Automation of Design Processes

[originally published in May 2018 on Medium]

Automation is a trend omnipresent in all the industries. The capitalistic demand for enhancing efficiency (of literally everything) leads us to The Growth — ideally exponential. Executing the projected goals and our plans could be achieved by two means — human intelligence or artificial intelligence. So-called developed countries are undergoing a drastic decrease in productivity; the correlation between demographic change, both negative population growth rate (1) in leading economies like Germany (-0,1%) or Japan (-0.2%) combined with an ageing society, and the low unemployment rate, seems to be obvious reasons. Some argue that today mass human migration, which in 2014 hit 59.7 M(2), would be a remedy. The actual political situation seems to deny it. Here I am going to present another more likely scenario — automation replacing manpower.

The invention of the first water-clock by Ctesibius in Ancient Greece could be considered as a very early application of automation. Throughout centuries we witnessed some minor and scattered advancement in that field — thermostat, centrifugal governor among many others. We had observed automation slowly becoming more common when factories got electrified at the beginning of the XX century. Jeremy Rifkin (3), an American economist, claims that the term automation has not been widely used before 1947. This world was introduced with opening of an automation department by Henry Ford in one of his factories. Contemporary implementations are based on an extensive research in the fields of computer science, mathematics or biology of the last twenty years. All the industries have been and will be, undergoing drastic changes. Architecture and design will not stay aside. The pace of the changes is accelerating faster than ever before. Technological advancements age and become obsolete from day to another. Whether we like it or not, the imperative of the economic growth is simply stronger and everything is going to change.

At the beginning of February 2018, PwC published a report identifying three waves of automation: the first one, that has started already and will last until early 2020 could potentially impact up to 3% of jobs. The second wave, projected to finish in late 2020, could potentially impact 20% of jobs. Finally, the last wave, nicknamed Autonomy, will last until the mid-2030s and could affect 30% further jobs. The second wave, called by authors Augmentation, would focus on repetitive task and exchange of information. What might be striking is the significantly increasing importance of the level of education, as simple jobs show a lower level of the risk of automation. The authors claim that the Augmentation wave will hit white collars. Would it hit architects too?

Fabrication, precasting, off-site manufacturing have changed the face of construction: more processes are automatized and digitized, tools are being modernized and updated. Recent achievements in digital and robotic fabrication could be perceived as “just” a continuation of what has started in the beginning of the XX century — such as Ford assembly line. Qualified and unqualified workers have been replaced by machines and robots. Today, the “physical” robots are complemented with “software robots” — Artificial Intelligence, that aims to replace our brains, not muscles.

Would machine steal my job? That is a question that many of us have asked… And it is not new. Also, there is no simple answer. Creative and design work is definitely going to change. Since the computers and software were tested in architecture in the sixties ( Ivan Sutherland’s Sketchpad) and seventies (Architecture Machine Group — Negroponte), a huge number of researchers and scientist began devising the way to automation. A book Automated Architect by Nigel Cross drew some scenarios how some tasks could be digitized. I strongly believe that the job of an architect (or any creative) is unlikely to disappear. Though the job description, responsibilities, abilities, and role would definitely change. There are innumerable tasks that a highly qualified and educated designer will, blissfully, hand on to the machine and algorithms. As I mentioned before, the second wave of automation is closer than we acknowledge. The main focus would be repetitive tasks that could be replaced by narrow Artificial Intelligence.

A job description of an architect includes the following tasks (in chronological order):

1.defining clients needs, preparing a concept, scheme design and a project budget

2. planning and execution drawings production, including detail drawings, perspectives views; coordination with engineers, consultants, specialist, contractors and clients

3. project review, application for the building permit and the construction supervision

Analyzing a few examples of each phase and defining whether and how could they be automatized could give us the volume of upcoming changes.

