PRINT March 2020


General Motors assembly line, Youngstown, Ohio, ca. 1972. Photo: Library of Congress/Getty Images.

ELECTRONIC COMPUTERS have been around since the end of World War II, yet for a number of reasons architects and designers did not start seriously tinkering with computer-aided design until the early 1990s. In doing so, the pioneers of architecture’s digital turn quickly stumbled on a groundbreaking realization: The computers they used to draw objects on the screen could also aid in fabricating those same objects right away. The integration of computer-aided drawing and computer-aided manufacturing (CAD-CAM) became a tenet of digital design, and the machinery for numerically controlled fabrication has been as influential and even inspirational for designers as the software at the heart of the computer systems themselves.

General Motors assembly line, Biel, Switzerland. 1946. Photo: Grisel/Lindroos/RDB/ullstein bild via Getty Images.

In the 1990s, computer numerical control (CNC) milling was designers’ tool of choice—a subtractive fabrication technology in which a drill moving along a continuous path carves seamless grooves into a solid volume. CNC machines were a perfect match for the new digital-streamlining (or spline modeling) programs that were then coming to market, and this felicitous pairing of software and manufacturing tool accounted for the rise in popularity of the smooth, curvy lines and surfaces that marked end-of-the-
millennium global architecture—a style inaugurated by Greg Lynn’s Blobs in the mid-’90s and, following Patrik Schumacher’s influential writings, now often called parametricism.

Gramazio Kohler Research, Gantenbein Vineyard Façade, 2006, Fläsch, Switzerland. Photo: Ralph Feiner.

A few years into the 2000s, however, the rise of 3-D printing brought about a significant change of tack in computational theory and design practice: Unlike subtractive CNC milling machines, 3-D printers—both cheap consumer and costly industrial-grade versions—mostly work by materializing tiny units of printed matter called voxels; hence the 3-D printing of even small and simple objects involves the notation, calculation, and fabrication of vast numbers of minuscule and identical boxlike units. At the same time, various “discrete” tools for electronic calculation or simulation, such as finite element analysis, cellular automata, and agent-based systems, became popular in the design community. These tools are designed to handle extraordinary amounts of unsorted and often meaningless data (now often called big data); they afforded engineers and designers the unprecedented ability to deal with discrete voxels one by one, and to eschew traditional (continuous) mathematical functions. Not surprisingly, signs of this new and increasingly pervasive logic soon began to show up in the work of technically driven experimental designers: Voxels, for example, were often left visible, sometimes ostentatiously, in numbers far exceeding the powers of human calculation, and at resolutions deliberately challenging the thresholds of human perception.

Gramazio Kohler Research, robotic fabrication of an element for Gantenbein Vineyard Façade in Fläsch, Switzerland. Zurich, July 17, 2006.

These early tools for computational fabrication—CNC and 3-D printing—shared two core aspects. First, they did not require any mechanical casts, molds, or matrices to reproduce identical copies. Mechanical matrices have an up-front cost that must be amortized through repeated use; the savings obtained through mass production are called economies of scale. This does not apply to digital fabrication: Thanks to the absence of mechanical matrices, every digital replica is, in a sense, a new original; therefore, making more copies of the same item does not make them cheaper. This is the technical logic of digital mass-customization—one of the most disruptive ideas ever conceived by the design community. Second, and mitigating to some extent the import of the above, digital fabrication, though well suited for scalability in numbers, appeared for a long time to be eminently nonscalable in size. Milling machines can mill any number of panels in a sequence, all the same or all different, but each machine must be bigger than the panels it mills. Most 3-D printing, even when based on extruded filaments, happens inside a printing chamber; to print a house in one go, a 3-D printer would have to be bigger than the house. Attempts at building giant 3-D printers having proved impractical, it was long assumed that digital fabrication would be destined primarily for the production of small nonstandard components (facade panels, for example) that would be put together later more or less by hand, in a process so labor-intensive and costly that digital-fabrication technologies were often dismissed as best suited for making teapots, chairs, and experimental architectural pavilions (teapots and chairs requiring little to no assembly, pavilions being mostly assembled by hand by unpaid architectural students). This is when architects realized that the ideal tool for the automatic, computer-driven assembly of any number of parts had been in existence for half a century, hiding in plain sight—or, rather, on the factory floor.

