Jon Andersen

Philosophically, AI has interested me for a long time. My nerdy side and fascination with the futuristic, combined with my general interest in technology and curiosity, drove it. When I heard about GANs (Generative Adversarial Networks) and Nvidia's faces, and saw the artwork Edmond de Belamy sold at auction, I had a big aha moment. It really caught my interest and pushed me to dive into what the technology could do and which tools were available at the time.

My interest in AI for architecture was truly sparked when the first generation of DALL·E and later Midjourney launched in 2022. That is also when I started experimenting with it myself. Initially it was a source of great frustration, dominated by chance, because there were significant control challenges — something we as architects need, even in early exploratory phases. But as a generative tool it was fun, fantastic for creating happy accidents and imagining new combinations and worlds.

I believe AI will mean that we as architects need to strengthen our core competencies even more. AI frees up time and resources so we can invest further in the essential parts of the process where we as architects truly create value. At the same time the technology makes image-making enormously accessible. We can use visual language actively to test as well as communicate our ideas with quality. It is universal, unlike many of our other forms of representation. That creates opportunities for us as architects to communicate architecture more easily and perhaps win over stakeholders to solutions that would otherwise be governed by stereotypical and generic ideas about space and architecture.

It challenges us, and we need to be highly qualified in our use of AI. We have to build awareness of how to use these tools critically, but also see them as a natural addition to our toolbox and an integrated part of our processes — saving time, money and resources and creating value for everyone.

That we can create images that can seduce, but also distort and manipulate, because scale, gravity (the laws of physics) and the spatial, three-dimensional aspect are not truly represented. There are also rights issues, built-in biases in the models we use, questions about which datasets they are trained on, and the danger of uncritical use.

I use it as a central tool in my own processes, but very much also in dialogue with students. I often try to get them to define qualified aspects to investigate and test, so it becomes a series of more targeted studies rather than random generations — where you sketch quantitatively horizontally instead of qualitatively in depth. It becomes an active part of one's sketching, informed by and closely tied to the process.

Claibe makes it possible for everyone to work with AI in a qualified, experimental, controlled and intentional way. Claibe's intuitive, tailored features, workflows and tools are built for architecture, design and creative practice in general. The tools give control and shorten the path to good results significantly, saving resources on many levels. Rights and legal aspects are accounted for, and the aesthetic bias is tuned toward a Scandinavian sensibility.