If you just Google AI in architecture, I would say 80% is parametric. The skill required to actually model that or script that is another step, and I don't think it's as momentous as Grasshopper and all these tools. This was, I think, a bigger shift forward. What I see now doesn't lead to a new paradigm. It leads to more productivity and, let's say, more versatility within the paradigm.
And Zaha was most compelling, convincing. I already thought she was the best, but then she came across very well, very genuine and not showing off, but convinced me that the first call would be to try to get into her studio, and it worked out.
It was very much a means to an end for her. She noticed that we can generate things faster, also maybe more beautiful, more consistently beautiful. I mean, early on when we started to script, she realized the superiority of that. So we started with Model Shop and Smalltalk first, then the fold depleted, then they became the soft folds, then blobs came in, and then the next one was these kind of swarms. Generative components was the bigger step, and then was when we started to use a lot of Grasshopper plugins and so on.
That's it, we're live. Welcome guys back to another podcast here on the Architect Network, and today we have a very special podcast for our 60th milestone podcast. Someone that needs very little introduction. We have with us today Patrik Schumacher, the principal of Zaha Hadid Architects. Patrik, thank you so much for joining us today.
Yeah, I've been wanting to have you on for a while. I feel like it's the perfect time. Typically we don't get a lot of the kind of principals from these big firms — we usually talk to the tech nerds, the code team. But in your case, you're such an expert in the area. You've obviously coined so many of the terms and things like that. It's inevitable we finally get you on, so I really appreciate your time today. I thought I'll start with a little bit of a trip down memory lane, because I've seen you talk about this once or twice, but I just wanted to touch on the moment you actually met Zaha and you kind of made this decision that you wanted to work for her.
Well, this has been a while ago. So I was a student at the South Bank. Actually, I came from Stuttgart. I encountered her work in Stuttgart around '84, '85. There were some publications — the GA in Architecture, she had won the Peak competition in '83, was published in '83, etc. There was an AA publication. So I got interested already. I came to London and we were highly aware of Zaha in the studio where I was learning, and so I took note of a conference which was at the Tate, which was scheduled in '88, before the Deconstructivism show in MoMA. They had, for some reason, launched the pre-conference in London rather than in New York, where all the participants were there including the curators like Mark Wigley and Philip Johnson. And Zaha was most compelling and convincing. I already thought she was the best, but then she came across very well, very genuine and not showing off, but genuinely trying to explain what she was up to. So it convinced me that the first call would be to try to get into her studio, and it worked out.
Yeah. And I think I've seen you explain that you were kind of thrown in at the deep end. You started to take on projects quite early, particularly with the Vitra Fire Station. I feel that was the big project.
That's right. The studio was super small. She had an assistant and then there were, I think, three or four other employees, and for competitions there was a kind of group of ex-students who came on evenings and weekends to pitch in for free. So that's the way we were. Vitra came along as a job, and I was the only German speaker and I was anyway in the studio every day until midnight, mostly by myself in the end at the last hour. So I took that on. It took a long time to get somewhere satisfying because it's quite difficult to do something beautiful out of a fire station which had a garage and a few smaller spaces attached to it in this kind of surroundings. We were struggling and had to postpone and frustrate the client multiple times because we weren't ready with anything we liked. He always said, 'Yeah, I like it, let's do it.' No, no, no, we don't like it, we're not happy with it. So it was an interesting struggle.
At that time it was mostly drawing, right? I mean, I had in my student time — so I was still a student at the time, I interrupted my studies to move to Zaha's studio — I had worked with computers a little bit, but the studio didn't have any. And so we were first of all exploring, we were just hand sketching, not even setting up perspective. In terms of the geometry, it was just plotting by hand curves into 3D space with coordinates, and that was also a bit of a challenge. So we had no computers. We acquired them a year or so later to sketch, but not yet to draw properly. That happened another two or three years later.
So one thing I've always wanted to ask you was what was Zaha's outlook on technology? I always thought this was like a component that you brought to the office. But what was her outlook on technology? Was she a bit of a tech evangelist herself, or was it kind of like a means to an end?