Gramazio Kohler Research, robotic construction of Pike Loop, New York, October 2009.

ENTER THE ROBOT. In the early 2000s, the industrial robot was widely seen as a mature, unexciting technology; industrial robotic arms had been known since the early ’60s, and they became commonplace in the ’70s, mass-produced by a number of American, European, and Japanese companies and adopted in particular by car manufacturers to replace manual workers on assembly lines. Frederick Taylor, the Progressive Era inventor of scientific management, famously saw the modern industrial worker as a gorilla, with the intelligence of an ox.1 In the Taylorist tradition industrial workers were assumed to be too stupid to learn more than a limited number of simple motions; in modern moving assembly lines the worker was expected to stand still, learn one gesture, and repeat it forever. When industrial robots were ushered in to replace human workers (mostly due to the rising costs of labor in the ’60s and ’70s), they inherited the same intrinsic stolidity. Industrial robots need a scripted program for each motion they make. When these programs had to be individually written or otherwise prepared by humans, it made sense to have each robot repeat the same motion, i.e., the same scripted program, ad infinitum, just as the human automaton it was replacing had been compelled to do since the invention of the assembly line. All the “intelligence” these early industrial robots needed was the memory of a few sequential moves, recorded on magnetic tape. 

Gramazio Kohler Research, robotic fabrication of a trussed beam for The Sequential Roof, Zurich, March 25, 2015.

As it happens, the best computer scientists of the time would have been hard-pressed to offer much more than that. The memory and processing power of even the biggest mainframe computers of the ’60s was a tiny fraction of what we now have in any cell phone; unsurprisingly, most artificial-intelligence projects conceived in the ’60s proved undeliverable and were abandoned in the ’70s. Cyberneticians of the ’60s, particularly in academia, had high hopes for the imminent development of intelligent industrial robots, but factory owners did not need intelligent robots to replace unintelligent workers—and there would have been no artificial intelligence to power those robots anyway.

Robots could carry out a theoretically unlimited number of different motions—without any supplemental cost, or at the same cost per motion.

Fast-forward to 2005, when Fabio Gramazio and Matthias Kohler, young architects then in their mid-thirties, began their seminal experiments with industrial robots in the department of architecture of the Federal Institute of Technology (ETH) in Zurich. At the time, personal computers could easily script all the instructions needed to drive the motions of a robotic arm. This was often done by a kind of reverse engineering: The final positions of the robotic hand would be set; then the computer would calculate all the motions needed to get there. Gramazio and Kohler soon realized that computers could also be tasked with writing sets of instructions for sequences of incrementally different robotic operations. For example, a robot programmed to pick up identical bricks from a given location could be instructed to lay each brick not in the exact location where the preceding brick had been laid, but next to or above it. Robots could also be told to lay each brick in a given course at a different angle from the horizontal alignment of the wall, or of the brick next to it, and so forth. Driven by a theoretically unlimited number of different scripts automatically calculated by a computer program, robots could carry out a theoretically unlimited number of different motions—without any supplemental cost, or at the same cost per motion. Gramazio and Kohler famously demonstrated the architectural possibilities of this new technical logic by designing and building a number of geometrically complex brick walls: The complexity of their geometry was, by design, beyond the reach of what most skilled bricklayers could manage—assuming that any such artisans still existed. More recently, Gramazio and Kohler applied the same logic of differential, or nonstandard, assembly to stocks composed of different components, each individually designed and custom-built: The roof of the Arch_Tec_Lab at ETH in Zurich, completed in 2016, included almost fifty thousand different timber elements, automatically assembled in 168 different lattice trusses, each with a width of nearly fifty feet, covering a surface of more than 24,000 square feet.

