AI in architecture: Zaha Hadid Architects on its pioneering use and collaborating with NVIDIA We talk to ZHA about AI in architecture, its computational design advances, and its collaboration with NVIDIA on design, data and the future of AI and creativity Sign up to our newsletter When you purchase through links on our site, we may earn an affiliate commission. Here’s how it works. Experimental results from teaching text-to-image AI to produce structurally and contextually meaningful outcomes, AI in architecture (Image credit: Zaha Hadid Architects) BY JONATHAN BELL PUBLISHED DECEMBER 13, 2024 IN NEWS The debate about generative AI in architecture and creativity as a whole looks set to rumble on for perpetuity. Creators are divided into two camps; AI is either a tool like any other, or it is an existential threat that looks set to unravel the intangible magic of human creativity. Architecture might consider itself the mother of the Arts, but it is an intractable morass of disciplines and data. Could AI tools offer a path through the complexities of modern construction without sacrificing the visual expressionism of architectural design? Founded by the late Zaha Hadid in 1980, ZHA has always been a pioneer in computer-aided design. Wallpaper* spoke to the studio’s Associate Director Shajay Bhooshan, a co-founder of CODE, the studio’s Computational Design Research Group, along with Lead Designer and CODE colleague Vishu Bhooshan and ZHA Director Nils Fischer, who has been at the firm for 20 years. How have machine learning, AI and CAD evolved and intersected over the years? Experimental results from teaching text-to-image AI to produce structurally and contextually meaningful outcomes Experimental results from teaching text-to-image AI to produce structurally and contextually meaningful outcomes (Image credit: Zaha Hadid Architects) AI in architecture: what does the future look like? ‘We’ve done machine learning for many years for floorplate optimisation,’ says Fischer, ‘it’s what many people would consider AI.’ The studio was quick to explore the potential of early generative software, notably DALL·E, OpenAI’s text-to-image model first released in January 2021. ‘We were as excited as everyone else,’ says Fischer, ‘we’re not frightened of breaking things.’ As Instagram’s most popular architecture firm with 1.4m followers, ZHA had a massive visual archive, both public and private. One of the first AI experiments the studio undertook was to upload this dataset into Stable Diffusion, another generative artificial intelligence model. ‘Our team has been curating data sets and training models,’ Fischer explains, ‘[but they] are like a five-year-old child – they have fantastic dreams but not what you actually need.’
