AIA Canada Society Journal: Navigating AI’s Impact on Visual Communication in Architecture

How does using AI change the role of architects in visual communication?

If the rise of artificial intelligence feels eerily familiar, it’s no wonder.  We’ve seen how social media reshaped human interactions, how streaming changed the entertainment industry, and how the internet itself rose to provide the infrastructure for all of it. The difference is that with AI, we find ourselves in the early stages of its adoption and societal integration, with an open and exploratory journey ahead. The question for architects: how shall we navigate this emerging and unproven path forward?

Artificial intelligence (AI) is a branch of computer science. It is concerned with performing tasks by computation that traditionally require human intelligence. Such tasks may include reasoning, problem solving, and language processing. Especially relevant to the architectural profession is AI’s ability to understand, interpret, and generate imagery—in short, to visualize.  

If we grant that visualization is architecture’s main currency, what does it buy for us? Just about everything. Collaboration, coordination, consensus, exploration of options, client communication, decision-making, construction, marketing, energy modelling, and more—they all depend, ultimately, on some kind of image production. Visualization is a fuel that propels projects forward.               

The natural question that arises is: if visualization stands as a core architectural skill, what happens when we entrust this task—either partially or entirely—to AI? As we navigate through a rapidly evolving landscape of increased AI adoption and the emergence of new tools at speed, the answers to this question are not yet clear. What is evident is that the implications span from the practical aspects of daily work routines to the very essence of what defines the role of an architect.

This Prairie house concept by architect John van Hemert was used as a reference image in Midjourney.
Based on the reference image, Midjourney was asked to imagine an interior courtyard for the same house. It was also prompted to use black wood cladding, to include a water feature and vegetation, and to render the image as if taken on a DSLR 50mm camera. The time spent using Midjourney was five minutes.

Visualization with Generative AI

Generative AI, a subset of artificial intelligence, involves training a computational model on a large set of existing data. In AI’s visualization tools, the models are trained on large sets of existing imagery. The model learns the patterns and structures in the data, and when given instructions from a user, it can generate new content by extrapolating from what it has learned.

Midjourney, one of the many generative AI image tools on offer, outputs original imagery from user-inputted text. While it is far from perfect, one only needs a short amount of time with tools like Midjourney to understand their seductive promise: rapid idea generation, real-time collaborative brainstorming with clients, unlimited options in a few clicks, often with compelling and surprising results. In those early stages of design, when the question ‘what if’ is asked so often, design teams can almost instantaneously generate propositions to react to. Ironically, using the tool can be likened to sketching, even though there is no drawing involved: it is imprecise, exploratory, and holds the possibility of discovering happy accidents. 

Other AI capabilities at various stages of development include: the automatic generation of code-compliant floor plans based on a footprint and functional program; the revision of floor plan options in real time to respond to footprint and/or program changes; the rapid production of options-heavy feasibility studies; the automatic generation of building forms that respond to site constraints; and the automation of repetitive and time-consuming tasks. What’s more, the tools generally dispense with steep learning curves and complicated interfaces, thereby democratizing their adoption beyond the architectural profession.  

Just as the benefits of generative AI are easy to spot—and even rally behind—with enthusiasm, there are pitfalls which can only be framed as questions at present. Are the AI outputs really any good, practical, or useful? How can we trust their quality, and should we? What if the data sets were trained with an inherent bias, either cultural or formal? Is having more design options necessarily a good thing? How about the mountain of possible intellectual property and copyright issues? Can machines really acknowledge and synthesize the complex array of project-specific contexts and constraints with sensitivity and elegance? How about liability concerns? Will staffing levels be forever changed in a world where so much more can be accomplished with fewer resources? Will there be a loss of craft and skill over time as human intelligence is outsourced? How will this change architectural education?

As bewildering as such questions may seem, there is a group uniquely positioned to navigate through the uncertainty: architects.   

As custodians of a multi-disciplinary process that can last years, architects are looked upon to provide collaborative leadership through the design process. Making sense of often competing cultural, economic, poetic, technical, functional, and sustainability objectives is a core job description.  While AI stands to transform how some of the visual products of design are produced, the design process itself will remain a complex collaborative act, so long as the need to synthesize the disparate concerns of human beings remains.     

Similarly, architects are harmonizers of complexity. While AI may reduce the complexity involved in performing certain discrete tasks (such as producing multiple rendered design options), its widespread adoption may indeed amplify the complexity of the design process as a whole. If AI’s alluring promise of democratizing design through its compelling outputs and ease of use is met, architects may find themselves cast in the role of ‘curator’ instead of ‘creator’ at times, as the sheer volume of design ideas brought to the table mounts. In such a role, hard-earned abilities such as context sensitivity, compositional discernment, and multi-disciplinary coordination may be more essential than ever to separate the signal from the noise.

Deep down, architecture is concerned with human well-being. As such, architects often find themselves taking on some big questions: Will a design enhance the psychological and emotional well-being of its users? Will it foster an environment conducive to positive social interaction and human connections? Will it be sensitive to the cultural identities and traditions of its occupants? Can it adapt to changing circumstances to support long-term well-being? Will it be comfortable?     

AI’s image-making power will not answer these questions with the click of a mouse. It may, however, offer the gift of time found through the automation of the repetitive tasks baked into the process of image generation by conventional means. In this sense, AI stands to augment architects’ capabilities, and to encourage greater focus on leading a complex and collaborative process in the service of human well-being.  

John van Hemert, Architect, AAA, CPHD, MRAIC is a practicing architect in the province of Alberta. He holds a Master’s degree in Architecture from the University of Calgary. His interest in Artificial Intelligence and its effects on culture began while completing his Bachelor of Arts degree in Science, Technology, and Society, also from the University of Calgary. Before becoming an architect, John was a computer programmer specializing in the design of large data structures.