Arturo Tedeschi is one of the world’s leading experts in computational design, and a consultant for companies from architecture to the apparel and automotive industries. I interviewed him 8 years after the publication of his most famous book AAD Algorithms-Aided Design to talk about the present and future of computational design, the computational designer’s role, Training, Virtual Reality, Ecology, 3D printing and Machine Learning applications to Architectural Design.
I first met Arturo in Rome in 2014, on the occasion of one of his Algorithmic Design courses that he runs all over Italy. Even then, a few years after the first version of Grasshopper, Arturo was considered a reference point for training in algorithmic design and parametric architecture. Today Arturo works mainly as a consultant for design studios and international brands. With his group, he’s also the author of research aimed at the continuous testing of innovative digital techniques, from VR to machine learning.
Giuseppe Gallo: The last 10 years have been crucial in the diffusion of computational design. How does this affect training?
Arturo Tedeschi: the training activity changed radically. I can see three moments in my teaching activity, but I think it can be generalised. During the first one, more or less from 2009 until 2014, computational design was a totally pioneering topic. I remember I started studying grasshopper and publishing my books when there was, also in Italy, a small community of people, architects, designers, and people interested in understanding how to push the boundaries using the new digital tools. That first age was more like an inner circle. It was really about having an exchange with people younger and older than me. I remember there were experienced architects suggesting a direction. For example, they were trying to solve specific problems: things they couldn’t do with traditional CAD methods.
And then a second phase, more or less around 2014, when I published AAD Algorithm Aided Design. People considered Grasshopper just an extension of CAD software, and in my opinion that was a bad moment. This is because students and young professionals were approaching Grasshopper as the latest tool to exploit.
I’m mentioning Grasshopper because it is the top of the iceberg in the computational design software, but when I started, there were so many tools out there. One really interesting was Generative Components, around 2010 if you were interested in computational design, you had Grasshopper 3D on one hand and generative components on the other hand. Grasshopper was for free, there was a community around the software, and it was the main reason a young designer decided to go towards Grasshopper rather than the other option. In the period between 2014 and 2020, Computational design training was about teaching and learning software. Grasshopper has become used in so many industries, for so many design topics, and suddenly, I found Grasshopper in automotive companies, fashion design, and so many different fields.
At that moment, training was about teaching and learning a tool. But, as you well know, I always try to defend a different vision: to build a strong foundation. This is because I started as a 1.0 architect, doing not diagrams but drawings, by sketching everything, by having a comprehensive idea, in a top-down process, not a bottom-up method. As computational designers, we need to have a strong foundation in doing research and studies, not just to use a particular piece of software, but to solve real problems.
And then we have 2020 when the pandemic forced a lot of students into their homes, training and teaching changed completely, becoming an online activity.
At that precise moment, there were many people who started teaching online. You could find a lot of webinars, recorded courses or whatever, a lot of possibilities, and I decided not to follow that trend.
I took a break from teaching because I guess I don’t fit anymore in this new wave of interest. Also, I am reorganising my idea, so the next step will be probably to offer something different. I don’t know yet precisely what to do, but I guess it’s not possible to teach computational tools anymore as we did in the last ten years, and mainly I guess it’s important that this kind of confusion, should be evaporated with a new theoretical framework to provide to students.
G.G.: Today there are dozens of different digital tools for design, and I also feel is a risk of getting lost in a purely commercial narrative. In this sense, as you said, building a methodological foundation is fundamental. Can you explain what it means?
A.T.: This is one of the main differences between now and my beginning: computational tools are not just Grasshopper 3D, but also something like Blender, something like Houdini. There is now a kind of grey area, where different tools merge, blend, or just stay one on top of the other.
Young students who are interested in advanced computational design don’t know exactly what to do. They just have a kind of idea of, you know, doing cool stuff, complex things. But, if you don’t start from research, if you don’t have a vision, or, this is also very important, if you don’t know the literature, if you don’t know the architectural design history or if you don’t know the evolution of tools: you are just fascinated by cool images out there. That’s what I mean when I talk about building a foundation.
This is what we need because computational design is not just about how to create complex shapes with a connection with geometry, or with advanced digital fabrication tools for example, not about doing cool, or about pushing the boundaries for the sake of pushing with no kind of idea behind that.
