From monuments to museums, houses to hospitals, architecture plays a fundamental role in everyday life. Generative AI can help architects create more functional and innovative environments, faster and more efficiently than ever.
It’s hard to overstate the complexity of creating an environment — be it a boundary-breaking skyscraper, temporary community installation, or even a seemingly simple tiny house.
Design is just one aspect of a larger architectural process that involves the consideration of building codes, structural engineering, zoning laws, environmental systems, and the client’s budget, just to name a few. Although technology like computer-aided design (CAD) has made architecture radically faster compared to the days of drafting, it still takes a monumental amount of brainpower and creativity to build a compelling structure or environment.
This is where generative AI can help. “Generative AI not only changes the ways architects work — altering the design phases and their traditional means of exploration — but also fundamentally changes the ways we see, image, and imagine the world around us. Ultimately, this transforms the way we operate on and design our environments for potential futures,” says Professor Emily C.S. Pellicano of Syracuse University School of Architecture.
By integrating generative AI into the architecture design process, architects can work more efficiently to create innovative built environments.
As with generative AI across all professions and use cases, architects are still exploring the many ways this new technology can add to their field. The following represents just a handful of the benefits of generative AI to working architects and architecture students.
“The thing that excites me the most about integrating generative AI into architectural practice is that the technology is becoming a collaborative partner in the design process,” says Pellicano.
Especially during the early phases of the process when architects are concepting, iterating, and developing their vision, generative AI can serve as a sounding board. For example, architects could query a generative AI model about recent or historical projects to get a sense of where their idea fits within the architectural landscape. Or, they could stress-test initial ideas using text-to-image capabilities.
For example, an architect could input a number of different text prompts into an AI image generator like Adobe Firefly to see what an unconventional concept might look like before sitting down to design.
One of the most exciting aspects of generative AI is that the technology doesn’t “think” the same ways humans do. Even the most creative people have preconceptions about what’s possible, and architects are no different. As Pellicano says, generative AI is “capable of making novel connections and identifying unanticipated relationships, patterns, and potentials that were previously withdrawn from our independent human perception.”
Unexpected or impossible-to-achieve results can open up new pathways for creativity that architects might not have explored otherwise.
Pellicano cautions students against using generative AI technology to simply get their assignments done more quickly — it’s critical that architects-in-the-making put in the time to expand their awareness and challenge presumptions. Instead of thinking about generative AI as a time-saving tool, think about it as something that can help architects spend their time better by minimizing rote manual tasks.
One of the most challenging parts of the architecture process is accounting for myriad factors from the very start of projects. If a curveball comes after the initial design phase is over — say there’s a misunderstanding with the budget, and now some of the building materials are no longer in scope — it can throw a project off the rails.
Because generative AI is trained on huge volumes of data that can encompass site constraints, budget, construction materials, code requirements, and beyond, it can help architects and their teams account for potential problems ahead of time and get everyone back on track (and budget) faster. Plus, having access to this type of rich data can allow architects to spot inefficiencies and opportunities they might not have otherwise, which can help architects get the most out of the space and make the best use of the materials — a huge plus for sustainability.
It’s important to remember that, while generative AI gives architects “superpowers,” it’s no replacement for critical thought and human empathy. The tremendous datasets that generative AI applications are trained on still can’t account for everything — especially the cultural, personal, and community impact of built environments.
Remember: generative AI is a collaborator, not a replacement for good thinking and human creativity.
While there isn’t a single generative AI application for architecture yet, students and professionals are already incorporating it into the design process as a complement to existing technology. There are generative AI use cases across the full architectural design process, but the schematic design and design development phases are a great place to start exploring the possibilities.
Here are some examples of what it could look like to integrate generative AI into the architectural process:
AI generators can help students and working architects explore a wide array of options fast before they settle on one design direction. They can start with an idea and text-to-image generation or use capabilities like Generative Fill and Generative Expand in Photoshop to build upon quick sketches, mood board elements, or photographs to get a feel for how the concept might look in practice.
In one of her architecture design studios at Syracuse University, Pellicano uses generative AI to help students find new ways to interpret, visualize, and understand the physical site they’re working with. Her students use a “neural style transfer” process and aerial site photographs to question, evaluate, and plan for the site. During this exercise, students start with aerial images of the site and then “transfer” alternative site images (for example, historical cartography, sensory or weather maps, legal tax maps, and so forth) onto the original aerial photo.
“This allows students to imagine alternatives outside of their preconceived notions of what the site should look like, how it is organized, or how it relates to its context in potentially novel and productive ways,” says Pellicano.
Starting with the base of a 3D render or a work-in-progress animation, generative AI can help architects quickly create vivid, illustrative videos using a text to video feature that give viewers an immersive experience of the design. Video is an important tool for helping architects choreograph the flow of space and how people should move through it.
Most architects will use 2D and 3D computer-aided design software to create a detailed vision of the structure they’re designing. With generative AI, they can quickly add context and style to images, such as landscape design, building materials, textures, and colors, and additional elements like furniture, hardware, and other interior design elements. Plus, with capabilities like outpainting — known in Photoshop as Generative Expand — architects can enlarge images or easily change the aspect ratio while maintaining the look and feel.
Turning a flat, 2D design into a 3D model can take a lot of time. With new technology, however, generative AI can make the task a lot faster and simpler. There’s still value in having a working knowledge of 3D modeling as an architect, of course, but near-instant 3D rendering can help expedite last-minute changes in a pinch using a 3D art generator.
Just as artificial intelligence has made its way into everyday life — consider autocorrect, for example, or the map app that helps you avoid a traffic slowdown in real time — generative AI is set to become a seamless part of the architectural design process.
“I expect all architects to begin to incorporate AI into their work, whether they are aware of it or not,” says Pellicano. “It will become more and more incorporated into our standard software packages and operate behind the scenes in ways we are largely unaware of, or in unassuming ways.” Generative AI technology will act as an extension of the architect’s creativity and vision, opening up new realms of possibility the way technology always has — from the ruler and compass to the drafting machine to computer-aided design software and beyond.
Pellicano’s advice to architects wanting to be on the cutting edge of generative AI technology? Get to know how generative AI works. Learn about the datasets different models are using, explore how content synthesis works, and experiment, experiment, experiment. It’s the unexpected results — the ones a human brain would likely never conceive of — that hold the most potential for innovation.
“I would recommend that anyone seeking to work with generative AI approach the technology with curiosity and the willingness to lose control a bit,” says Pellicano.
Contributor
