You are on page 1of 3

STUDYING

ARCHI TECTURE
DA NIEL KOEHL ER
Why should one study architecture if anyone as difficult to understand, which, above all, into Alberti’s ultimate architect and Harari’s

AFTER AI
can design anything anywhere instantly at any halved the cost of planning and making. Homo Deus. By that, architecture and a re-
time? Today—much faster than predicted— search project that spanned over five hundred
Deeply linked to the economics of a building,
anyone with access to a cellphone can draw years comes to an end. When natural language
the composition of a building, therefore,
anything of their desire at a resolution indiffer- can process three-dimensionally, there won’t
served as a kind of benchmark to value the
ent to reality. The implications for architecture be any need for an extra architectural lan-
design. Aesthetics equated with the utility of a
are obviously ground-shaking. This essay tries guage; for any plans, sections, or similar medi-
building. Alberti identified three aspects that
to grasp some of it—its implications for teach- ations that translate intention into form.
shaped research in architecture for generations
ing and learning in a studio environment, and
to come: first, a beautiful design uses first a Natural language models translate not only the
for preparing students for the significant chal-
limited number of parts which keeps the need- desires of clients but anything that is easier to
lenges that face architectural practice today.
ed equipment and workers on a construction communicate through words—like communal
For hundreds of years, and in Western cultures site manageable; second, those parts should decisions, legislative codes, and, in general,
since the publication of Alberti’s treatise De have a clear form that is easy to produce and any kind of knowledge. The superiority of AI
re aedificatoria in the mid-fifteenth century,1 transport; and finally, those parts should be models is already beginning to shatter. For
architectural knowledge was defined as a composed into a correlated arrangement using example, the text-based model GPT4, which
notational science. Architects invented ways symmetries and other connotations to make it is readily available through a number of free
to notate, draw, compose, or codify—not to easier to comprehend the building. Since then, online writing tools, utilizes over one hundred
build (that was the role of the craftsperson) research in architecture invented notation after trillion different parameters. By comparison,
nor to house (that was the realm of the client). notation to design more with ever less. The the average adult uses only up to thirty thou-
The role of the architect was only to mediate Baroque celebrated architects who could draw sand different words.5 That means a neural
between both.2 Plans, sections, and other complex forms by calculating the transfor- network uses three billion unique assets per
architectural representations quickly notated mation of simple shapes. In very similar ways, one word you use. By looking at vast amounts
comprehensible forms on one hand, and eco- five hundred years later, the Digital celebrated of data, digital machines learn to repeat cor-
nomic—and, later, cultural—investments on architects who could shape any free form from relations between different features, which
the other: the matter and methods to build.3 even less than parts: particles. then can be interpolated and automated. Thus,
Made by humans for humans, buildings had AI can automate any task that is particular
Alberti’s treatise did not contain any draw-
to be easy to understand. If a client’s inten- enough to be identified as such and that is
ings, images, or other notations. For Alberti,
tions were easy to understand, more complex frequently executed.
ultimately, an architect, like a god, would form
shapes could be built with less. A symmetrical
matter with words. And here we are today: Despite Richard Sennet’s famous expression
drawing was considered beautiful because
Networks like OpenAI’s DALL·E 2 draw that “for great craftship, it needs first a passion
the mirrored half of an original was only half
realistic images from text descriptions only.4 A and then 10,000 hours more of it,”6 any profes-
digital image is built from pixels, two-dimen- sion that relies on classic crafts will most prob-
sionally arrayed square-y points. Technically, ably be automated. Subsequently, architects
FIG. 1 An Urban block in
Vienna designed by Vienna it takes very little additional information to will have to find new ways to contribute as a
and its data. Co-authored turn a pixel into a voxel and an image into profession by valuing a different set of skills—
with Midjourney AI, Daniel
Koehler, and an uncountable architecture. In limitlessly replicable ways, skills that one will have to learn together with
number of situations tagged as
Viennese in photos.
the network can be applied to anything at any AI. The implications fall into three categories:
place instantly, turning anyone with the desire predictions of the professional contributions