It was very much a means to an end for her, and she noticed that we can generate things faster and also, it's more versatile and also maybe more beautiful, more consistently beautiful. Particularly when it came to the use of splines, or also then early on when we started to script, she realized the superiority of that with respect to what she was looking for in architecture — this consistency, fluidity, poise, and complexity with a lot of coherence embedded, which is difficult. You need a lot of talent and confidence to generate that. It comes also out of the rapid hand movements which have an inherent physics to them. So if you're very talented, you can generate a lot of these effects — fading effects and swelling. The kind of dynamics of a fast-moving pen has these continuously changing coherent curvatures, like whiplash kind of geometries. And then we used a lot of French curves from the shipbuilding industry and had specially made sets of curves to translate that. But still, it always required some kind of hand sketching which was then hard-lined, and also the 3D element was hard to get by. So then it was model making and so on, and it relied much more on talent, which was exceptional, than using those processes which were more fail-safe in what we were looking for. And she noticed that, and it was faster too, of course.
Yeah, interesting. I was always interested — so at what point did the technology components start to play more and more of a role? For example, you guys are synonymous with Maya.
I mean, we came across earlier versions of this — Alias|Wavefront — which we saw. We didn't have — there were these Silicon Graphics machines which I think Columbia University purchased in '84. We were there in '93, sorry, '94 — we were there in '93 — and then Harvard in '94 where we used a lot of software and also already splines and curvature software. My first — so we started with Model Shop and Smalltalk, this was more for 3D generating perspectives rapidly. And then something which was later called VectorWorks was something I used a lot, which had beautiful spline tools, and also started to use that for setting up all the drawings but also work on 3D in this form. I mean, it's interesting that the first iterations of folding from, let's say, '93-'94 were all faceted because the NURBS surfaces weren't round yet. A year later they merged and everything shifted more elegantly and more fluidly. I had it at the South Bank in '87-'88, I was using Swivel. We had sporadic explorations, and Columbia University went paperless in '94, one year after we were there. But we had tools available and there was expertise flying around in the school — Greg Lynn was teaching, Jesse was teaching — it was all forming at the same time, what became folding in architecture and the idea. But we weren't immediately able to have these same tools in the office. So we were using them in the schools first. Then in '96, of course, when DRL was founded, we had a lot of these very quick — we came to fully paperless and we were kind of flying, with all this stuff.
Yeah, it's quite interesting that you explain the process. It's almost like the visions were there and the software was catching up to allow you to do that.
Yeah, that's why the software was drawn in. We also had used Xerox machines to distort — one-way sizing or rotating the drawing under the machine to generate some of these blurs or curvature distortions and so on to generate these more fluid compositions. So yeah, it was a craving, and that's what pulled these computer graphics tools into architecture — that desire. Even somebody like Greg Lynn, he said some of these works were done by hand first. Think of Kipnis. I was very keenly observing Kipnis at the AA '91-'92. They did quasi computer graphics all by hand at the precursor to the DRL — designers where they did the Dong competition and then the Haiku master plan — seminal works. And I was immediately drawn to that. Kipnis also, American, he came to the office, he was looking at our work and picking up stuff, and we were in conversation early in the early '90s. When we came to Columbia in '93, we plugged in and observed. Greg Lynn was already a hot ticket there, he had just started teaching the first time as well, coming out of the Eisenman office. We had observed closely what Eisenman was doing with the kind of folding stuff even already in '89-'90. It was an exciting time.
Because what was fascinating, within the year of '93 and into '94 — '93 was also the publication of Folding in Architecture — which was, you know, with Kites on one hand, with an article, then Greg Lynn, and Greg Lynn had published something one year before. But then what I noticed at Columbia, everybody else who wasn't yet on this track — who did quite different things, you know, ColeDonald or Hani Rashid, Ben van Berkel, Alejandro Zaera-Polo — there were many people out there, Jesse Reiser — they had been doing quite different things, some kind of deconstructivism or phenomenology, whatever. And they all shifted instantly, noticing these potentials and seeing some of these compositions. It was absolutely rapid, and within a few months, convergence onto a certain paradigm where everybody was looking at everybody else and picking up the same tools, the same — that was very exciting to see. It was compelling, this whole generation of architects into this direction, and when you had the Yokohama competition, everybody produced versions of this.