Gramazio Kohler Research, The Sequential Roof, 2016, Arch_Tech_Lab, Federal Institute of Technology (ETH), Institute of Technology in Architecture, Zurich.

ALMOST FROM THE START, industrial robots were endowed with rudimentary sensing capabilities; simple sensors of touch and velocity were needed for basic feedback and to automatically correct motion and grip. Today, computers can easily interpret all sorts of sensory data, and they can do so fast enough to allow for real-time interaction (at least in the case of relatively slow movements like those of a robotic arm). Early roboticists thought that intelligent industrial robots would soon learn to pick and choose loose components from a random heap or straight out of a bin. That remains a tall order today, but thanks to the combination of advanced sensors and artificial intelligence, or machine learning, industrial robots can now be tasked with making independent decisions on the fly, based on their detection of unpredictable factors. This mode of operation used to be called adaptive fabrication, and it was originally meant to cope with local incidents in a traditional, highly controlled industrial environment. When generalized, however, the adaptive or intelligent skills of today’s industrial robots are the harbinger of a truly revolutionary approach to design and making.

Institute for Computational Design, robotic construction of ICD/ITKE Research Pavilion, Stuttgart, June 13, 2015.

Modern design is an authorial, allographic, notational tool of control based on calculation and prediction. Its customary modern vehicle, the engineering blueprint, is based on the assumption that humans (the workers on the receiving end) will do as told, and that materials (all the physical parts to be put together) will behave as expected. To that end, industrial workers were trained to forsake all intelligent skills they might have possessed, and industrial materials were standardized over time to make them compliant with modern mathematical models (and not the other way around): Steel is a case in point. Stones, as found in nature, are all different from one another, so modern engineers had to process them to make them bland and tame, homogeneous, predictable, and calculable—thus creating an artificial stone, which we call concrete. Ditto for timber. Engineers cannot work with timber as found, because each log is different when split from a felled tree; for that reason, the timber we use in building is “engineered” and served up in standardized, heavily processed formats—as plywood, particleboard, laminated timber, etc. 

Today, we can use computation and robotic labor to reproduce at least some of the preindustrial artisanal economy and its inherent, circular sustainability.

Today, however, intelligent robots tasked with building a wall in an open field could in theory scan the horizon, pick and choose from the stones and boulders in sight, analyze, combine, and assemble those pieces as found to minimize waste, and pack them together without any need for infill or mortar. Likewise, intelligent robots could, theoretically, compose with the random shapes and structural irregularities of natural timber, fitting together each log as found, or almost, without having to bring it to a faraway plant, slice it, or reduce it to pulp, then mix it with glue and other chemicals to convert it into boards with standard measurements and tested structural performance. Preindustrial artisans—living as they did in a world of physiocratic penury, where manufacturing and building were at the mercy of local supplies of materials and labor—had to make do with whatever they found on-site. Today, we can use computation and robotic labor to reproduce at least some of this preindustrial artisanal economy and its inherent, circular sustainability. Achim Menges’s Institute for Computational Design in Stuttgart, in particular, has shown the potential of adaptive robotic fabrication for dealing with the variations—some would say the vitality—of natural materials, in addition to or even partly in place of traditional architectural notations and predictive structural design. Also noteworthy is that some recent buildings by the celebrated Japanese architect Kengo Kuma show evident formal affinities with the discrete computational work being discussed here—without any appeal to computational theories, and driven exclusively by Kuma’s creative reinterpretation of traditional craft.

Institute for Computational Design, robotic construction of ICD/ITKE Research Pavilion, Stuttgart, June 13, 2015.