70 PL ATFORM 2022-2023 TE ACHING FOR NE X T 71


that an AI-infused culture will value, the skills have precluded education plans, and are also of knowledge transfer, and with AI that can
necessary for those contributions, and the available for marginal fees as a vastly greater augment missing skills, a class can be diverse,
ways AI will transform learning itself and its number of participants share costs. assembled from students with different skill
environments. levels based on their intent to contribute.
Available anywhere with an internet connec-
In similar ways, grading can be much fairer,
Until recently, most learning environments tion, a student can choose the best available
at once tailored to the individual student’s
within schools of architecture still followed content from the best teacher from a global
learned skills while also subtly leveraged
principles invented in Victorian times as a collection. In response, Donald Clark, an
against a globe-spanning database of compara-
response to the beginning of industrializa- expert on teaching methodologies, proposed
ble assignments.10
tion. Not by coincidence, the typical amount the concept of “flipped lessons,” or courses
of time involved in architectural education where students watch lectures for homework Design environments require less of a chat bot
equates to Sennet’s ten thousand hours of and use classroom time to apply their content than a drawing bot. With the encyclopedic
apprenticeship by learning eight hours per day, to practice.8 In this model, a teacher’s role knowledge of a building—including legislative
five days per week, for five years. In many cases, flips from instructor to mentor or tutor in an codes, material features, structural behavior,
a typical classroom still contains as many stu- interactive practice, serving as a curator to turn building protocols, and so on—such a drawing
dents as one teacher can instruct, and we still abundance into comprehensible lessons. In this bot would augment any professional knowl-
teach as though students are training for only scenario, teachers curate content that provides edge onto the lines, shapes, or blocks we draw.
one role that they will perform for their entire a platform for studying, and courses become Freed from the need to notate in conventional
lifetime, for which they must memorize a body stages for a particular discussion. As a result, media and equipped with an XR interface, you
of knowledge that is a scarce resource.7 students become constructive participants who can immediately draw or better choreograph
shape innovation through projects analogous to architecture in space. Freed from the need to
Before digitalization, schools were places of
the format of inquiry-based learning. instruct by simulating a given knowledge, de-
exclusive access to knowledge. Schools were
sign studios can finally become places to design
designed as a kind of Wunderkammer of a This is precisely where AI systems now step in.
knowledge. Students can join a school not to
profession, with the goal of preparing students Already, in daily life, Google responds to que-
learn but to study—in line with the classical
for any situation they might encounter in their ries with AI-generated summaries. The same
notion of a university in which students do not
professional lives. Therefore, schools had to AI linked to the interface of a chatbot turns
learn today’s facts but study future challenges
provide a condensed simulation of reality on into a personalized teaching assistant with an
to build their careers ahead. When knowledge
which students could be trained, tested, and encyclopedic knowledge of the curriculum one
is freely accessible anywhere at any time, the
certified. The standard curriculum in architec- needs to master. Trained with the data of one
value of knowledge will increasingly lie in its
ture still resonates with such motivations. In student’s life, the bot will understand—possi-
authenticity and original contribution.
design studios, students repeatedly arrange, in bly better than the student themself—the most
increasing complexity, a set of spatial queries suitable approaches to their learning, such as So, why should you still gather in a studio to
onto a virtual site reviewed by the studio’s the times of day when they are most receptive design together? Because in the same way that
professors. The standard studio outcome con- and when they are best left to relax for their one network can run with multiple instances,
tributes only to a virtual, simulated context emotional well-being. Mentors can use AI to we can design now with many hands. Never
that is valued only as proof of mastering a set translate a course’s content into personalized before could designers jointly draw multiple
of skills of a prospective profession. Such a programs based on their own unique experi- responsive shapes simultaneously. Unlike
simulation of designing a building never re- ences and preferences. The bots can adapt to the modernist who needed to craft a single
flected the reality of architectural practice. In each student’s level of knowledge and speed thing ten thousand times, and then again and
reality, a team of architects designs a building of learning, sharpen their skills, and improve again, designers in an AI-infused culture will
in a far more extended period than a studio weak spots.9 Freed from a one-size-fits-all pace design a single thing in ten thousand different
provides. However, the studio had to compress
for students the wonders of their prospective
buildings to come. FIG. 2 Thirty-six sectional

Digital technologies have enriched learning Digital technologies have studies on urban blocks made
of mass timber, infused with
vertical foresting. Labor time
enviornments enormously, and AI systems and expenses: 0.35 GPU hours
might be the last melt below the tip of an
iceberg. Classes based on a lecture format are a
enriched learning environments to $1.75. Image by Daniel
Koehler.

FIG. 3 The city as a forest.


great example. Digital lectures, once recorded,
can be replayed infinite times to infinite par- enormously, and AI systems What would it take to live with
a forest? With the pace of cur-
rent AI models shouldn’t we
ticipants and provide their content asynchro- expect an exponential increase
nously at any location. It is not necessary and
no longer efficient to watch lectures gathered at
might be the last melt below in research on new materials,
construction methods and
alternative living? Image by

the tip of an iceberg.


Daniel Koehler.
one place at a dedicated time. Today’s Massive
Open Online Courses (MOOCs) make
knowledge accessible to remote locations or
individual circumstances that might otherwise

72 PL ATFORM 2022-2023 TE ACHING FOR NE X T 73


NON-REPEATABLE
FIG. 4 View into a residential
area of a community-owned
building with a water harvest-
ing system that mitigates tem-
peratures while offsetting rent
for farmers. “Tower 2031,” by
Erin O’Brien (MArch, Year),
Paperclips Studio, spring 2021
(instructor: Daniel Koehler).

ways. AI technologies offer us a new world the climate crisis. Architecture schools will 1. Leon Battista Alberti, De re aedificatoria (Mainz: Jacobus
Cammer, 1451).
of knowledge that is plural, non-linear, and be freed from teaching drawing and other
non-repeatable by nature. The skills necessary forms of notation, and studios can become 2. Mario Carpo, The Alphabet and the Algorithm
(Cambridge, MA: MIT Press, 2011)
to craft unique, authentic knowledge through places to envision and design proposals for
3. Antoine Picon, The Materiality of Architecture trans.
the fusion of human and artificial intelligence the future. And we cannot have enough of Abigail Grater (Minneapolis: University of Minnesota
coined a guideline for K-12 education six such proposals; the challenges to the built Press, 2020).