In the AA, Kipnis had left, but there was Baram Shidell and Hansel doing stuff. The AA caught on quickly as well, particularly once we had founded DRL. Moyes had come to the AA from Harvard where this had spilled over from Columbia. He brought a lot of people. The AA was too big, there was always other things going on. Columbia went nearly 100% down this path. Harvard had also other things going on. But the AA — we were kind of 60-70% in this direction. It was an exciting time. And then we also quite quickly realized that we have to go through all the building typologies. We were the ones who went into urbanism more strongly and tackling issues like construction and taking our own technical issues and questions and reinventing, completing the typologies. Zaha was very excited about participating in that and investing more and more — machines coming in, and started hiring from DRL, from Columbia, from Harvard — those people where we had developed the skills and concepts.
Yeah. So I guess it was a moment of computer-aided design — like I don't want to say CAD because you think of —
No, but CAD was an element of that in a sense that we realized — when the competitions we had done, we had to make drawings by hand. They had to be big, and we had ten people trying to draw all the plans of a competition, lying over, making, you know, hurting their backs and scratching. This is what was still going on. And then so once we had this kind of precursor VectorWorks and I got myself onto competition, I could substitute for all of these, doing all the plans as layers of one set of drawings, and do a 3D on the side. The productivity gain was massive, but also I thought some of — you just had more chance to adapt and readapt and evolve the drawing because it's just too slow and painful to change drawings. And also they weren't consistent. When you have to build a model it's going to be difficult if the drawings are not matching up. It was a massive productivity and quality uplift which we experienced with what we were trying to do — these more complex, adaptive, more difficult compositions in spaces. I think — so I don't know about other firms, of course it was also faster to do more normal kind of minimalist drawings, but we weren't yet anywhere near constructing anything. So this was tool sets of being more proficient in competition output.
Yeah, there's elements — I'll close the loop on this — but there's elements of what you're describing, it feels like the era we're in today with AI and stuff, where it's a huge productivity gain, you know.
That's right. Yeah, I think exactly that. It sounds a lot like that.
And so from there, fast forward, you're then kind of in the parametric era. You, of course, coined the term parametricism, which is sort of the era I left uni and came into this world. I think the first tool I ever used was called GenerativeComponents. It was a very important tool and inspired this whole idea of parametric urbanism — that you can populate a landscape surface with components which were buildings. And that was very important. I mean, there were a number of these shifts — the first was the fold depleted, then they became the soft folds, those single-surface projects, then blobs came in and you had more closed volumes, and then the next one was these kind of swarms — swarm formations — and GenerativeComponents was important but also these flocking algorithms and so on. And the bigger step was when we started to use what you then called a shift in tectonism — with the world of Grasshopper plugins and so on, with mesh relaxation and form-finding, and the Rhino world. This whole world was another step, and the investment in fabrication — the first instances were the excitement with milling machines, and again Greg Lynn was there, but then it became — as you know, schools of architecture invested heavily in developing fabrication sensibilities and feeding that back into design generation. That came after the GenerativeComponents thing and parametric models became very important. And when I was at university I was looking up to firms using this, and I think at the time Zaha, Foster's Specialist Modeling Group were big players, and a firm I really looked up to was Future Systems, trying to connect this world of digital with fabrication and boat building with the Lord's Media Center. And then this explosion of Rhino, Grasshopper, and parametricism was born, and we had tectonism and then parametricism 2.0. What are your thoughts on parametricism looking back now? Because on one hand it's a style, and it was extremely successful as a style.
I think it's — it's kind of interesting to reflect that when I first mentioned it, and I thought about I need to find a term for this — there was a Smart Geometry conference in 2007, and I tried it out in a lecture and it was a good response. I then had the Banus lecture, we wrote the Parametricism Manifesto, and I had a session of the Dark Side Club which was inviting a number of young architects, Maya amongst them and a few others, to present under the headline of parametricism. It's interesting retrospect — I didn't notice at the time yet that this was also the period where this kind of huge convergence and movement forward soon came to an end. The financial crisis happened in that year and then the European debt crisis the next year, and there was a lot of these opportunities for young architects all over Europe. We had projects in every city in Spain nearly, and plenty of amazing opportunities in Dubai. Spain, for some reason — retrospect, I was wondering when it stopped to develop. There was the shift in the schools, partly Occupy Wall Street and anti-capitalism, and also maybe lack of opportunity, and this kind of shift of sensibility. The moment of announcing parametricism, it was already doomed, in a sense, of spreading further. I mean, for me it was clear this is it. It was the way I saw it spread through the '90s and into the early 2000s and take over the AA, all the main elite universities, and then bring it to Chicago, etc. That seemed to be unstoppable. And then it didn't continue to spread. Of course, the people who were working and who were committed, we all kept developing and deepening it — like at ETH, there was a core group. A core group of people were just keeping progressing, but there weren't new generations flowing in, or maybe only our students which we still can draw in and excite. But it wasn't spreading any further.
