Machine Learning

To suss out the future of architecture beyond the wave of automation currently sweeping the field, Mario Carpo turns in his latest book to a genealogy of avant-garde bricklayers. He begins his epic tale with the Renaissance celebrity architect Filippo Brunelleschi, perhaps the closest embodiment of Adolf Loos’s famous quip that “the architect is a bricklayer who’s studied Latin.” Then there’s Frank Bunker Gilbreth, the wily American brickmason who studied the scientific method. Rounding out the cast are Fabio Gramazio and Matthias Kohler, two Swiss architectural researchers who began setting robotic arms to the task of piling up intricate brick walls at the height of computational enthusiasm in the early 2000s.

Among the greatest charms of Carpo’s book is his pairing of poignant historical details with careful analysis sourced from media theory and science and technology studies. Parsing the differing attitudes of Brunelleschi and his contemporary Leon Battista Alberti toward drawings, for instance, Carpo shows that what was at stake in the masonry of the Duomo in Florence was nothing less than the architect’s intellectual property. The success of Brunelleschi’s paranoid control of the cathedral jobsite led a few years later to Alberti’s insistence that workers should be content to build exactly what he himself designed, without the creativity of medieval masonry craft. A building should be a “copy” of the architect’s drawings, just as the concurrent invention of moveable type allowed faithful copies of text (without the idiosyncratic flourishes of medieval illuminated manuscripts). Carpo similarly shows how Gilbreth, the fin de siècle pioneer of scientific management, began his career studying efficient masonry techniques before becoming a productivity guru. Gilbreth took Alberti’s logic to the next level, scripting human movement to match the relentless flow of the assembly line. Gramazio and Kohler’s robot arms then fulfill the project of automation by removing human hands entirely and allowing machines to do what they do best: moving bricks to precisely the right place, without a whimper of complaint about boredom or repetitive stress injuries.

Robots are the true stars of the book. Carpo is a fan of fabrication, but he’s not so swept away as to suggest that a robot can “study” anything. He does not traffic in loose analogies; for him, the very notion of artificial intelligence should elicit the question of what “intelligence” even means. Written before the release of ChatGPT and the current AI craze, Beyond Digital focuses on the fundamentals of computational automation and thereby manages to avoid overemphasizing recent developments. In many ways it is a continuation of Carpo’s previous two books, The Second Digital Turn (2017) and The Alphabet and the Algorithm (2011), which were published in Cynthia Davidson’s “Writing Architecture” series. Beyond Digital leaps out of that series and into MIT Press’s general catalog, probably to reach a wider audience of designers and artists. It’s a smart move because Carpo has an epochal story to tell, and he’s both a born storyteller and remarkably attuned to the contemporary mood. I admit his straight talk—architects today “feel no attachment to the machines they use” and “nobody likes robots”—stings a little for an old-school technophile like myself. (My own book on the artistic joys of assembly lines and algorithms came out last year.) But he’s right.

Sometimes the analysis is a little too rational, and his writing veers into the terrain of technological determinism. He describes three technical logics—artisanal, industrial, and computational—that mesh with three design logics: craft, mass production, and mass customization. “The technical logic of computation is now changing the world as deeply and irrevocably as the technical logic of mechanical machine-making did at the start of the industrial revolution,” he summarizes, “but as the digital turn has only just started, we are still trying to understand what it is precisely, and how it will pan out.” (Despite his books’ definitive titles, Carpo’s prose suggests a more cautious consideration of momentous change currently underway.) In a throwback to midcentury art theory that focused on medium specificity (think Clement Greenberg and early Rosalind Krauss), Carpo delights in pointing out the unique affordances of automation technologies. “Artificial Intelligence is not a spaceship; it is a time machine,” he says in a typical passage, because “it predicts the future by having unlimited access to the past.” He suggests that architects ought to use technologies in the ways they’re meant to be used, perhaps forgetting that, like artists, they usually go about their work with more than a pinch of irrationality. Creativity is not a clockwork mechanism. A counternarrative to Carpo’s would insist that art—and architecture that aspires to that status—always goes against the grain, misusing techniques and pushing technologies beyond their intended boundaries. Conceptual art flourished in the 1960s, for example, as designers turned algorithms into poetry.

But artists probably don’t need any encouragement to imaginatively exploit the potentials of new technologies, and anyway Beyond Digital should be read more as a labor history than as a theory of art. Carpo draws on Marxist themes: His narrative is driven by responses to the shifting landscape of industrial and postindustrial production, and his measure of progress is the degree to which designers and workers are freed from drudgery and the alienation that breeds therein. His story has an air of utopianism about it: Readers are asked to “imagine a world … where economies of scale have ceased to be”—once widespread mass customization marries the best elements of mechanization and craftsmanship. Carpo plants the seeds of multitudinous sci-fi scenarios. What happens to the construction industry in a world of robot masons? What becomes of fashion (and industries like architecture that bank on being fashionable) when products are effortlessly tailored to consumers’ whims?