years ago as the “four Cs”: Communication, environment are daunting. According to 4. “DALL·E 2,” OpenAI, April 6, 2022, openai.com/dall-e-2.
Collaboration, Critical Thinking, and calculations by UN-Habitat,14 to address the 5. Alberto Romero, “GPT-4 Will Have 100 Trillion
Creativity.11 To immerse ourselves into the housing crisis we need to build the equivalent Parameters—500x the Size of GPT-3,” Towards Data
Science, September 11, 2021, towardsdatascience.com/
new design media, architects should similarly of the existing US housing stock each year, gpt-4-will-have-100-trillion-parameters-500x-the-size-of-
emphasize emotional intelligence, soft skills, across the globe. At the same time, according gpt-3-582b98d82253.

and cognitive flexibility. to the Intergovernmental Panel on Climate 6. Richard Sennett, The Craftsman (New Haven: Yale
University Press, 2008).
Change (IPCC) report,15 we need to find ways
Wilson and Daugherty, scholars on the future
to reduce the carbon emissions of what should 7. Calum Chace, “The Impact of Artificial Intelligence
of learning,12 elaborated on the fusion between on Education,” Forbes, October 29, 2020, forbes.com/
be a doubled building stock to 10 percent of sites/calumchace/2020/10/29/the-impact-of-artificial-
humans and AIs: within augmented working intelligence-on-education.
today’s emissions. Combined with the twen-
environments, communication translates into
ty-fold increase of building production and the 8. Donald Clark, Artificial Intelligence for Learning: How to
the skill of an intelligent inquiry—knowing Use AI to Support Employee Development (London: Kogan
cap to 10 percent of the current emissions, we Page Limited, 2020).
how best to ask questions to an AI agent.
need to teach future architects how to design
Collaboration becomes the capability of 9. David Karandish, “7 Benefits of AI in Education,”
and operate buildings some two hundred times THE Journal, June 23, 2021, thejournal.com/
reciprocal learning, like teaching AI agents articles/2021/06/23/7-benefits-of-ai-in-education.aspx.
more efficiently than we do today.
new skills and in turn learning AI-enhanced 10. Kyoungwon Seo, Joice Tang, Ido Roll, Sidney Fels, and
processes. Creativity depends first on the skill Soon, networks like DALL·E 2 will fully auto- Dongwook Yoon, “The Impact of Artificial Intelligence
on Learner–Instructor Interaction in Online Learning,”
to design processes to increase available time mate any aspects of notating a building, from International Journal of Educational Technology in Higher
for distinctively human tasks and learning. In a sketch to plans to the codes for robots to fab- Education 18 (2021), doi.org/10.1186/s41239-021-00292-9.

augmented environments, critical thinking is ricate and build. Finally! When AIs can draw 11. “The 4 C’s to 21st Century Skills,” You for Youth, US
Department of Education, 2016.
exercised by integrating judgment, choosing anything, architects will have the resources to

NON-LINEAR
a cause of action amid machine uncertainty, design what we need more than anything. And 12. H. James Wilson and Paul R. Daugherty, Human +
Machine: Reimagining Work in the Age of AI. (Boston:
and endless rethinking—such as developing more than anything, we need to design new Harvard Business Review Press, 2018).
mental models of AI agents that improve col- architectures. 13. Hermann Hertzberger,. “Friendly Architecture – In the
laboration outcomes or correct bias. Footsteps of Structuralism: An Interview with Herman
Hertzberger,” interview by Daniel Koehler, Prospectives
1(2020).
When asked what resources he needed for the
1990 founding of the Berlage Institute, one 14. “UN-Habitat: Annual Report 2020,” UN-Habitat, United
Nations, 2021, unhabitat.org/annualreport.
of the leading architecture schools of its time,
15. “IPCC Sixth Assessment Report: Climate Change 2022:
Herman Hertzberger replied, surprised: “For Mitigation of Climate Change,” Intergovernmental Panel
a school, you don’t need anything. A school is on Climate Change, United Nations, 2022, ipcc.ch/
report/ar6/wg3.
a movement. It is a place of study, communica-
tion, interaction, support, and presence.”13 For
most of history, schools have not been places
to instruct or simulate an environment for
problem-solving. They always have been places
of confidence in the questions of the future.
FIG. 5 “Machine Inclusive When the work that the architect was known
Learning Environment,” by
Victor Trautmann (MArch,
for has been automated, they will be freed
Year), Paperclips Studio, from their craft and can focus on what they
spring 2022 (instructors:
Casey Rehm and Daniel should be concerned about: the housing crisis,
Koehler).

74 PL ATFORM 2022-2023 TE ACHING FOR NE X T 75

You might also like