pulp #46: algorithms
The date is
Friday, March 30, 2018
BRIEF NEWS. And then an exceptional interview, with DAVID REEVES of ZHA.
BIOPHILIA RISING. After VIVIAN LOFTNESS talked about the importance of designing buildings that bring the outdoors in - even in very cold, harsh climates, say, Chicago (pulp #45) - SOM announced plans for a classic black steel and glass Miesian SOM tower... with, just at the height of its neighbors, a 5 story open deck in Fulton, Chicago. Project name? "The Porch." Correlation? Causation? Who can tell. ArchD
MONOPOLY HOUSES. They just keep coming: 3/22 3/25 3/30
CIRCLES. JUSTIN MAROZZI writes Baghdad's relatively recent origins - in 762 - as the capital of al-Mansur then the Abbasid Caliph. It was a perfect, walled, circle, one kilometer in radius. Today's Green zone (still walled, geometrically more of a rectangle) is just to the south. Guardian
POSITIONING. In Common Edge, DUO DICKINSON offers a thoughtful piece on the home, "at once infinitely personal and culturally pervasive. They can be as simple as a glass of water, or as complex as an eight-course meal." CommonEdge. And then MARIANELA D'APRILE writes about #MeToo: "If you’re a man in architecture and nervous about how to get involved in this movement, just show up in earnest. Bear public witness to abuse. Don’t turn a blind eye." CommonEdge
CONTRA-COLLAGE. MARIO CARPO continues Metropolis Magazine's coverage of the post modern (po-mo) revival. Let's do this: THOMAS DE MONCHAUX (Metro), in his February book review of po-mo books, made a list: Po-Mo, Po-Po-Mo, No-Mo, and Mo-Po (More Postmodernism). To this list, Mario Carpo adds PoDig - Post Digital:
The only difference between yesterday’s Postmodernists and today’s Post-digitalists would then be in the degree of their aversion to technology: the PoMos were violently against all modern, mechanical technologies of mass production; the PoDigs seem to have adopted a strategy of technological nonchalance (even Photoshop is OK, as mentioned). The PoMos fought against technology; the PoDigs don’t care about technology.
Of these PoDigs, Carpo does not approve:
If we don’t care about technology today, it is not because there is no technology out there, but because there is too much of it; it’s not because we are bored, it’s because we are quitters. And, as always, if architects stop caring about technology, someone else will in their stead.
Find now, below, an interview with an architect who decidedly does care about technology, David Reeves of Zaha Hadid architects:
FEATURE: INTERVIEW WITH DAVID REEVES
ALGORITHMS, & GAUDI
David Reeves is a designer, programmer, and aspiring hermit. He is currently a member of the Computation and Design (CODE) Group at Zaha Hadid Architects where he leads the office’s research efforts in geometry processing and digital form finding. He also currently co-runs Research Cluster 3 (RC3) within the MArch Architectural Design (AD) programme at the Bartlett School of Architecture.
Links: Spatial Slur / ZHA Education / Videos (Vimeo) / Downloads (Github)
DR: David Reeves
NK: Nicolas Kemper
DR We only have 45 minutes, because security is going to kick me out
NK Amazing. They kick the workers out at ZHA [Zaha Hadid Architects]?
DR No, the Architectural Association. At ZHA they never kick out the workers – it is a 24-hour fun factory.
NK That was my suspicion. Can you introduce yourself?
DR My name is David Reeves, I work at ZHA architects as a computational design specialist. Which I guess just means that I like to design things with algorithms, and writing code.
NK So I have a theory about the flowing organic forms in ZHA’s work: there are two computer programs that produce them, Grasshopper and Maya, each embodying a different philosophy. If you ask, say, Patrik Schumacher, he will say the forms are parametric: they come from having combined 40 factors with the right algorithm in Grasshopper. If you asked Zaha or her students, they would say the forms come from artistic intuition, sculpted and molded in Maya.