Another important influence comes from the growth of social media: Instagram fuelled a new culture based on images and this kind of instantaneous, quick feedback. So you publish something which looks astonishing, and powerful in terms of visual effect, and that’s it. Everything is about this idea of gamification, pervasive, everywhere, and everything is based on quick scores. If you get, you know, 2000 likes, your image, your product, your idea is good, otherwise is not. Again, everything created this kind of grey area, where is difficult to understand what a computational tool or platform is nowadays.
Many students come to me and ask: “Ah, ok, I see you do cool stuff. Is it Houdini?” Or “Should I learn this software or the other?”, “Is it about doing everything using virtual reality?”
From my perspective, everything looks like the same thing and should be intended in total continuity. Look at form-finding strategies, and their master, people like Gaudí, Heinz Isler, Sergio Musmeci, and Frei Otto. They started with analogue systems, studying and generating complexity with an intelligent approach. But now there is this kind of misconception on how to use tools, and how to manage them.
G.G.: You talked about literature. Can you suggest three books useful to build a strong foundation in Computational design?
A.T.: When during lectures, webinars or whatever, people ask me to suggest some good book to improve computational design skills I always suggest books about mathematics, geometry, and the history of architecture.
My first option is always Architectural Geometry, by Helmut Pottmann. I also suggest collecting all the AD. Architectural Design, not Architectural Digest. It’s important to be very precise on that. I have a lot of them, and each volume is like a photograph of a particular moment in the evolution of digital tools, and advanced architecture, any of them is focused on a specific topic and they are very useful. AD also offers a different perspective on this world. There are issues about digital fabrication, Artificial Intelligence and other interesting topics.
Two other books that I liked so much, maybe not easy to find are, Fluid Totality, and Total Fluidity, two publications by Zaha Hadid Architects. Zaha taught in Vienna, and those volumes gather some of the most interesting projects designed by her students. It’s not a brand new publication, but I feel a lot of energy inside those proposals. They are super original, with different tools and methods, so you can find form-finding strategies, and the best uses of Maya, to apply the physics, within a design process.
G.G: Environmental sustainability has gained enormous space in the market and communication. You’ve been involved with it since the beginning of your career. Can you explain why computational design is necessary for this?
A.T.: You know, ecology nowadays is more a label than a purpose, and that is a bad thing. Why is it a label? Because Architects always faced ecology and sustainability only on the scale of the component: a window, or a façade system, that works without thinking about the whole system. You can create a column, a steel beam, or again a module of a façade, and it works because the process is sustainable. You use recycled materials and think about the lifecycle, but you are not considering the system, just a single part of that. When you create a wrong building with the right components, you will have a bad building. So when we started the ecologic pattern course in Italy, my main idea was triggering in the mind of people this idea of system, many things that work together in a kind of organism.
Parametric design and computational design are about creating organisms rather than systems based on addiction. In general, architects create by adding things one on top of the other, using a metaphor, like the AutoCAD layers: one thing, one layer on top of the other. If you want to overcome this limitation, you need to think in terms of organisms, and so when you create a script, for example, in a tool such as Grasshopper, everything is controlled at the same time: the overall shape, every single component simultaneously. Of course, and we can come back now, on training and teaching, you need to invest a lot of time on how a building works, not just on how to use the latest environmental analysis software. So it’s about culture again, that’s my idea, creating organisms.
There is also another very important topic, talking about ecology and talking about sustainability, trying to create artworks, or architectural design, with the ability to educate people. One of my latest works is named Become. It is an installation for Adidas in Berlin. Our client wanted to create an installation aimed at being a metaphor, a sculpture that communicates this idea of transition. Things that become another one: bottles, then particles, fluid and then filaments, metaphorically used to create clothing, shoes, etc.
Become is a 17 meters high sculpture, we used recycled materials to create the sculptures, and everything was 3dprinted. It may sound like a contraction, but if you go on Google or Pinterest, and you write something like a 3d printed installation, you will get a cascade of images about lattices, grids, and lattices, again and again.
And you know what, the weird thing is when you see a 3D printed object which is made by a repetition of things. I truly go crazy when I see something like that because you can do standard things and put it all together. So I think 3D printing is a technology allowing designers to meet the complexity and richness of digital art.
G.G.: What you are saying makes me think about mass customisation: something virtually possible since the advent of 3D printing, but which has not happened until today. Is it because of the still limited diffusion of this technology or the market?
A.T.: I see three reasons. The first one is, as you mentioned, a technological funnel, or limitation, 3Dprinting will be the solution, but it will take time. Adidas has made one of the first successful attempts with the Futurecraft project.
Using a novel technology, called Carbon 4D, they can print an entire outsole in a few minutes, and so they created a kind of small production, using 3D printing.
