural scale in new terms. Rather than relying on rela-
tive size, we might be able to mine flattening agents for the thresholds at which they attempt
to induce dimensionality or difference out of flattened fields (and vice versa). This is also to
acknowledge that flattened environments and the legibility they afford may impose new limits
and sensibilities toward disruption and disturbance.
Perhaps the most important issue raised for the discipline is the existence of and ease of
access to SIFTs despite architecture’s general unawareness and/or disinterest in them to date.
 Drawing Distruptions 228
TAD 1 : 2

r Figure 13 (Top). Adjacent frames,

serial rotation series.

r Figure 14 (Middle). Alternating

frames, serial rotation series.

r Figure 15 (Bottom). Fifth frames,

serial rotation series.

w Figure 16. Composite SIFT

portrait.
TARON & PARKER 229

while reproducing the existence of the other. Familiar paradoxes

of sustainability and social equity emerge whereby an awareness
of the other may negatively impact their own viability. For exam-
ple, does the breakdown of legibility in the SIFTed images of the
sukkah reveal a deficiency in the algorithm? And would improving
the fidelity in SIFTed images undermine their ability to agitate or
animate the latent information residing in the gap between com-
peting modes of representation? Or if we were to ask the question
in reverse of the conventional orthographic modes of represent-
ing the sukkah, would this somehow frame the singular symbolic
understanding as being over-limited or somehow less significant
that the compounded images of the SIFTs? Do new modes of rep-
resentation demand acknowledging every one of them and does
that come at the expense of allowing any one of them to operate of
If the making of models has become a kind of distraction or design exist? The answer to this question may simply require more time;
placebo—taking too much time, presenting too much difficulty, time less as a matter of “figuring things out” and more as a matter
and expending too much labor—than why not instead make use of of discovering what has been, is, and will be happening.
the models, buildings, and algorithms already in play? It was easy If architecture is to take seriously the expanded constituency
enough for SIFTs to conquer the surface of the planet by aggregat- of multiple algorithmic users (human included), then the notion of
ing satellite, street view and personal image sets; it is only a matter social and political agency would necessarily be extended to those
of time before every surface has been captured within their gaze. things capable of authoring or affecting change in those systems.
However, if architecture is to take advantage of them, it needs to For too long, architecture has insisted on humans as a primary

PEER REVIEW / SIMUL ATIONS

acknowledge that it is already on camera. if not sole actor in social bodies. But their wane has revealed a
kind of vacuum that architecture has yet to sufficiently address.
Socially Orienting the Discipline If geo- and bio-engineering are toolsets for making, thinking and
For the discipline, it may also mean that the production of new governing the world, than architecture is perhaps best equipped
architecture may not necessarily mean producing new physical to tackle the way that images themselves are constructed, con-
structures. Instead, architecture could invest in reanimating the struct, and provide instructions for how the world is made and to
built environment in new ways using technologies that “see” the be made into the future.
world in different ways. This establishes a kind of socio-political
project for architectural distortions. If, for example and as we con-
tinue to prove to ourselves, we are incapable as a society of author- Notes
ing new paradigms or modes of governance due to the relative 1. Luciana Parisi, Contagious Architecture: Computation,
permanence of the built environment as an infrastructural regime, Aesthetics, and Space (Cambridge: MIT Press, 2013), 67.
than our best means of recovering or reauthoring those regimes 2. For more on this subject, see Vernor Vinge’s lecture at D:GP
is through their algorithmic reimaging. Stuxnet was a good (albeit 1 conference session titled “Design and Post-humanism.”
extreme) example of this in action. Here, an algorithm was able to Here Vinge argues that humans might be best at outsourcing
change the way that an operating system “saw” a nuclear reactor cognitive function, which is to say, the development of
technologies that are designed from the very outset to
such that it would allow itself to function differently and by the supplement or exceed the limits of human thought. The talk
design of the algorithm. Does this make SIFTs a kind of Stuxnet for can be seen at the following link: https://www.youtube.com/
architecture? This is a question focused not on using SIFTs to pro- watch?v=-CJSEYYTQas.
duce new objects but rather a call to discover new readings of exist- 3. Alberto Perez-Gomez and Louise Pelletier, Architectural
ing buildings in an attempt to understand the world that they are Representation and the Perspective Hinge (Cambridge: MIT
producing. In Extrastatecraft, Keller Easterling speaks of a world Press, 1997), 3.
where actors say one thing while deliberately doing another; in 4. This was Walter Benjamin’s primary argument in The Work of
many cases, because they are indistinguishable from one another.26 Art in the Age of Mechanical Reproduction. This line of criticism
But what if this is a condition not just of individuals and politicians, deeply extends into Guy Debord’s Society of the Spectacle,
Jean Beaudriallard’s orders of simulacra, Frederic Jameson’s
but of things in general in an increasingly integrated, networked, Cultural Logic of Late Capitalism and Cognitive Mapping (to name
and communicative world? This means that architects should begin a few). W. Benjamin, "The Work of Art in the Age of Mechanical
with the assumption that their own projects are operating beyond Reproduction." Illuminations, edited by H. Arendt. New York:
their intentions and quite possibly against them. In the case of the Schocken, 1969 [1936]), 217–251; G. Debord, Society of
the Spectacle (London: Rebel Press, 1983); J. Beaudrillard,
sukkah, the meaning and legibility of the Hebrew lettering is so
Simulacra and Simulation (Ann Arbor: University of Michigan
fundamentally important to its significance and yet SIFTs operate Press, 1994); F. Jameson, Postmodernism, or, The cultural logic
in complete ambivalence to this intention. What emerges is a dis- of late capitalism (Durham: Duke University Press, 1991).
torted translation of symbolic meaning and legibility of one user 5. Perez-Gomez and Pelletier (see note 2 above), 93, 96.
group into another that can possibly coexist without any aware-
6. K. M. Hays, “Prolegomenon for a Study Linking the Advanced
ness of the other. Each reinforces a group’s belief in themselves Architecture of the Present to That of the 1970s through
 Drawing Distruptions 230