I think that was to do with the times — the political times, the economic times — and this kind of vogue takeover of the institutions accelerated. I think now, glimpses out of this — if you look at the Bauhaus but also the whole AI thing and what people use it for, as well as the metaverse as a frontier — where I see mostly parametricist approaches. So I've recently been thinking that parametricism will now finally happen more broadly.
Do you think there's a component where, in that era of parametricism, people kind of got their hands on Grasshopper and these other tools, they suddenly had the ability to create really expressive complex forms, right? And I felt like the reason there was this boom and maybe bust of them actually getting built was because there were only a few people that had the knowledge to actually build them. Do you think there was this disconnect between the architects coming up with this but not necessarily having the knowledge of transforming that into actual constructible geometry?
Yes, I think there's an element of this. There was an over-exuberance — there were excesses which could have been compromising, giving, putting to disrepute some kind of fundamentally very sound and highly rational movement. The other thing which I always thought, when it came — the schools and avant-garde movement around parametricism was highly technologically oriented, Smart Geometry and so on, which was positive, that was good. But what was lacking, and was hard — there was no dialogue with these, on the sociological front, on the socio-economic front, on those people who are looking for societal meaning. We were looked at as nerds and CAD monkeys. The AA, for instance, the atmosphere was — there's a presumption, imagine, that the DRL doesn't have theory. I mean, this is an absurd presumption. But that is another element why in the academic world there were other themes and topics which the protagonists of parametricism were just not engaging with. And so I saw my role to engage more with this aspect. I see historically and socio-economically the development of post-Fordism, and some of that was understood on the technological and production side — non-standard, flexible specialization, customization, mass customization — all these phrases were understood. But that this has interesting and actually liberating and advancing sociological, societal consequences — with a knowledge economy, it not only transforms the solution space through all the tech means we have but also the problem space — urban convergence, complexity, mixed-use, integration, etc. These are the elements I was focusing on, writing about and thinking about. There wasn't enough of that, which is partly why in the academic world we've lost some ground as a movement. But also there was on the one hand too much technology focus but then there was also not enough depth of skill in many of the early protagonists.
But I think now — what I find exciting, as I said, if you just Google AI in architecture, I would say 80% is parametric — trying to generate. So there's a desire there, of course. The skill required to actually model or script that is another step. But it shows where a certain generation of architects want to go, and I find that very optimistic.
Yeah, I think that's an interesting point. It's undeniable that with AI, people very quickly start to gravitate towards parametric fluid — there is an interest in that. And I think the public would love to see more buildings like that. That's also part of why you guys are so famous in the general public, right?
That's right, very super popular. I mean, there's no other major firm coming anywhere near in terms of popularity. And the other thing is China. China is still — although they had Xi and some traditionalists criticizing some of the innovative and unusual forms, this is still going on. And recently somebody said parametricism is a kind of national style of China, but against the kind of desire of the Communist Party. It's a groundswell of some kind. So unfortunately now China is slowing down temporarily in terms of this development push. But I feel that as China slows, there's a rise of the Middle East, and I think there's a lot of talent — we've been teaching Indians and Chinese for the last 30 years, a lot of them also going back and doing work, and that's another wave. So I think it will come, and as these technologies diffuse and as contractors become more geared up to handle this complexity logistically and through their tooling, eventually I think it will spread more. Because in the end, particularly with tectonism, you have a lot of potential material and then economic gains as well — projects where you can use shell action, or tensor forms, or more sophisticated structures, or more sophisticated shading systems — you gain things. In the end, things are lighter and cheaper. That's the outlook. There will also be better utilization of complex sites, better utilization of spaces if you have something more closely adapted to different parts of the site and different parts of the program rather than using too much repetition and symmetry. I think eventually this will come through — not only surcharges which is maybe still at the moment the case, but will turn the opposite way — efficiency gains. That's something I consider very important.