You don’t have any kind of customisation tool, so this is just a first iteration, a first wave, where you buy a non-customisable 3Dprinted shoe. The next step will be about collecting data, and that is a huge topic on how to collect data. Thanks to that, you know there are a lot of apps that you can use for training activities, running, jogging, etc.
Everything can be considered fuel for an algorithm able to shape your perfect shoe, or a cloth or other devices, for example.
The second is that according to marketing analysis, people don’t want customisation. In general, the trigger to buying something costly is status.
And so, the status is not about customisation. I mean, we want something that cool people have. Like a watch or a car. They extend our identity, and so we buy things that represent ourselves but that people well accepted. So if I customise my car, I am not sure others will appreciate it. It is easier to buy a Ferrari or a Lamborghini.
The third reason is that companies want to protect their outputs, so they will allow probably, customisation, but within a certain set of limits. When a brand markets a new product, they don’t want the final client to destroy completely their idea. Otherwise, people won’t recognise it as part of the company’s production.
GG: How has the figure of the computational designer changed 15 years after the first Grasshopper release?
A.T.: My idea about the role and identity of the computational designer is the following: I always use a music analogy. Nowadays, in the music system, you have the main star, and you have the music producer. The main star is someone that has a good voice, or some powerful idea of creating a guitar line, or is very good at playing melodies on the piano. But they need a producer or an entire team of producers to turn a melody into a hit that works on YouTube. In every field, from automotive to graphic design, you have people that are, let’s say, pure creative people (even if I don’t like the adjective creative) and then you have someone that supports the development. In architecture and industrial design, this role is that of a computational designer.
Of course, this figure evolved with time. At the first age, 2005-2009, a computational designer was an architect, not a designer. I remember talking about it with Ross Lovegrove, and he said: “Architects have this ability to think in terms of systems, of course, it also depends on architectural schools, but they are more avant-garde compared to industrial design schools, which are more focused on the idea of a product, to solve some specific problems”. So by that time, a computational designer was an architect, with a passion for tools, mathematics, and geometry. Maybe he didn’t have a precise idea of what to do with this kind of knowledge. But over the years this profession found so many confirm out there and now I think we are becoming music producers for the architecture and design field.
GG: When you look at contemporary digital architecture and one of the first digital turn, it’s not always easy to see the differences in terms of computational optimisation. Can you give some examples?
A.T.: The first digital turn was a moment of showing off, as in the early production of Zaha Hadid Architects. This is also true for Gehry, even if his production is not completely aligned with this wave, because he is basically a sculptor with a big office able to realise his vision, but I don’t see an algorithmic idea behind his architecture.
What I see now, from the aesthetic perspective, is that architecture is simplifying. It’s not getting more and more complex as it was at that time. There is still complexity, but you find it in details or small parts. The building, in its final shape, is absolutely simple. You can see just a simple repetition of elements, or a very subtle pattern, but you don’t find anymore this kind of showing off.
It is changing from a fabrication point of view. The world of construction is evolving, even if it is not visible at first sight. If you analyse a building from 2006, you can see several imprecisions. This is not because it was constructed and built in a bad way, but because it was not possible to achieve that level of complexity we can reach now.
Back then, computers had a different velocity. It was easier to design; it was easier to represent, but it was more complicated to build with the same level of perfection. Algorithmic architecture always had a hidden goal, which is seamless continuity: creating a fluid organism without discontinuity. It is also true for aerospace and automotive. Take as an example the latest Range Rover Velar: if you look at it from a distance, you cannot see any kind of gap, because the evolution of fabrication tools, reached some level that is closer to this idea of seamless design towards which architecture is moving.
At the dawn of the first digital turn, this was not possible, as an example, let’s look at Gehry’s dancing building in Prague, where to save fabrication costs, they approximated a free form shape, made using glass, by planarizing each panel with no advanced computational design tool. You can see gaps between panels, air and water can flow through them, technically this is called a not watertight panelization. Now it is possible using simulation processes with a physics engine such as Kangaroo physics. Within the platform, you can basically move each vertex of each panel of your façade until you get the whole building planarized. So you get a set of panels completely planarized, with no gaps between them. Of course, you have a price to pay, which is a kind of approximation of the original geometry.
G.G.: I remember this technique from one of your courses. If I’m not mistaken, it involves the simulation of springs, and it is much more laborious than you can imagine reading the description. This leads me to another question: Some may have the idea that creating computational processes is enough to use a tool, but in reality, it is much more complex. Can you tell us more?
