Kaan —
How do you describe your approach towards architecture or urbanism?

Roberto —
First, I'm more focused on urbanism and urban design. That has always been my area of interest. My approach to it is very much affected by data and the use of data to analyse and structure the urban environment. The data, for me, is a kind of new metric, a different way to measure things and to represent them.

Data, by definition, doesn’t come with any given or particular scale. It's a system of obstruction that can be applied to many factors in the city. Some of them could be empirical, like things that exist in the city, but these empirical data sets may not be sensible to humans. For instance, pollution could be entirely non-material, being entirely non-empirical in its manifestation, but still real. Non-empirical data could be varied from happiness to issues of discrimination in the city. These considerations very much affect how I see my approach to design.

Kaan —
What led you to pursue digital technologies in architecture and urbanism? What triggered your interest. Was it an event, a book, a project, or was it just your interest?

Roberto —
I've thought about this many times. I think it was a mixture of things. Architecture education was a single five-year course at that time, and halfway through it, I don't know why I got interested in computers. Then I got a birthday present from my parents, a computer i486, which doesn’t even make anything, but it was okay for me. Then I installed AutoCAD light, basically AutoCAD, but with fewer functionalities. As soon as I started using it, I felt incredible. The possibility that caught my attention was that I could have designed conceptually, meaning that I could simply concentrate on the process. By controlling the process, I could have a randomized or a totally clear result.

Kaan —
You are the author of Digital Architecture Beyond Computers. In your book, you present a fragmented yet interconnected historical narrative, using computer-related technical terms as chapter themes. How did you develop this non-linear, associative structure, and what advantages does this way of story-telling offer over a more traditional framework in discussing computational design?

Roberto —
Funnily enough, probably the structure of the book is the thing that took the most time—even if, in the end, I didn’t really talk about it much in the book. It took a long time because I didn't want to create a purely historical and linear narrative. After all, chronological sequence was perhaps not the essential quality of digital or computational design.

Hiroshi Kawano, KD 29 - Artificial Mondrian, 1969. © Hiroshi Kawano, Photo, ZKM Center for Art and Media Karlsruhe

The core quality was always processes and operations. That, to me, was more important than what happened when, or if something happened before something else. It also matched the content and the conceptual structure of computation better by emphasizing the process and assigning less importance to the chronological account.

For instance, when two things work well together in the narration, this structure provides an excuse to move freely through time and jump 150 years. However, I still used chronological organization in each chapter because it would've become very difficult to read if everything was just a random sequence of processes.

Kaan —
If digital architecture goes beyond computers, how could it play a role in defining architectural intelligence in the age of AI?

Roberto —
Let's go back to what I was saying about when how I fell in love with computation. As I said, I could find a way to think or design through processes not through products. I was thinking about how to make something, not what I would make. At the time, that was easier for me with the computer.

The traditional idea of computation in a creative environment for pretty much 70 years (from the end of World War II right up until the diffusion of these very powerful AI platforms) was based on using the knowledge coming from many domains of intelligence through the computer. Your job as a creative person was to formalize that knowledge into some set of instructions you could give to the machine to do a task. These instructions could contain randomness, meaning you didn’t really know what the machine would do, but you knew the process.

Grazia Varisco, Variable Light Scheme R. VOD. LAB, 1964 and Marina Apollino, Dinamica Circolare 65+S, 1966, installation view in Electric Dreams, Tate Modern, 2024. Photo O Tate (Lucy Green)

Now, with AI, the situation has inverted. An AI or a neural network is nothing but a software that designs itself. In other words, you give the software an input and you want a certain output. The task that the software needs to perform is how to take the input to an output, which means the software designs the process because the process is what links the input to the output.

Due to the flipping of this relationship, our intelligence assesses the output. Our job is to surveil the automated outputs of AI-invented processes. I am still unsure about this word, but I would say that our forms of intelligence need to be more curatorial. Like in an exhibition, a person would probably not be the author of the work that is going to be displayed, but they still need to read and organize it in a way that makes sense. If it's an exhibition on paintings, the curator has not painted anything, but they know why these paintings need to be together in a space. If I'm not making things too complicated, we are basically in a situation where the machine produces outputs, and we have to think of forms of intelligence to interpret them—not to create them, but to interpret them.

Yuan Ping, Wang Huiye, Liu Jie, Wu Yu, Ping Yuan, Sensory Balance, B-Pro Urban Design RC14, Bartlett School of Architecture, University College London (UCL), London, 2024.

Kaan —
What aspects of the design process in the studios do machine learning and wide data sets impact?  What kind of urbanism results from the implementation of ML models, conceptually or practically in the studio environment?

Roberto —
First of all, the studios that I've taught are always focused on London because I want students to be able to go and see places. But the novel element, perhaps the most unique component, is the use of data in the process of design.

Basically, in the studio, the work is very much about studying cities through data. Part of the task is to use data just like any other material an architect uses, like wood, steel, or glass. You have to know the material. You need to break it and try to make a model with it. You have to test it. Students do the same with data, which means testing data sets. They learn how to apply statistical methods to data to highlight, maybe the distributions of things or correlations between data sets. Students have to learn, which is always the most difficult part—learning how to read these objects.

But why is it difficult? Because they are confronted with scatter plots, with techniques of visualizations, that are far from the traditional ones architects or urban designers use. Just like you need time and experience to start being able to roughly read an architectural plan, whether it is a church or factory, in 10 seconds, students begin to get familiar with these data sets in tutorials. Once you have become more literate, which parts of these representations are useful for your design starts to be clear. There may be outliers, data that do not fit a certain statistical distribution, but it's still interesting to work on and to discover relevant correlations for their design.