Building In The Future
How architects around the world are using generative AI, and what it means for young designers.
I recently travelled to Stockholm and saw an exhibition at Liljevalchs+ about the architectural practice of Dorte Mandrup. Mandrup leads a Danish architectural studio by the same name and is known for her buildings in remote environments (see the Icefjords Centre in Greenland). The project that stood out to me in the show was The Whale, a visitor/research centre 300 km north of the Arctic Circle with a structure resembling that of a breaching whale. It wasn’t until after I checked round-trip flights to the Arctic town of Andenes that I learned The Whale was in pre-construction and had only secured 30% of the funds needed to start the project.
The Whale’s renders were likely created from pencil and paper drawings that gradually became digitized. But, it prompted me to explore how architects are using generative AI to bring new projects to life, and how generative AI is changing the design process for young architects.
The Whale, a proposed new visitor/research centre in Andenes, Norway designed by Dorte Mandrup architects. © Dorte Mandrup.
1. Zaha Hadid Architects
Architects at Zaha Hadid (one of the world’s pre-eminent firms) use machine learning throughout the design process: from ideation, to stress-testing the legibility (navigability) and structural integrity of their projects.
During a roundtable discussion (“AI and the Future of Design”), Principal Architect Patrick Schumacher spoke about generative AI. He is mesmerized by how many potentially great projects can be generated with DALL·E 2 and Stable Diffusion. He says, “Not everything is useable but often there’s one or two great sketches that could be taken forward.”
Examples of concept images generated with DALL·E 2 in the style of Zaha Hadid Architects. © Patrick Schumacher.
Schumacher shared a series of interior images (below) generated with DALL·E 2 in the firm’s signature style. The Changsha Meixihu International Culture & Arts Centre in Changsha, China borrowed from these concepts, and the final project has been recognized for its imaginative design.
Top: Patrick Schumacher showed the audience these images generated with DALL·E 2 during the “AI and the Future of Design” virtual conference, April 2023. © Patrick Schumacher and Zaha Hadid Architects. Bottom: Changsha Meixihu International Culture & Arts Centre, China designed by Zaha Hadid Architects. © Virgile Simon Bertrand and Zaha Hadid Architects.
Schumacher encourages all architects who are working on competitions and early ideation to use generative AI tools and see what is possible. Zaha Hadid Architects employ a small team of AI Research Scientists, who are now working on creating a specialized model for the design firm to use for their own ideation and competitions.
2. Rojkind Arquitectos
Rojkind Arquitectos was founded in Mexico City by Michel Rojkind, who is known for his “experiential innovation” and adoption of new technologies. He says the relationship between clients and architects is changing as a result of generative AI, because of the speed and accuracy of text-to-image models.
Rojkind is now able to “bring clients on the ride” from the earliest stages of a project. Prior to Midjourney, Stable Diffusion, and DALL·E, Rojkind’s firm would have to create collages of images (similar to Pinterest) and share these with clients once complete. It was time consuming, and hard to extrapolate these flat images to three-dimensional models. AI affords new ways of working with clients: now, an architect and client can sit together and generate images in real-time, increasing the speed of feedback and establishing a shared vocabulary.
Rojkind also commented on his evolving relationship with AI saying that, “AI is learning from us, and we are learning from AI.” To him, generative tools have secured a place in the design process.
3. Xkool Technologies
Wanyu He is CEO of Xkool Technologies, a software provider for architects. Xkool was originally focused on machine learning for data and optimization to help architects work within budget constraints, material constraints, and other inputs.
The company is now expanding their products to include text-to-image and image-to-image generation. He identified a gap in the market: current generative AI tools didn’t allow for sufficient user inputs, beyond a text prompt. Xkool responded by building a series of fine-tuned models that work with multi-modal inputs. Architects can upload paper and pencil sketches, basic floor plans, and text prompts to generate 3D models. This approach saves time and helps architects iterate on early concepts.
AI-chitect a product by Xkool © Xkool.
Generative AI is a game-changer for architects, especially who are at the beginning of their careers. With generative AI young architects can iterate rapidly, bringing new and imaginative designs to clients. The technology not only streamlines the design process but also provides valuable insights and optimizations, enabling early-career architects to create more innovative and sustainable buildings. As the field embraces these advancements, the future of architecture will witness a new generation of architects who push the envelope, challenge conventions, and shape the built environment in extraordinary ways.