Ideologies of Media, the Experience of Cities in Transition, Joshua M. Taron is an Associate Professor of architecture at the
and the Ongoing Effects of Reification,” Perspecta, 32, (2001): University of Calgary’s Faculty of Environmental Design where
101–107.
he co-directs the Laboratory for Integrative Design (LID). His
7. As was one of the obsessions of Benjamin (see note 3 above) current research focuses on the way in which buildings mediate
and then later Deleuze. Gilles Deleuze, Cinema 1: The the way various infrastructural systems experience one another.
Time Image. Trans. Hugh Tomlinson and Barbara Habberjam
(Minneapolis: University of Minnesota Press, 1986).
Matthew Parker is an architectural designer with DIALOG, and
8. D. Lowe, “Object Recognition from Local Scale-Invariant
Features,” in Proceedings of the Seventh IEEE International a sessional instructor with the University of Calgary’s Faculty
Conference on Computer Vision (Washington DC: IEEE of Environmental Design. His current research focuses on the
Computer Society, 1999), 1150–1157. ability of algorithmic observation to transform, mediate and re-
9. D. Lowe, “Distinctive Image Features from Scale-Invariant animate architectures’ image. Matthew holds a Post Professional
Keypoints,” International Journal of Computer Vision (2004): Master in Environmental Design and a Master of Architecture
91–110. from the University of Calgary, where he received honors recog-
10. Zaera-Polo, "Politics of the Envelope," Volume 17 (November nition and the AIA Gold Medal.
2008): 76-105.
11. Hawthorne, 2007, October 20). “Nokia Theater and plaza
send out mixed messages,” Los Angeles Times (online), October
20, 2007.
12. Hulchich and Spina 2008, Matters of Sensation, Artists Space
Exhibitions, September 25 - November 22, NY.
13. K. Balliet, “Possible Volumes,” Offramp, online (2015).
14. Ibid.
15. Andrew Atwood and Anna Niemark, Couch on Couch (2014);
Andrew Atwood and Anna Niemark, Possible Table (2013); A.
Atwood, “Rendering Environment.” The Expanding Periphery
TAD 1 : 2

and the Migrating Center (Toronto: ACSA Press, 2015),

297–300.
16. Atwood, “Rendering Environment” (see note 14 above), 300.
17. Besler and Sons’ On the Resolution Frontier (2015);
Ian Besler and Erin Besler, Along the Resolution
Frontier (Log, 2016), https://medium.com/@IanBesler/
along-the-frontier-of-resolution-e1eeb1643d33.
18. Besler and Besler (see note 16 above).
19. Ibid.
20. Ibid.
21. Felicity Scott, In Distributed Urbanism: Cities after Google
Earth, edited by G. Hawkins (New York: Routledge, 2010), xii.
22. L. Kurgan, Close Up at a Distance: Mapping, Technology, and
Politics (New York: Zone Books, 2013), 11-12.
23. The letters form the word sukkah in Hebrew. Their integration
into the plan of the sukkah is a necessary design requirement.
The stereotomic projection from three directions enables
the form to assume any one of three orientations while
maintaining adherence to the “building code.”
24. When two image components exhibit zero-magnitude UV
vectors, along with a zero-value max flow they function as
identical; whereas, when an input image and a query image
have less similar SIFT features the values for the UV vectors
and the pixel’s displacement energy increase as the correlation
between pixels and their associated SIFTs decrease. Thus,
the lower overall vector magnitude and energy displacement
between two images the more alike two images are and
vice-versa.
25. This is most clearly articulated by the lineage of critical
discourse originating in Corbusier followed by Koolhaas
and most recently P.V. Aureli. P.V. Aureli, The Possibility of an
Absolute Architecture (Cambridge: MIT Press, 2011), 1-46.
26. Keller Easterling, Extrastatecraft: the power of infrastructural
space (London: Verso Books.