OTHER SCHOOLS of computational design take no interest in the variations of natural materials, preferring instead to emphasize the combinatory and modular features inherent in the technical logic of robotic assembly. Bricks have been made to measure for human manipulation since the beginning of time; today, more powerful robotic arms can easily deal with bigger and heavier chunks. In particular, the automated assembly of standardized, modular chunks (mostly made of industrial-grade, processed timber) has become the visual marquee of the so-called Computational Brutalists (see in particular the work of Gilles Retsin and, with different premises, of Daniel Köhler at London’s Bartlett School of Architecture, though similar technical and logistic considerations also underpin work produced at the Southern California Institute of Architecture, at the University of Southern California, at Massachusetts Institute of Technology, and elsewhere). The averred predilection of some Computational Brutalists for the mass production of a limited number of conspicuous prefabricated components may appear odd or retrograde in today’s post-mechanical context. Modularity was a staple of the industrial revolution: By using interchangeable components, we can make more stuff out of fewer parts, thus maximizing economies of scale through the mass production of identical components. But digital fabrication does not work that way: When digitally fabricated, nonstandard parts do not cost more than standard ones. Likewise, computer-driven, identical robot movements are not cheaper than variable ones; today’s intelligent robotic arms can (in theory) perform endlessly different or repetitive motions at the same cost per motion. One may therefore surmise that the rising appeal of computational modularity (the nonstandard assembly of standardized parts) may be due to other, nontechnical factors.

Institute for Computational Design, robotic fabrication of a wall for Sewn Timber Shell, Shenzhen, November 2, 2017.

For reasons too complicated to explain, and perhaps inexplicable, the curvy smoothness that characterized end-of-century digital design is today often seen as the architectural style of choice for neoliberals, neoconservatives, and free-marketeers of all denominations around the world. This symbolism is now so pervasive that many on the opposite site of the ideological divide ended up, often unintentionally, championing stuff that looks the opposite. In eighteenth-century art theory, the opposite of “smooth” was “rough.”2 Today, the design community seems to agree that the opposite of architectural smoothness is architectural chunkiness. Dissenters of all ilks have a long history of allegiance to expressions and representations of aggregation, dissonance, and disjointedness in the arts. In the same tradition, the architectural chunk is a de facto rallying cry of today’s activist Left.

Gilles Retsin, Diamonds House, 2016, Belgium. Rendering.

More controversially, similar aesthetic and political choices today often team up with nostalgia for the technical system of mechanical modernity, for the social system it begot, and for its main avatar, the industrial factory.3 Many who regret the rise of robotic manufacturing never saw an industrial assembly line—let alone worked on one, evidently. But that’s not the point. Today’s nonstandard robots—as redefined and reinvented by architects and designers—will not automate the moving assembly line: They will eliminate it. They will not replace the industrial worker: They will create the automated version of a preindustrial artisan. One hundred years ago, Le Corbusier thought that building sites should become factories; by a curious reversal of roles, today’s robotic revolution promises to turn the postindustrial factory into something similar to a preindustrial building site. The intelligent, adaptive, “agile” robots being developed by the design community are likely the future of manufacturing, but the social and economic import of this technological revolution—unleashed, almost accidentally, by research in computational design and architectural automation—far transcends the ambit of our discipline, and raises questions of greater consequence, and of a more general nature. 

Mario Carpo is Reyner Banham Professor of Architectural History and Theory at the Bartlett School of Architecture, University College London. 



1. Frederick Winslow Taylor, The Principles of Scientific Management (New York: Harper and Brothers, 1911), 24, 34, 36. The first moving assembly line for the mass production of an entire automobile was inaugurated at the Ford plant of Highland Park, Michigan, in the fall of 1913.

2. See in particular William Gilpin, Three Essays: On Picturesque Beauty, on Picturesque Travel, and on Sketching Landscape (London: Blamire, 1792), 26. 

3. As an anecdotal example, the radical UK organization Novara Media, noted among other things for its vocal support of Jeremy Corbyn and for the pro-Brexit position of some of its leaders, was named after the North Italian industrial town where Elio Petri shot the film The Working Class Goes to Heaven (1971). The movie was a parody of the daily life of the Italian industrial proletariat of the time, but it was shot in a real factory, showing the actual machinery, tools, and assembly lines then in use for the production of electromechanical elevators.