DR That is fair to say. Though the tools are modern, the approach to ideation is pretty traditional: the way design space is explored is largely intuitive. Now Patrik would say that intuition is the result of someone considering many constraints simultaneously based on a history of prior experience. But ultimately when you work with some sort of polygon modeling software like Maya, it boils down to moving vertices around - a very sculptural approach.
NK As opposed to Grasshopper.
DR With Grasshopper, the process tends to become more systematic. In general, the Computation and Design (CODE) group – of which I am a member – advocates working more systematically during early stage design exploration, but the time investment associated with developing design tools makes this a challenge. I personally feel that the development of design tools are a fundamental part of the design process: As a designer, you should be developing tools that help you explore the space of possible outcomes in a way that is fundamentally relevant to the project at hand. Because of the time investment involved in this, however, more often than not we fall back on a more familiar, sculptural approach.
NK So what precisely does the non-sculptural approach look like?
DR 95% of it looks like writing code. Sometimes, this makes it a difficult approach to justify internally, because progress isn't always apparent to a room full of architects and designers. The main challenge is capturing design intent and translating it into an algorithm. You are going from something that is typically ambiguous and volatile – design intentions start change wildly sometimes from hour to hour – to an algorithm, which is an incredibly precise set of instructions that can be executed by a computer. It can be tricky.
DR So it varies from project to project. Sometimes there is a lot more space for that, to develop more substantial tools and methods, and sometimes – just because of time constraints, and pressure on the team – we fall back on traditional design methods.
NK That does seem like one advantage of developing tools – you can keep developing them across multiple projects.
DR Yes, exactly.
NK What is the ideal outcome for the design process? What would you like it to look like?
DR I think it is less what would you like it to look like, and more what does it allow you to explore. Across the whole group, every time we set out on a software development exercise, the goal is to build some sort of model that allows us to explore design space in a more intelligent way. So rather than say we are manipulating the model, by virtue of the way we build the model, it steers us away from exploring areas of design space that don't make sense. In a way it allows us to more quickly converge on a "higher quality" design output - it augments our ability to explore design space.
NK Step back. How did you get here? What is your story?
DR It was largely reactionary to my academic experience. I went to Carleton University in Ottawa for undergrad. During this time the school had a pretty clear dogma rooted in a phenomenology and representation. There was a lot of focus on drawing as a form of representation, rather than drawing as a form of design exploration.
NK What does that mean?
DR Well, the first design studio we took was actually a drawing class. For one of the first exercises, we were given a small section of cast in place concrete wall and told to draw it in absolute detail. We were basically told if it didn't take us more than 80 hours to finish, we'd fail. So everyone in the studio was just perched in front of their patch of concrete wall, for a week. We all thought we did a great job, submitted, and everyone received 0s, and were told we could not draw. Because we could not draw photo-realistic cast in place concrete.
NK That is one hell of an exercise.
DR So after wrapping up 3 more years of that I felt obligated to figure out what computers did. Because I felt immediately dated, as a 21, 22 year old who just finished university, I felt obsolete from the get-go. Computers seemed like magic. I was thrilled to learn about them. I taught myself how to program, and write little plug-ins for things, and gradually moved to more substantial programs.
But I often wonder if I would feel the same way if I did not come through it through this really weird, reactionary experience.
NK How would you contrast writing code against writing text or prose?
DR With code, you finish writing, and then the thing you just spent a long time writing does something. Very immediate. It is a tool that can be instrumentalized in a very tangible way.
I now have a particularly painful relationship to writing texts. It takes a very long time to write, just standard text. And I think it has gotten in some ways worse since I have spent time writing or programming. My brain tends to nit pick at English the same way it would nit pick at code, when really the same rules do not apply: there is all this room for ambiguity in English, that you cannot afford in writing a computer program. You cannot be in the same headspace to write both.
NK Would you say the contrast between prose and code is similar to the contrast between the systemic and intuitive approach to design?
