Beyond the Blank Page: Generative Design and the Future of the AEC Industry

Figure 1: Urban option

For more than 40 years, computational practices in the architecture, engineering, and construction (AEC) industries have been taking a detour from their full promise. As if we lacked more than 5,000 years of historical buildings and more than 2,000 years of writing about buildings, and as if commercial practice and academic study had not produced oceans of knowledge concerning the best ways to build, popular software environments for building professionals still present every new project as a blank screen, as if none of that rich history had ever happened.

The earliest academic experiments applying computation to architecture and engineering problems date to the early 1960s, beginning a period of exciting development and speculative advancement that lasted more than a decade.

In contrast to that auspicious beginning, the first commercial software offerings for architects and engineers focused on the most mechanical aspect of practice—the production and management of construction documents. Arguably, the adoption of CAD increased the production speed of project deliverables and the succeeding advent of BIM helped increase productivity further while enhancing the coordination quality of construction documents. However, the adoption of computational technologies to support only one narrow aspect of building practice ignored an abundant and provocative body of work in favor of an increasingly elaborate translation of physical drafting to the digital medium.

There was always an alternative to this detour, the primary exploration found in the earliest papers and experiments with architectural computing: the various practices we now collectively label as "generative design." To aid in the recovery, examination, discussion, extension, and application of these practices, in 2018 I publicly defined generative design as "the automated algorithmic combination of goals and constraints to reveal solutions," attempting to avoid the shortcomings of more anemic definitions designating a single technical approach such as topological optimization or option creation as the totality of a much broader endeavor. The methodologies of generative design have the potential to fundamentally advance the AEC industry by capturing and contextually applying vital building expertise.

Figure 2: Tower options

While the capability of generative design to deliver a practical infinity of solution variations has been its most visually compelling characteristic, the true importance of these practices is their scalable, repeatable, and distributable embodiment of building knowledge.

Popular software environments supporting AEC primarily help record and communicate building decisions arrived at by some other method. Generative design practices seek to incorporate the means of decision making directly into the design environment. Whether it is by including widely understood inhabitant needs such as minimal ceiling heights, horizontal and vertical circulation, structural integrity, emergency exiting, or environmental comfort, there remains no reason why the AEC industry must continually answer those needs limited by the consulting speed of one hour per hour, the figurative speed of light for the consulting business model.

Whether captured through the straightforward method of coding into software their best practices or revealing and applying those practices through the automated mining of existing data through machine learning, we need no longer recapitulate the slow acquisition and employment of building knowledge solely through the direct involvement of the individual building professional; our machines are capable of generating and helping us explore an infinity of solutions constrained by requirements and in service to our goals. To ignore the possibility of producing better buildings through scalable computation employing captured professional expertise in a world growing toward 10 billion inhabitants is not merely an opportunity discarded, but will eventually represent a collective failure to address one of society's most vital needs.

In their book The Future of the Professions, Richard and Daniel Susskind explore the role of the professions, which can be loosely defined as those people afforded certain considerations and privileges in return for conserving, extending, and applying bodies of human knowledge important to the maintenance of a healthy society. By their measure and by the measure of anyone honestly observing the current delivery of medical knowledge, accounting practices, legal skills, or building proficiency, it is obvious that absent public subsidy, professional expertise is delivered parsimoniously to a high and thin economic stratum, and most often, not at all.

Defensible statistics are difficult to discover, but since the late 1960s it has been popular lore that the quantity of buildings enjoying meaningful involvement of an architect is approximately 5 percent of the inhabited environment, one of many factors leading David Susskind to remark, "Often in judging the merits of a new system, the appropriate comparison is not with what the best human expert would achieve, but with nothing at all." And so, we see popular use of web services such as WebMD, Rocket Lawyer, and Intuit—all of them filling gaps in societal needs for medical, legal, and accounting advice that would otherwise remain unaddressed.

Further, the lack of systems in the AEC industry to meaningfully capture and deliver expertise slows development of the professionals themselves. The widely maintained belief that building design apprentices require approximately two decades of practice before being deemed "mature" in their profession is not merely a reflection of the demanding complexity of modern buildings, but an indictment of a failure to leverage available information systems to contextually deliver common building practices.

Figure 3: Tower option

Do we still need to devote valuable apprenticeship time to repeatedly depicting regulation commercial restrooms for every single project starting from the equivalent of a blank page? Do we still need to send novice building professionals scurrying to code books to discover how to assemble those restrooms? Do we still need to directly draw every fire stair, every exiting scheme, every multi-unit residential arrangement, every building service core, every office plan, as if we had never created anything similar before? Do we truly need to act as if there were not an elaborate set of regulatory and physical constraints governing so many of these decisions? Must we still practice as if it were impossible to capture and deliver such expertise and such constraints implicitly in our software environments?