A.T.: It’s not about the tool. If you want to apply a process of geometrical optimisation, of course, you need to have some basics of differential geometry, to understand what curvature means. You need to understand how to split a freeform surface into a set of panels: this kind of process has a theory, you need to study that, and you need to do research. You need to understand how masters of digital architecture solved this kind of problem because this is a set of theories and methodologies that they have, let’s say, 20 years. So you need to have a culture. But first, you need to study how to solve this kind of problem on a piece of paper. If you cannot do it on a piece of paper, you cannot apply this kind of process.
Because the particle spring optimisation logic works with a simulation of physical springs, understanding the physics behind that is important. So you have a complex recipe. First, you have to know how to create a recipe, then need to manage the ingredients, and balance them. This is something that you learn over years of experience, not something you can gain just by reading a manual or the instruction on a box.
The same thing happens with computational design. And you know, when you understand, the science behind that, you can easily also cross platforms, it is no longer about learning one tool, you can easily jump from one tool to another, and you are prepared for the next big thing because everything is evolving at the speed of light, so only if you have a strong foundation, you can also merge different things.
I’ll show you an example. In 2011, a client asked us to create a 3D printed shoe, that is called the N:US shoe. The fashion-designer idea was to create a complex lattice, and so he was thinking just about the aesthetics, not about the structural behaviour. At that time, it was so complicated to create a free-form three-dimensional intricate lattice that we discovered topology optimisation.
Topology Optimisation is a process that allows you to start from a volume. For example, you put some force and constraints, and so the algorithm removes or adds material based on the physical behaviour of the object you are designing. With this project, we used a subtraction logic, using a method that gets rid of redundant material.
When we finally created that shape, it was just because we studied topology optimisation. Similarly, creating your theoretical spine is the only thing that will allow you to face the evolution of tools and their integration: a topic which is getting increasingly important. Even if we use Rhino and Grasshopper as the main platform, the situation forces us to play with different tools, and sometimes create our tools, that’s why it’s always important to have a precise idea of what is behind the script.
GG: As a consultant, what are the obstacles you see most often in implementing computational methods in architectural offices?
A.T.: The usual answer is: “We have always done it in this way”. This is something that often happens in architecture. Industrial design is faster in absorbing waves and trends.
For example, car companies are super interested in computational design tools, because they allow them to speed up the processes, speed-up iterations, and make it easier to evaluate different shapes and design alternatives. This is also true in terms of production because you are faster and in terms of aesthetics since you can create something different from the other companies.
Except for a few offices, architecture is closer to craftsmanship. So it is very difficult to overcome the rigidity of an internal process, and it is true also in terms of construction. The average architectural office has some design leaders that are chosen by a client because of their history, and the line-up of buildings already created, so it’s very difficult to change the method. And in general, you can change a method if a firm can see an opportunity, in terms of business, to speed up the process. I see this very limitation within the architectural world.
An opportunity, but this is probably something about the next years, is about using Machine Learning and AI in general. I believe it will be a game-changing technology for architecture, more than parametric tools. The world of architecture has no proximity to the world of programming. If you look at the dose of AI that is contained in a video game today, we are talking about light years compared to what is happening in architecture. It is a business problem, and it is necessary to create an ecosystem in which a developer, a researcher, is interested in working on the topic.
Artificial intelligence is already present in some architecture software, even within Grasshopper. As an example, I am thinking of the Clusterization tools included in the Lunchbox plugin developed by Proving Ground. We were talking about panelization. Imagine you have a complex facade, with different orientations. Before Lunchbox, to find groups of similar elements, we analysed the vectors considering a plane and based on that, we could create groups. It was a half-day job, but today we can do it in real-time thanks to machine learning.
The problem is that when you propose the development of similar things even to medium-large studios if it does not hold up in terms of financial sustainability, you will not go forward. This is because investment is seen as a cost and not an opportunity. There is a lack of culture and the inability to see where we are going. There is no ecosystem, there is no real incubator of programmers and Information Technology around the world of architecture. The limit is not technological, but cultural. Few large studios are dealing with it, such as Foster and partners, which since 2006 has worked on methods for the development of intelligent layouts. Methods will certainly play a role in the future of architectural design, but they need more attention from architects.
GG: Contemporary computational design owes a lot to personalities like Luigi Moretti, the inventor of parametric architecture, and Sergio Musmeci, the author of the famous bridge over the Basento. Why are they so important?