The first example might be planning a parking. A task that, along with a shop, an office or a hospital project, architect or landscape architect has to deliver. The tedious work of finding an optimal spots distribution, fluid circulation and good balance between small, large and handicapped spots is considered less challenging and less interesting than façade or layout design. Parking planning is more like a jigsaw puzzle, a game with a strict building code on top. When I presented this idea to my fellow architect colleagues, they were not simply interested, many asked when it is going to be available. Since last year I and my colleagues have researched and developed a tool that aims to solve this task. Today we imagine that only by inputting the outline and project location (to determine the local building code) we could generate a series of options for a parking. Potentially the time needed for the planning would take less than a minute. The preliminary studies revealed that an algorithm employing optimization techniques could fit up to 30% more spots into the same outline.

During the eCAADe (Education and research in Computer Aided Architectural Design in Europe) conference in 2015 in Vienna Michael Wimmer from Institute for Computer Graphics and Algorithms showed a project that tackles the problem of furniture distribution in a bathroom. An algorithm learns from examples of where and how furniture should be arranged in a hotel bathroom and later distribute them accordingly in any bathroom given. The algorithm exhibited some many imperfections, but the idea is clear. A machine learning system can learn a typical distribution or arrangement and apply the rules to solve a similar case. Among the projects and research in the design field, Machine Learning algorithms seem to be actually the most popular and powerful medium to continue with the radical shift in the design field.

Once a project enters Design Development phase, adjustment to actual building code and local regulations must be made: fire code, structural code, the height of steps, angle of vehicular ramps, doors widths and hundreds of elements need to be designed and reviewed. Even a minor adjustment could lead to major changes in evacuation scheme. We could look on a BIM model of a building as pure information on characteristics and relationships between elements that have to follow a certain logic, which is defined in building codes. We are likely to find a hint on how to proceed with data. During my last meeting with Beuth, a daughter company of DIN, the world’s leading institute for standardization, I was told that there are about ten thousand building norms, only for Germany, and they tend to change often. Is it possible to keep up? Today larger architectural offices tend to hire people, architects or engineers, whose job is to verify whether a project complies with all the codes. The question raised was whether it is possible to create a software that analyzes natural language, the sort of natural language used to write a law, interpret it, and based on that verify if a project follows building code. That is an ambitious idea, more ambitious would be if an algorithm could propose a solution, the optimal solution to solve the problem.

Zumthor, Ricci, Pezo von Ellrichshausen and others might disagree with these concepts and path — parking could be and should be, an important part of a design scheme. Designing a bathroom is as important as the entrance hall of the museum, one might say. This reminds me of a meeting with an architect who works for David Chipperfield (DAC). We were talking about how automation and optimization of design can accelerate their process and allow them to make an informed design decision and reduce the operation cost of a building. Strangely the conversation ends up in the conclusion that DAC is “a design-driven office” and they are not interested in such an approach. One could not disagree. The pluralism, also in architectural design, helps to maintain a balanced development and to discover new opportunities. The automation of the processes has much to do with mass-production or mass-customization. Sadly architecture becomes a product. Undoubtedly, there will be always a space for the great masters and minds that could follow their path creating extraordinary and unique jewels, bringing us joy.

I have just presented three possible applications of how architect job could be automatized. I believe every architect, and every engineer could imagine a plethora of other application and case studies. Still, they are not replacing the architect; they are enhancing the productivity and equip with time for a concept, design strategy or for building physical models. Design of some typologies could be highly or even completely automated — mall design, distribution centre, business hotels or corporate offices to kick-off with the list. For typologies, where efficiency is the most important driver, automation could (and most likely) would team up with optimization and supply only the “fittest” options. Optimization of costs would the most important driver of this architecture (?).

Architecture sometimes is defined as a compromise between art and engineering; it might be that in the future architects would design only exceptional buildings, a true masterpiece, other buildings will be generated by algorithms and build by machines.

(1) The World Factbook, https://www.cia.gov/library/publications/the-world-factbook/fields/2002.html

(2) Worldwide displacement hits all-time high as war and persecution increase http://www.unhcr.org/558193896.html

(3)Rifkin, Jeremy (1995). The End of Work: The Decline of the Global Labor Force and the Dawn of the Post-Market Era.

(4) Waves of Automation. https://www.pwc.co.uk/press-room/press-releases/waves-of-automation.html

(5) io9.com mentions narrow AI. 
http://io9.com/how-much-longer-before-our-first-ai-catastrophe-464043243

Powered by WordPress.com.