At Hypar, we decided the answer to these questions was not only "no", but has been effectively "no" for decades, a period during which the AEC industry has been distracted from the real benefits of computationally enhanced building design by the prospect of producing more drawings faster, as though the design industry were merely a drawing factory rather than an important repository and delivery mechanism of critical human knowledge. While the reality of contractually prescribed project deliverables is not likely to change for generations to come, the methods by which professionals arrive at their decisions and maintain their institutional knowledge must undergo a radical transformation to effectively scale beyond the limitations of the consulting delivery model.

Absent debilitation, professionals are most knowledgeable on the eve of their retirement, after which their knowledge and understanding are lost to subsequent projects except as filtered through the records they have produced and the protégés they have cultivated. The expertise of those successor professionals is not available when they take ill, take vacation, or undertake other obligations. Those professionals cannot consult to more than one project at a time, even if the rapid redirection of attention between tasks has provided an illusion of such capability.

Figure 4: Office options

Figure 5: Office option

Further, the ability of professionals to consistently apply their own hard-won expertise has been revealed to be highly and sometimes dangerously limited. In his book The Checklist Manifesto, surgeon Atul Gawande cites numerous anecdotal and statistical demonstrations of the failure of surgical teams to properly deliver on critical procedures and outcomes and concludes, "...the volume and complexity of what we know has exceeded our individual ability to deliver its benefits correctly, safely, or reliably." Can building professionals honestly say that they surpass specialized surgical teams in consistently delivering their knowledge every time?

None of these limitations apply to software, which can assume a more central role in augmenting the delivery of knowledge everywhere and into every context where it is required. Internet connectivity coupled with scalable computation can deliver building expertise tirelessly and consistently nearly anywhere in the world where buildings are being built with or without the direct participation of learned professionals.

As we surmount the limits of our biology by creating and applying tools to our problems, the abilities of some of those tools are regarded as so necessary in certain focused aspects of professional building practice that avoiding their use has become a signal of backwardness and even of ineligibility for participation in many projects. How long until the failure to take advantage of automated expertise becomes a barrier to building at all? Is there a structural team working on a project of any significance that does not employ at least one, or more often two, software analysis platforms to confirm and extend their professional judgment? Is there a process designer attacking a significant problem who does not embrace the insight of an automated study? How long until the inverse of such analytical processes as represented by generative design becomes not merely a welcome addition extending traditional practice, but a requirement from clients desiring assurance that multiple alternatives have been examined and accepted or discarded according to rigorous and defensible criteria?

All projects start with a catalog of client and professional goals and a variety of constraints, whether of function, budget, regulation, site, or public responsibilities. If a building project can be understood as the intersection of these goals and constraints, and more practically as the intersection of multiple interdependent systems of accommodation, we need not assume generative design necessarily leads to the production of fully realized buildings through an entirely automated process.

Figure 6: Formwork option

Figure 7: Duct option

The production of discrete building systems or spatial solutions can be combined with professional direction and judgment to result in improved solutions enjoying high confidence of all project stakeholders.

Generative design processes allow for illustrated explanations of explored alternatives and by what metrics they were measured, even if that measurement is the articulation of a design strategy more fully embodied in one alternative than another. Today, it is possible to inform and extend professional judgment of localized project choices using scalable techniques of exploration and optimization. Many design and engineering firms have internally produced such processes, but almost as many confide that they find the distribution, adoption, and maintenance of their solutions challenging.

The answers to these issues are nearly as varied as the firms in question—in some cases leading to the creation of entire software companies to both realize the benefits of captured expertise internally and financial gains through the technological and compensated distribution of their expertise.

At Hypar, we believe such investments in capturing professional expertise for scalable delivery and the realization of associated financial rewards are the future of the AEC industry, but that the barriers to that future remain too high. Currently, few building professional firms have the time, money, and expertise to create software companies to preserve and extend their knowledge, and so at Hypar we've sought to lower those barriers by providing commonly combined technologies on an open platform employing an open source data model.

We believe the industry must adopt the means to build on one another's experience by providing and using technologies embodying critical building expertise, and we seek to facilitate its compensated combination and exchange. Such practices are routine in commercial software production; the use of publicly available or licensed software libraries is now so common as to render building any software from a blank page not only an expensive indulgence, but tantamount to professional malpractice. The software industry constantly recombines its collective expertise through the media of various toolkits and libraries, so that no software project, as replete in goals and constraints as any building project, need ever start from scratch.

We build software on the strengths of literal generations of others' achievements. We build buildings the same way, but for decades the AEC industry has indulged a long digression from a body of work that holds out the promise of effectively and employing past successes and commonly available knowledge to deliver the best building practices to all projects, everywhere, through the agency of ubiquitous and scalable computation.

Whether privately held as competitive advantage used consistently internally, or publicly offered to realize new sources of revenue by capitalizing upon latent worldwide demand, the AEC industry has the opportunity to move beyond the blank page currently on offer in its software environments to delivering its critical expertise at the right time, to the right places, to everyone who needs it, through the practices and methodologies of generative design.

Anthony A. Hauck is founder and president of Hypar, which is focused on advancing the art and practice of generative design in AEC through captured building expertise and scalable cloud technologies.

Appears in these Categories