A.T.: There is something very important about Luigi Moretti. He was a successful professional, not just a researcher. He created not just the first parametric architecture approaches, but also renowned buildings such as Montreal Tour de la Bourse, and the Watergate complex in Washington, D.C. As a professional, he understood that a combination of parameters and the use of calculators would benefit the architectural outcomes, and invested time and money to research and develop his ideas.
Also with Musmeci, there is something similar, even if he is well known for just one important building, the bridge over Basento. He also created a lot of research on freeform structures. But you know the idea behind the work of Musmeci and Moretti as well, is this idea that you can flip the general direction in creating things, where the form is the unknown. This is to me their most important heritage if we want to compress and stress everything in just one phrase. Back then, architecture, design, and engineering were top-down processes. You had a general idea, designed a shape and then just analysed, studied and made a refinement of specific elements and dimensions.
Then they said, No, it’s not like that, we can shape buildings as nature does, so you have to flip the problem, form is the unknown, the parameter are forces, environmental conditions, ergonomic conditions, and everything is shaped by those constrains that will fuel and shape the final object.
Moretti approached this problem as an architect, as with his famous stadia, working on visibility, distances, and the relationship between architecture and the human body. Musmeci was a pure researcher, and he finally came out with a building which is probably one of the very first buildings based on a purely natural approach. I started studying at Musmeci in 2002. I am from Calabria, and Musmeci Bridge is a few kilometres from my house. The first time when I saw it I was a kid and to me, it looked just different, so it was fascinating. It looked like tree branches. I couldn’t find a metaphor at that time. Some years later, I came back to that image as an architectural student, it was still impossible to conceive something similar.
Then, when I started learning computational tools, and when Kangaroo came out, my first idea was: I want to create that shape with computational tools. A process in which I came back many times, improving it. I made three videos on that: the first process and shape were not that accurate, and then I did more structural analysis to improve my process. This takes us back to the first topic. I was very naïve at the beginning, I just tried to use the software. But then I studied, I collected all the books that I found about Musmeci; I went to the archive in Rome; I collected a lot of information and now I can model the building almost as it is.
GG: In 2018, you designed the Oyster chair, the first piece of furniture designed in Virtual Reality, something different from any other computational design process you usually follow. What’re the advantages and limitations of this design approach?
A.T.: We designed in Virtual reality with the Idea of not being contemporary, but primitive. It may sound strange, but when you work in virtual reality, you are experiencing the use of your hands in space, so you start from scratch. Since we didn’t have any design aid as Osnaps (Object Snaps), we created a three-dimensional grid, and we used that to trace the NURBS curves. We created everything using curves and surfaces, starting from scratch using our hands in a virtual world. That’s why the object looks so primitive, so simple because if you are in a pure virtual reality, you cannot be that complex. If you see something created in virtual reality, and someone says “we did it by hand”, that’s not true, it’s not possible right now.
The advantages? A trivial one, you are inside the design environment, and not in front of it. This makes a lot of differences because you can go around your model, you can have a feeling of the final shape, and you can see it on the proper scale. Anytime we usually design an object, we are stuck in front of the computer and we are just using our mouse or devices and it is different, you cannot feel the object you are working on.
During the pandemic we did another object, it was a concept car, and we designed it in a VR environment where one of my colleagues was inside, and I was outside the car. We were talking simultaneously to understand the impact of any kind of modification, both inside and outside. I was looking from the outside, from a different perspective, understanding a design choice, impacting the exterior part of the car: things that you cannot do with traditional CAD software on a screen.
Another thing that you can do is to simulate materials, so you can better understand reflections when you go around, understanding if you created an object and your intention is doing it with no seam. You go around the object and you can understand if it works or not.
The cons are, that it is physically tiring, I don’t know if this is again a trivial one, but I feel it as a tiring process. Any design activity can take a lot. Imagine standing for hours moving your hands. Also, it’s tiring for your brain and your eyes.
You cannot stay inside Virtual Reality for a lot of time. I don’t know if it’s a kind of individual response, but I cannot stay over 1 hour, so, with the current technology, it’s not possible to imagine a design process completely moved and translated into VR, maybe in the future we will have different things.
Also, these VR visors are heavy. They look light when you wear them for two minutes, but after some time of working, it gets really frustrating.
You also need to do some coordination with the joysticks. It’s like playing the guitar, doing something with your right hand and another thing with the left one. You need to get some coordination, and each software has its own rules, as you can imagine. But I am confident that in a few years we will have a better technology that will simplify the work.