DR Yeah. There is a definite parallel happening there but I haven't been able to parse it in a meaningful way.
NK What are you teaching?
DR I am running a masters studio with Tyson Hosmer, who I work with at ZHA, and Octavian Gheorghiu who works at Foster and Partners. The three of us are doing sort of part-time studio, running the show split three ways, because we all work full time.
NK And what is the studio about?
DR I am really trying to get students more engaged in the technical aspects of computational design. Many students come here to learn about computational design, but, developing the required depth of technical knowledge can be difficult to fit into a 12 month design program. More often than it should, this results in students skimming through the internet looking for "cool algorithms" that they can immediately apply to their project without really understanding what they do or how they work.
So we spend a lot of time trying to build up these skills in a hands-on way: lessons, tutorials, mini-hackathons – either go through language features, or try to implement a particular model and dive deep on some of the technical difficulties. And students are rising to the challenge. They are coding on their own, making games.
NK When you say a game, what precisely do you mean?
DR We're pretty loose with our definition of a game. For me, a game would be an interactive simulation. You are simulating something that you can manipulate, influence, or impact in some way as it happens. So it is maybe less of a game in the traditional sense and more of a model that allows us to play with simulation.
NK When you look to architectural history, who would you say are your fore-runners?
DR I think the obvious choice for anyone who works computationally would be someone like Frei Otto or Gaudi – just because of the way they positioned the model within their design process. It was something they used to explore design space. The design model was something that augmented their ability to conceive and ideate. Whereas more often than not a design model is a record of decisions that have already been made. So when you are designing with modelling software like Maya or Rhino, the model is recording a series of decisions, or actions you've made.
For me, a design decision would be any action that changes the state of the design model. Most design models don't change their own state - they require design decisions to be made externally
For instance, Gaudi’s catenary models: you pin one string at one end, you pin it on the other end, and it organizes itself into a shape that happens to be very productive structurally. But you are basically working with a model that is able to change its own state - it makes design decisions about how it should be configured autonomously.
NK What do you call this kind of model? Interactive?
DR Most other disciplines would just call it a model. That is what, say, engineers would say a model is. When you say ‘I am modeling something,’ they say ‘What are you modeling? What is the phenomenon, or process, or thing that you are simulating and trying to understand? That is what models are usually used for. ‘This problem is too complex to approach analytically, so I will model its behavior over time and get an approximate a solution.
NK So if you cannot solve something analytically, you solve it through modeling
DR Yeah, that's more or less correct. To solve a problem analytically, it must have a closed-form solution i.e. one that returns an exact answer in a finite number of operations. For example, the points of intersection between 2 circles can be solved analytically with some simple trigonometry.
Unfortunately, the large majority of problems don't have nice neat closed-form solutions so they must be approximated numerically. This typically involves creating a model that attempts to solve the problem by gradually moving closer and closer to the exact solution. While it might never get there, it ideally gets close enough that you can't tell the difference. At this point, the model is said to have "converged".
Design professionals are unique in their use of the word model.
NK You mean, for design professionals, the model is representational
DR Exactly – more of an inert thing that captures their decisions, whereas in every other practice it is an active thing that presents you with solutions.
DR And that is exactly what you would use a model for. That is the reason you would model something in other disciplines: because you cannot really figure something out, so you say, ‘I have to model this, to understand how this works.’
NK Whereas in architecture it is representation.
NK Yours is a long-standing fight to claw back territory from representation in architecture.
DR Yeah. I would say that is a fair summary. A very well put summary, given the experiences I have had – why are we drawing to represent? Why can we not use drawing as a means to explore or generate?
NK Can you talk about nature?
DR Not with a huge amount of authority. But some of the models I like to build are based on natural models. I find that some of the models people develop to explain how, say, ants build extremely sophisticated architectures – some processes are quite potent, when adopted as tools that we can use to order, or structure patterns. Tools to create order. With ants, you have a phenomenon that is inherently well suited to be represented algorithmically: simple-minded creatures able to produce extremely sophisticated structures.
