About Horses, Cars, and AI in Architecture
With digitization, design is increasingly becoming automated. This trend can be compared to the advent of the automobile, because the metaphor of the car is well-suited to portray the influence of digitization and design automation on the role of architects and their clients.
The automation of the architectural design process has been researched and taught under the label “cybernetics” for many years at various European schools of architecture. The discourse revolves around the recurring question of how the architect can create the necessary space for decision-making within a structure of algorithms and parameters. Should he consider himself to be a programmer or an artist? An engineer or theorist? Because if an algorithm designs a house, then who is the real designer?
However, this very same question about authorship has an extremely trivial background: the author of a design earns money from this achievement. Just because the software of an architectural firm automatically generates hundreds of buildings — is the architect then to be paid in full hundreds of times (e.g. according to the SIA HOA) for this service? In recent years, various howlers have evolved around the “generative figures” to justify this demand that allowed more eccentric forms. Most of these developments result in sounding out the limits of what is physically possible as a digital mannerism. Emblematic of this are the complex individually manufactured facades of the current architectural canon, which are more of a “generative decoration” than purposeful design.
Of course, this artistic use of the now highly optimized digital tools is completely opportune, but it shoots past the initial question of cybernetics¹. The reason is that so far the wrong stakeholders were addressed by digitization. Architectural theorists who deal with art and creativity, of course, are pushing the creative potential of the new technologies to the forefront, whereas literary scholars find more joy in new word creations than in technological issues². For fear of deterministic design automation, which could jeopardize architects’ jobs, the non-artistic currents of design automation have been neglected at universities for years — but the occasional theoreticians on architecture faculties must prevail and justify themselves in a biotope dominated by designers. But what if we were to design these tools not for designers but for decision-makers? What if we could work with the right stakeholders as a target group? To this end, it’s worth taking a look a hundred years back to the first days of the automobile.
THE REAL ESTATE DEVELOPER AT THE WHEEL
The automobile was without question one of the most important technological developments of the Industrial Revolution. Not because it made the horse superfluous in propelling a carriage, but because it was able to set one speed record after another and was thus able to raise this quality of movement to a previously unimagined level. An analysis of the car from a systems perspective allows for it to be clearly reduced to a “black box”³. The driver makes direction and speed decisions to reach a destination, the vehicle, as a black box, carries out these “commands,” and as a result, the driver reaches his destination. This process is identical for automobiles as well as horse-drawn carriages. Just as before, drivers purchased the strongest horse possible⁴, so too do they choose the most powerful type of engine in a car. One engine may be better than another, but the task of efficiently reaching a destination remains the same. What is hidden inside the black box “vehicle” therefore only plays a minor role for the driver. Whether horse or engine, for the driver, only the quality of the implementation of his decisions is relevant.
If one applies this picture to the meanwhile increased occurrence of design automation in architecture⁵, then alarm bells should be beginning to ring loudly for one professional group. Because the big mistake of the architects is that they have so far viewed themselves as drivers in this system and were supported by most theoreticians in this assessment. However, more and more researchers are currently identifying the systematic relationships in designing more comprehensively and are coming to a completely different conclusion⁶: the classical role of the architect in the car metaphor is more like that of the horse. He does not steer or hold the reins, but instead the client decides. Just as reaching a destination quickly and safely is relevant, so too do builders have clear destinations in mind. Therefore, the reorientation away from the designers to this new target group for the development of design automation has major implications. While designers still measure themselves through publications in glossy magazines and therefore use automation to produce more and more spectacular images, a radical neo-positivism is emerging via Big Data and evidence based design. This means that computer-aided planning is moving more and more away from artistic design and towards objective KPI-oriented optimization. As a last resort, architecture can be defined with scientific precision using economic indicators such as return or life cycle costs.
Just as cars were not designed to replace horses, the primary goal behind design automation is not to replace architects. The vehicle as black box was replaced by a new variant, which was able to fulfill the task of locomotion more efficiently. The counterpart to the metaphor of locomotion in architecture is design. And for this purpose, a system has been in place for years to distinguish “better” designers from “worse” designers with architectural competitions. In this case the developer, with his objectives, is testing various black boxes until he arrives at the desired result. Again, more and more of the participating designers are using algorithms to deliver better results. This starts with simple solar studies and extends all the way to complex optimizations in hospital construction (e.g. the work of Aditazz). The ratio of human to automatic workload is increasingly shifting in favor of the computer, until human performance in design eventually falls behind and then gradually disappears completely. Not only can computers generate and evaluate many thousands of variants with the press of a button, they can also see relationships where human cognitive abilities fail. And just as automobiles led to a new quality of speed, design automation will help to achieve the goals of real estate developers faster and better, and increase architectural quality.
A NEW KEY TECHNOLOGY
Design is clearly in transition as digitization progresses. The large number of stakeholders in this ongoing process makes it difficult to identify key technologies whose implementation does not run the risk of becoming obsolete in a matter of months. Figuratively speaking, are we still relying on fast horses or even slow cars? And if so, which car brand promises sustainable success? Or should we even look for a complementary technology that will prevail beyond the limelight? To answer this question, it’s helpful to stretch the car metaphor one more time:
What is likely the most well-known complementary resource that has gained in importance as a result of the rise of automobiles, is clearly fuel. Initially oil and, in recent years, more lithium for batteries owe their value largely to the social need for mobility. On the other hand, if we look at computer-aided design machines, keep an eye on the price of data centers, fiber optic connections, and graphics cards. Of course, there are also players here in Switzerland who are pushing forward into this market. However, there is a second key technology that has sprung from the need for mobility and is even more closely interwoven with the technology history of transport than the issue of fuel: the topic of navigation. Because no matter whether you drive a Porsche or Trabant, once you’ve taken the wrong route, you can only reach your destination late or in the worst case scenario, not at all. It’s no different in architectural design. At the core, navigation is very closely linked to the destinations set by the driver. When and where does he have to decide on which directions and at what speed? How can he most efficiently reach his destination?
