Framing the debate

Much of what we hear about the rise of AI is presented in an over-simplified, perhaps even populist binary. The received narrative goes something like this: on one side, AI is going to free us from drudgery at work; on the other, it is a threat to jobs, particularly ones that involve creative work.  

The starting point for architects and designers confronting today’s AI has to be taking a critical view of this framing. There are dangers and disadvantages to seriously consider, but there are also potential advantages – and the complexity of the issues at stake demand more nuance than a black-and-white choice.

Two fundamental questions need to follow. First, what exactly is AI in its current and foreseeable forms? Second, how can it be used in architecture and design?

What exactly is AI?

On the first question, some perspective is required. Without getting into the philosophical nitty gritty of what constitutes human consciousness or creativity in the first place, the important point to note is that much of what is called AI today might actually be better understood as something like a data-crunching search engine – a phenomenally powerful search engine, but a search engine nevertheless. 

Yes, generative AI does indeed produce or ‘create’ things, but it doesn’t do so ex nihilo. Platforms that have gained notoriety in recent months, for example, construct images based on a brief text prompt but they do so by processing vast amounts of data that already exist. It uses images and descriptive associations created by humans to come up with new forms. While these new forms might seem remarkable, they are effectively produced through a capacity to process unfathomably large quantities of data; it is a mechanical form of ‘intelligence’ based on numbers and connections, limited only by the amount of data available and the computational capacity to process it. 

If that matches your picture of what human creativity is or has always been, then so be it (the Deleuzians will have you). However, it seems fair to say that many of us have a different understanding. Human intelligence surely involves something more qualitative than the quantitative nature of data processing, however powerful it might be. 

Leaving aside thoughts on where AI might get to in the future – will it ever be genuinely generative rather than data-driven? – its current state leads to the question of how it can, will or should be used in design right now. As with most technology, it’s often not the intrinsic qualities that matter so much as the ways in which it is employed.

Related: The design philosophy of Woods Bagot’s Domenic Alvaro

Human-AI collaboration

Collaboration is our best chance to make something valuable out of AI in architecture and design – that is, collaboration between humans and AI. It’s really about going back to the core principles of the design process. Designers are taught to work iteratively in a process that involves repeated refinements, all the while anchored by an overall concept and parameters such as costs or client demands. 

Once upon a time, this iterative design process involved more tracing paper and hand-drawing after hand-drawing. Digital design tools have changed much of that and in fact the emergence of CAD and other digital tools presents something of a case study for the AI debate. New tools can save a great deal of time but the equation remains similar: how can new technologies be embraced without losing sight of what is already valuable? The best designers, for example, don’t use digital renders alone but find ways to integrate that form of visualisation into a wider design process. A similar problem presents itself with AI and the question of how to keep a technology in its place, so to speak, will be considered in a future comment piece.

The sheer power and speed of AI means that parts of the design process can be greatly improved. Say, for example, that a building needs more windows on one façade or perhaps a little less Le Corbusier in its form – a quick prompt for an AI generator could create new images in seconds. The crucial point, however, is that these new images are not a simple solution, nor have they appeared out of nowhere. They are curated by a designer – formed out of a concept, subject to parameters, human preferences and the knowledge of the architect, and then used as a spur for further iterations. 

A new image is created based on previous iterations, and the designer can evaluate it and move on with further refinements. The only difference is that these links in the iterative chain can happen much more quickly. The greatest risk for architects and designers seems to be that the means become the end – woe betide those who forget that this technology forms only one part of a much wider web of the creative process.

Innovators are already working on platforms that involve human-AI collaboration in real time. This might involve working with an continuously evolving image, for example, by sketching on a tablet by hand at the same time as providing text prompts. As we will see in later pieces, the key skill then becomes a curatorial one as the human designer still needs their skills, knowledge and judgement to direct the process and then decide when it’s finished. We will also be considering implications for jobs and education in design.

We think you might also like this book review on Richard Francis-Jones’ Truth and Lies in Architecture.