As a model, or as a path to generating order, nature has some pretty fascinating precedents.
NK Are there paths to order outside of nature?
DR The one we are more familiar with is the top-down approach, which is essentially what the job of an architect or artist or designer is: a central decision maker, and I will decide how to order this – whether it is a master plan, or a building, or I don’t know – shoe. You are imposing some sort of order on that thing.
For a long time people looked at social insects like ants and bees, and assumed that they their impressive structures came from centralized decision making - one of them is in charge of giving orders to the others. But computational models played a large role in recognizing and understanding the more decentralized processes at work. This insight would've been difficult to have without being able to model these processes. You have to model the interactions between populations of simple individuals to see that they can generate surprisingly sophisticated structures and higher level patterns.
NK Hm. People who say they are looking to nature in the context of architecture do it because they are, for instance, attracted to the more graceful lines, the proportions, the biophilic aspects that put the bodies more at ease, and you are saying that it is actually about the decision making process.
NK Top down versus bottom up.
DR Yeah, exactly. So the times I refer to nature, when I am most interested in it, is when it provides a precedent for how I can go about developing the individual components in the models I am making - how I can go about programming them to make their own decisions. How to program a model to change its own state over time, in a decentralized way: each component of the model updates itself according to other components in the model. And it turns out that is pretty much exactly – to the best of human understanding – how a lot of highly complex patterns form themselves in nature.
NK Do you see there being a political aspect to horizontal versus vertical design, where the parts make the decision, as opposed to there being one decider – were we once more of a horizontal system, as opposed to a vertical one?
DR I cannot speak to political self-organization, but in design, any vernacular city you see is a demonstrator that a top down authority is not always required. You can get stuff done without it. It is just a matter of establishing a productive set of rules of interaction.
NK Does nature still inherently then produce the graceful geometries, curves, lines and proportions that people find so attractive in nature? Or is that just a coincidence?
DR It's a nice thought but nature can do some pretty gnarly things as well. And often the gnarly things are just as performant as the beautiful – they are satisfying a number of criteria really well. They are not beautiful in the sense of an organic aesthetic. There are just as many examples of nature being successful and really messy, as nature being successful and beautiful –
NK What is an example?
DR Slime mold comes to mind immediately. So there's a particular species of slime mold that looks for food by fanning out over its environment. Once it finds a food source, it reconfigures into remarkably efficient network structures to better distribute nutrients throughout its blob-like body. But it is not like – I do not think many people would look at this process and describe it as a beautiful thing. Or at least not in the same sense that they'd say Zaha's work has an "organic beauty" to it. This mold is pretty nasty looking the whole time. But it is accomplishing some pretty magical feats of engineering and design which is more of a performative beauty than an aesthetic one.
NK You have modeled slime mold?
DR Yeah. That was one of the natural precedents I began looking at more closely. As a model, though, it is hard to make it do certain things. It is difficult to tame. Which kind of brings up this principle of ‘directability’. How much can I control it, or steer it in certain ways? It seems like a paradoxical property to want from a model like this - because on the one hand you want the model to be autonomous and present you with novel solutions that you might not have considered but on the other hand you need it to do certain things.
The more applicable versions of these models are the ones that do facilitate some degree of directability. You can tame them when you need to. But when you are freer to explore other possibilities, you let it run loose.
NK You said at the beginning that computers seem like magic: do they still seem that way, now that you understand how they work?
DR Yeah! Or the closest thing to magic that humanity has. You have this machine, you told it exactly what to do, but it is presenting you with things you could never have imagined.
Copyright © 2018 pulp,
All rights reserved.
pulp is an architectural circular.
Our Comics Editor is Sarah Kasper.
You can see all of our drawings at:
Our mailing address is:
282 W. 115 St, 10026
New York, New York
TODAY’S COMIC IS BY