Everyday, we interact with technology that expands our abilities and alters the way we live: work, learn and play. Artificial Intelligence (AI) is one such technology. Virtual assistants like Siri and Google Assistant have become a ubiquitous part of our daily lives. Our personal devices use Natural Language Processing (NLP) to understand and respond to our voice, and they have become indispensable to many.
Innovations that once took decades to mature are now reaching market readiness in a fraction of the time. While it took the telephone 75 years to reach 100 million users, it took Twitter 5.5 years, and ChatGPT only two months. Cloud computing and digital reach have contributed to the accelerated growth of these platforms, but societal shifts are also a factor with the increasing cultural aptitude to more swiftly leverage technology. This acceleration is not without its challenges, however, as it demands human adaptability, business agility, and thoughtful policy regulation to ensure that the intended benefits are aligned with societal values.
A week after OpenAI launched ChatGPT, my brother, who works in tech, asked if I’d heard about it. He said that he believed the impact of the new AI would be on par with that of the internet. Earlier in that Fall I had given a keynote on architecture and technology and little did I know, that talk was severely outdated. It was a reminder of the pace of change, and that we -architects- are at the receiving end of technology. This last year brought huge strides in AI “decentralization” — the trend of expanding access to advanced AI technologies that were previously available only to players with access to massive, centralized, proprietary data sets. Products such as ChatGPT, Mid Journey, Dall-e2, Stable Diffusion, have risen fast and are only projected to expand, offering a huge role to play in design and architecture.
From early computational design in the 1960s, computer-aided design (CAD) systems, building integrated modeling (BIM), to rule-based parametric systems — the role computers play in the design process have evolved over decades from simple drawing tools to intelligent parametric design system — all which laid the groundwork for the integration of AI.
Creativity is irreplaceable by AI; human designers can translate complex emotions and cultural contexts into tangible designs that address the nuanced needs of the human experience; something AI cannot relate to.
Generative AI text-to-image generation tools like Mid Journey, Stable Diffusion, PromeAI, Veras and LookX have allowed us to expand our thinking into visuals very quickly. Like the digital virtual assistants, these tools use NLP to generate images that correspond to natural language text prompts, making them great tools for storytelling, to visualize ideas, and to layer vignettes to support design narratives. At Perkins&Will, we are using AI-generated imagery to explore visual references, to interrogate moments within a project, to instantly render images from a sketch or a wireframe 3d model, or to explore detailed façade tectonics. In concept ideation, text-to-image tools might help create a poetic visual to capture an idea or provoke new thinking, but designers are also at risk of being distracted. Image generation tools can create noise that disrupts an intentional and well-structured design process when not considered thoughtfully.
As designers play with this astounding ability, it is important to remember that AI generated visuals are not designs nor do they reflect conscious resolutions to design challenges. The intelligence needed to create images from textual descriptions does not equate to the problem-solving skills inherent to designers across multiple disciplines. There is creative value in starting a project by developing a clear concept based on the designers’ comprehensive understanding, then using generative AI tools to process the concept, augment and explore further. It is crucial to be intentional with how and when to use generative AI in the design process. Our role as designers in interacting with Generative AI is to direct the process, relaying our understanding of the project, its context, design ethos, and identity.
Another game-changing application of AI in design is performance simulation and prediction. Building performance simulation for daylight, wind and energy are helping designers better understand the effects of design decisions and reach increasingly more stringent performance targets. However, time constraints are currently limiting for design exploration, especially when evaluating a wide array of design options enabled by generative design.
Machine Learning surrogate modelling offers drastic speed improvements to simulation processes and has allowed designers to access performance analysis and prediction in near real-time, cutting down the time-intensive computation required for traditional simulations. For example, Perkins&Will’s AI engineers introduced a method for predicting spatial distribution of annual daylight metrics using a raycasting-based encoding of the inputs and a deep-learning surrogate model¹. Using synthetic data, surrogate models trained and achieved low average errors and are giving a high degree of confidence over predicted simulations. This allows designers to interact with real-time feedback on daylight quality and performance while making design iterations, comprehending how design decisions influence the ability to achieve performance targets.
In the future, AI has potential to optimize production workflows, bridge the gap between design and manufacturing by connecting design, engineering, and product so designs are optimized for production from the outset. Once built, predictive analytics based on sensors data would help smart buildings measure their performance, anticipate and respond to human behavior, and actuate a change in passive or active building systems to change light levels, noise, or air quality and adapt to human needs across the day. Since the most responsible architecture is often in not building new, I imagine AI serving to query and analyze existing building stock in urban areas, evaluate against local codes and regulations, and suggest opportunities for adaptive reuse and transformation. As technology advances and its applications grow, the possibilities become unlimited.
Our highest value as designers is to think critically and be creative. AI can provide us with tools to unleash creativity and preserve our highest value as humans. A logical start is to employ AI to automate the mundane, help us arrive at solutions faster and more accurately, so we can spend more time on the creative aspects of our work. As AI becomes more integrated in the way we work, critical thinking is needed to ask the right questions and evaluate the outputs of AI tools. Creativity is irreplaceable by AI; human designers can translate complex emotions and cultural contexts into tangible designs that address the nuanced needs of the human experience; something AI cannot relate to.
Preserving the human element — creativity, intuition, and ethical judgment — is paramount in an AI-saturated landscape. Architects are to remain at the helm, directing AI as a tool for innovation, as a powerful assistant, elevating human ingenuity rather than eclipsing it, ensuring that the heart of design remains human, even in an increasingly automated world.