To find the best way, as a driver you need seamless maps that are up to date and that have the necessary accuracy. These three points are essential: Maps for which you don’t have for the whole target area don’t help, because you’re unable to determine the best route. Unfortunately, maps that aren’t up-to-date enough don’t help either, because the information could be wrong by now. And the maps that are too large-scale also don’t help because you can’t discern the necessary information.
If we apply this picture to planning, the road maps that help the driver to decide on a route correspond to the simulations and prognosis models that can then be used by real estate developers to guide the architects. In the early phases of planning, you already know the basics of possible tenants, rental prices, functions, disruptive factors, or added value and can more precisely specify the planning scope. For example, from the outset you’re aware that you want to stick to the norms of the SIA or the building code and at the same time you want to achieve certain qualities. In addition, you also have a better idea of the financial considerations behind the projects. All in all, you cover all sorts of topic areas more or less without getting too in-depth. That is to say, you want simulations or calculation models relevant in all dimensions that provide information seamlessly, that are current, and that are equally detailed — and that don’t just set individual priorities. Because as a real estate developer, you have to be able to extensively navigate and the more data and simulation results that can be used as a basis for decisions, the better the best possible and most efficient solution or the better the return later on.
Whether such data is now available as a street atlas, in which you have to look for the best route yourself, or whether this data, similar to a navigation device, already suggests different routes, is theoretically only a matter of usability. But only theoretically, because in reality, the data and the amount of possible decisions is so extensive and complex that even with study and work experience, countless hours are needed for the development and evaluation of solutions.
At the moment, the situation is comparable to a scenario in which you’d have to first combine the data from many road atlases in order to be able to plan an optimal route. This pain point not only costs money and is time-consuming, but until recently wasn’t even technically feasible. Until last year anyway, because since then, a new ETH spin-off⁷ is taking up the cause of the solution to this problem. Archilyse is the first company to offer a comprehensive and easy-to-use simulation service for various stages and areas of application of the real estate value chain. Whether residential or workplace qualities, retail or security issues — so far subjective architectural qualities can be quantified with this service. And little by little, general aspects such as standards checks or automatic zoning arrangements are being integrated, so that as a developer you only need a single, easy-to-use source of information. And one that also offers functions that are most likely comparable to those of a navigation device. Not only is all the relevant information collected in one place, but additional analyses help with the customized analysis of this information.
The issue of mobility was revolutionized by the replacement of cabinets full of maps by a small GPS device. Everyone was thus able to choose the best routes and directions for themselves without much effort. In the meantime, it doesn’t matter if you’re traveling on foot, by bicycle, by car, by boat, or even by horse-drawn horse-carriage — navigation devices are omnipresent and can help regardless of the type of drive. The same will be true for real estate development: no matter which architect or design vendor you work with, the question of whether the proposed solutions are going in the right direction will have computer-aided verification.
In this new technological ecosystem, no stone will be left unturned in the construction industry. Since all sorts of architectural qualities suddenly become easily measurable, it is possible to compare designs to an extent that was hitherto not financially feasible. From questions about target groups and associated rental price estimates, to deterministic methods for optimizing cleaning costs, all the way to functional influences of the architecture on the users⁸, the effects of the architectural design can be transparently depicted. The measurability of architectural quality will inevitably favor the further expansion of design machines and strongly influence the profession of the architect. They will probably continue to move away from an artistic understanding of their work and begin to use more algorithms like the big planning companies⁹.
tl;dr
Designing will be inexorably automated and the impact on architects and their clients will be significant. This trend began years ago, and various companies have been enjoying economic success with their algorithm-based designs for quite some time. In addition, this trend is favored by the development of various complementary technology, which brings about additional stabilizing side effects. Looking at the entire system in a context of technological history, we are experiencing live the transition of the oldest and, at the same time, the last technical discipline from an art to engineering science. As a result, the entire field of the real estate industry and all its stakeholders is in a state of upheaval. Of course, the exact development cannot be predicted, but you will not be able to sit this trend out.
¹ Initially Richard Buckminster Fuller had more ecologic goals, https://en.wikipedia.org/wiki/Operating_Manual_for_Spaceship_Earth
² e.g. next to ETH Zurich the wording “pre-specific design” is now also propagated at TU Wien, https://www.degruyter.com/view/serial/247566
³ a “black box” is a theoretic construct that helps to focus on the outer effects of a component rather than on its inside, https://en.wikipedia.org/wiki/Black_box#Other_theories
⁴ competitions helped to identify the best horses, almost as in architecture.
⁵ e.g. the work of https://www.archistarr.com/, http://www.aditazz.com/, or https://www.woodsbagot.com/enterprise/superspace
⁶ an important role plays the work of Albena Yaneva and Bruno Latour, https://www.researchgate.net/publication/287718080_Mapping_controversies_in_architecture
⁷ e.g. the startup founded by the author himself, https://www.archilyse.com/
⁸ Architecture is much more influential as many want to admit. Bad floor plan layouts lead for example to domestic violence, addiction and drug abuse or stress and sick building syndrome, while good layouts help to improve the cognitive capacities of children, reduce recovery times in hospitals or reduce stress and even migraine.
⁹ e.g. Woods Bagot with Superspace
