Ahead of the curve

Will we live differently when buildings are created by artificial intelligence? Will the architect’s profession still exist in the future? And who owns the copy­right in a world where computers take over creative processes?

One office that embraces the new technology is Zaha Hadid Architects. The London-based practice made a name for itself with parametric computer-aided design that defies the rigours of the right angle and instead uses organic, fluid forms to realise a new complexity and sophistication in architecture. Since the death of the company’s founder, Zaha Hadid in 2016, the office has been led by her former office partner, Patrik Schumacher. We talked to him about spatial composing, verbal sketching and journeys into the metaverse.

What do you think artificial intelligence is: a friend or an enemy?

P
S
It is a friend. I have always been excited about new technologies, and we have been using different computational methods for many, many years. The architecture discipline has been empowered by the computer revolution for ­decades, especially in the early ­design stages, with generative ­algorithms and so on. AI is another version, another way. A new tool that makes creation very easy for everyone. So there is a massive democratisation taking place. But I’m not afraid of that.
-
-
-
-

When did you start working
with AI?

P
S
As far as the new wave of image creation through prompting is concerned, we have been aware of it for a few years. We started experimenting with the artist, ­Refik Anadol, who was also one of the early adopters. For the DDP Design Museum in Seoul in 2022, we collaborated on the project Architecting the Metaverse. This was before the image generator DALL-E 2 came out.

The results of the first version of DALL-E were not coherent enough. They were a bit collage-like, and therefore not quite harmonious. When DALL-E 2 was released 18 months ago, it was a real leap forward in terms of the results we got to see. At the time, architect and theorist Neil Leach spoke of it as a Sputnik moment. And the results were truly spectacular. We realised that this could be very strong.
-
-
-
-

What is the key to creativity in the process?

P
S
I think it is in the system, in the prompt. If you work a lot with AI, you learn how to prompt and how to do it faster. You know what you can do and what you can’t do. There is a lot of trying out and learning and getting familiar with the system. It is a fascinating world because it is more than just automation. I think it increases creativity and also productivity enormously. You have a surplus of things.

This has been the case with many of these algorithmic tools. We were able to do things that would not have been possible without them. So it gives an additional, creative potential. At the moment we are looking for functions that are more usable for professional applications in a highly directional way. In this way, we can set certain parameters which the output has to meet. Therefore, we are currently developing our own versions of these tools.

-
-
-
-
Can you explain this in more detail?

P
S
We have a whole research team working on the development of our own system based on Stable Diffusion, an open-source system. When we model something and take screenshots, we can elaborate and render them in detail. Or we create a series of images and if we like certain aspects and parts, we can freeze them and then iterate around them. So we can home in on something. It is quite fascinating.

We are using our own tools the more they evolve. And we ­also run these systems on our own server farm. Because during ­COVID, we really fully upgraded our systems and have a centralised data centre with all our machines. At night we can use it as a render farm to generate AI. So that starts to work out quite well. We are excited. I think the discipline needs to tailor these systems to our own more particular needs. Basically, they are very open-ended at the moment.

-
-
-
-
What are the advantages of it?

P
S
The images are generated in rendering quality, so clients can quickly imagine something, even if the visuals are not to scale or in the right size. They are like concept ideas. So that is what we have been using it for. I mean, none of our own designs that are out there have been influenced by it yet. So this is more recent work and nothing built. But it brings interesting input into the process, getting more ideas ongoing in a project: not only the first sketch and then running with it, but more as an iterative feeding into the project. This is still at an early stage.

But especially in the world of competitions, where you only have a few weeks to create a complete, fully rendered, photorealistic design, that is where it all happens. Of course, we know how to execute and how to deliver complexity. That is not easy to imitate. But we want to be faster and have more diversity of options early in the process to be more innovative, more creative and also help the client to see more possibilities.

-
-
-
-