Partner James Loder has long been an advocate for the integration of new technologies in our design processes, while not becoming reliant on or beholden to them. As the floodgates of generative artificial intelligence (GAI) based tools have spilled into our world we’ve created a working group that shares their individual areas of research with the wider practice, their focus is wide ranging as we grapple with previously unrealised issues around ethics, security and intellectual property to name a few.

James joined Architecture Australia in a roundtable discussion moderated by Roland Snooks and Gwyllim Jahn around the impacts of this technology on our practice.

How is GAI different to other forms of concept-generation practices?

We see GAI as a tool to be used in support of conceptual development and not as a comparative method. We find it is most effective when we have a clear direction and can provide specific prompting to produce images that complement our conceptual narrative or a particular design decision. Using AI image generation to produce focused and curated images in lieu of searching for relevant precedent imagery is a much faster and more targeted method in support of an idea. The value of these images lies in their capacity to be evocative and compelling in unexpected ways, but their purpose and intent is almost always predetermined by the traditional practice methodology for conceptual development.

What are the unexpected qualities or characteristics that have arisen in this process?

Utilising sketches, diagrams, renders or other graphic content as visual prompts often produces counterintuitive outcomes that, through their realism, become convincing alternative approaches to a particular design task and present new avenues for exploration and experimentation. This process goes beyond the passive engagement of inputting key words into a black-box AI model and assessing the outcomes – instead, fostering an almost collaborative relationship where the back-and-forth exchange of drawings and images produces outcomes that often defy initial expectations. A balance between the designer’s intuition and the AI’s ability to process and reinterpret visual cues results in a blend of human creativity and machine intelligence.

Who are these images intended for?

Initially, our intention was to utilise these AI-generated images strictly for in-house purposes – mainly to explore and refine conceptual ideas. But, as we experimented with various prompts and observed the diverse outputs, we discovered that certain images held significant value in communicating an idea to a client in the early phase of conceptual development. This approach has allowed us to move beyond the conventional reliance on precedent imagery, which was often confined to examples of our own past work or limited to abstract imagery and diagrams.

Ultimately, what makes these creations intriguing is the human touch. When a machine generates content, it often lacks emotional connection – the human element. There might emerge a new appreciation for what’s “handmade” or derived from human creativity. Even though anyone can replicate imagery, the images, no matter how beautiful, require interpretation for actual construction. Humans want to hear from other humans, not machines.