But I also emphasize — it's not only desire, it enhances the legibility of the built environment. If you use convex-concave, if you use curvature, you can bring parts of a building together rather than allowing composition to fall apart. So you can have more difference at play which distinguishes situations and adapts functionally, but also recognizably different to identify sub-parts, but also express how they belong together and function together. There's a legibility enhancement for the built environment coming out of that. We can certainly see that when we do master plans — that we can generate a more unified geometric and morphological character, identity, and image for an urban district, yet have it differentiated. Compared to that, a lot of times you either have monotony or a random spill and you don't generate character and you don't generate global form except very simplistically with a symmetry axis or something. These advantages need to be talked about and explained, and when we have a chance to explain and present master plans, we can win them. There's sometimes also an intuitive appeal — 'oh, this is my area' — they can identify it, they see a certain hierarchy expressed. So I think with buildings there are social functionality advantages. I'm very much into orientation, navigation, legibility, interface, while maintaining complexity and variety. These kind of intuitive advantages just come through — that's why we win.
Win competitions as well. I think it's the other part of course is still let's say something which will wear off, novelty aspects or what people kindly call the iconic. It's not something foregrounded but comes out of a very great process. But these buildings still look other and unusual, and I mean that helps sometimes in some context; it doesn't help in other contexts. But I don't think it reduces to this. So it's good, it helps us sometimes to get work. But I think there's a deeper rationality, and this would be that when this novelty aspect wears off through this being a more universally applied set of paradigms, you can still maintain a lot of diversity because the repertoire or this paradigm allows for much more internal diversity than other approaches. Yeah. And the novelty element might wear out, but at that time we have a more consistent, cohesive environment.
Yeah. And how important has the CODE team been in Zaha? So I was trying to find the exact time you guys founded CODE. I get the sense it was around 2007, 2008 maybe.
Yeah, that's probably right. Well, it's very important on a number of levels. So first of all, as a service to a lot of the major complex geometry projects, they're doing a lot of work, taking a lot of optimization and workflow optimization and controlling. And it's not only the CODE team; we have also diffused experts. We're not tied into the CODE network. But CODE is also of course doing original research and developing the workflows, developing the tools, developing the algorithms and scripts. So they go in and apply them to many projects. So that's one aspect. But it also connects, and you know, with more experimental projects like this 3D printed bridge, this kind of collaboration with something like the Block Research Group, and small pavilions like the one in Mexico City, the computational knitting formwork, or this kind of compression-only concrete printed masonry if you like. There's been a number of these researches and pavilions. They show the way, and some of them don't have yet full application, but they connect up with some studio research at the Venice universities and this whole folding things. We use them in experimental studies, and then we have small structures, maybe some furnishing things, pavilions, and then we scale up to larger projects. Combination of, you know, I mean we had for instance the Serpentine Pavilion, and then we do the proper permanent Serpentine North Gallery. From there we went on to, for instance, the Tokyo Stadium using similar ideas of shell and tensile structures. So it's been very important to me, and it's actually something which is still not fully diffused within the office. The use of such tools and the sensibility of let's say tectonism over and above what we had in previous versions of Parametricism is still a rolling out process.
Yeah. Um, because skill levels and also sensibilities, and we have to keep pushing and disseminating internally.
It of course always lends itself more to some projects rather than others. I think every project could and should have more of that, and that is why we're going to build CODE up, increase the numbers of people. But CODE is also taking on new, and we have different research groups now. We have nearly 10% of researchers who are doing partly research, partly I mean research group taking on new topics like what we had now at the BNA, the participatory urban development platform, this idea using meta and components and choice processes to develop German ensembles is one aspect. I have the kind of agent-based, this is not CODE but another research team, occupancy simulations, and we're interested in corporate planning. So we have a whole suite of tools for corporate space planning as well as generative furnishing, machine learning to have optimized furnishings. There's a number of research groups. We have the environmental engineering front-end team. So CODE is part of an overall technology cluster which we have. It's the biggest, the most kind of diverse part of it, but the technology cluster is now, you know, it's exactly 10% of our staff, the kind of umbrella. And we're also doing new workflow things. We're investing in Omniverse, but as you know, and we have founded a metavise company, so we're heavily using Unreal. And we have a sub-team of CODE doing AI tool development. And in every project now we're training up these, you know, LoRA networks, and we're exploring already with Sora, with video generation. And we have done this whole thing of training, you know, we're teaching tectonism to AI, so we can build up some of these tectonic articulation inputs, for instance, generally discarded and translated all into curved folding, or learning topology optimization patterns. So the AI research team again is, so CODE although it's large, we don't have enough to disseminate, you know, all of these elements to everybody.