Carpo’s conclusion seems inevitable: Mass customization takes command, to adapt a phrase from Sigfried Giedion. And it may indeed already have come to pass in some precincts. Nike.com offers customizable footwear with little fanfare. Prodding architects to catch up with the times has been a theme of Carpo’s since 2011, when he made the case that architects should design algorithms rather than individual buildings; the algorithm can then spit out any number of similar structures. Greg Lynn’s Embryologic Houses (2000) remain the best demonstration of the idea: Take user input, then “grow” a house to fit those needs. Tellingly, however, none were built. Carpo’s antenna for vibe shifts may be picking up premonitions that take decades to develop.

Ultimately, it’s hard to say whether Carpo’s analysis really holds up as well for architecture as it does for small goods like shoes, and his analysis may be missing a crucial perspective by narrowly focusing on the production side of the economic picture. We are after all in a consumer economy. To say that architects are more than glorified service providers—that they can meaningfully affect the production process—is a stretch. Carpo doesn’t entirely neglect consumer preferences, but at the timescale he’s looking at—from the fifteenth century to the present—they have little sway over supply-side shifts. He offers a fascinating anecdote about how some people strongly disliked early printed books because they looked cheap in comparison to their hand-illuminated predecessors—but of course they didn’t just look cheap, they really were cheap, and it didn’t take long for affordability to trump aesthetics. Here Carpo draws from material presented in his first book, Architecture in the Age of Printing (2001), and indeed his scholarly consistency over the past quarter century has allowed him to probe deeper into more nooks of media theory than perhaps any other architectural historian.

It’s at the hinge between production and consumption that oversimplifications begin to appear in Carpo’s quick sketches of consumer choice. While many people now like the look of artisanal production, this aesthetic preference is far from universal. In “Ornament and Crime” (1908), Loos pinpointed the modernist mood in his predilection for everything anonymous, from standardized men’s suits and dress shoes to blank façades and plumbing fixtures unencumbered by decoration. Loos’s diatribe hung on the immorality of wasted labor, and a rejoinder to Carpo might begin by focusing on the complexities of waste. He points out that the mass production of building materials is less efficient than mass customization would be, citing the identical steel beams in Mies van der Rohe’s Neue Nationalgalerie (1968). Beams customized to match their loads would have saved tons of raw material. But central to the logic of modernism is its visual rhetoric: The uniform backdrop of ceiling, walls, and floors allows the artworks on display and the people milling around to stand out in dramatic relief. The expense required to achieve the minimalism of the Neue Nationalgalerie’s gridded ceiling is also kind of the point; luxury depends on extravagance. The calculus of excess is more complicated than measuring pounds of steel, however. Which consigns more matter to the landfill: an interior defined by a framework of oversize beams that remains unchanged for decades or a gallery space meant to be gutted and refashioned every few years?

This line of questioning is a necessary follow-up to Carpo’s persuasive but not-100-percent-airtight narrative. More examples need to be gamed out. Which is more wasteful: a world with just a few choices of mass-produced shoes or a world with an unlimited variety of mass-customized footwear? (I’d guess that mass customization across that industry would also mean a stress test of the world’s shoe racks.) Or take housing. If everyone’s housing were to somehow be continuously molded to fit their changing needs, there would be less wasted space (unused bedrooms when the kids go to college, and the like). But wouldn’t this entail a constant and extremely excessive process of tearing down and rebuilding domestic space?

The principle of loose fit—and the corollary that some wasted space is a good thing because it allows for flexibility and adaptive reuse—is not just a utilitarian ideal. It’s also at the heart of architecture’s claim to provide a public good. Now’s a good time to have this sort of conversation about values. Should we all want custom teapots? It seems narcissistic to me. The greatest oversight of Carpo’s book, in my reading, is that he doesn’t ask this sort of ethical question directly. He tends to see the bright side of production technologies but not the dark side of the consumerism they induce. I’ll take IKEA instead, along with the industrial-era dream of simple, cheap, durable water boiling devices for all. This is where the discourse on climate change, which is edging toward an ethics of degrowth, seems to be headed.

While Carpo’s take on mass customization is probably not the last word on the matter, he may have the last laugh at the expense of ’90s-style digital architecture. He’s been one of the keenest observers of the phenomenon since the era of folds and blobs, and in Beyond Digital he saves his harshest critique for what he calls “complexism”—“one of the most pernicious and nefarious ideologies of our time.” The complexity theory championed since the late 1980s by theorists such as Sanford Kwinter and practiced in various ways by the likes of Lynn, Reiser+Umemoto, and François Roche appears in retrospect a little too complicit with neoliberal economic enthusiasm, with potentially dire consequences. As Carpo puts it, “if we let pandemics self-organize we let millions die; if we let societies self-organize, we relinquish democratic life.” It’s fascinating to have witnessed, in the span of a couple decades, a wholesale shift from posthumanist experimentalism to humanist values.

Carpo is, in the end, a humanist—just as you might expect of a Renaissance scholar. His soft spot for six-axis robot arms doesn’t mean he’s totally pro-tech, and in this respect Beyond Digital tracks one of the most captivating aspects of the current era. We’re all digital device connoisseurs, social platform ethnographers, AI anthropologists. Advice columns help us navigate the social niceties of using generative chatbots. (Automating a first draft of your wedding vows is OK, I’ve heard, but if you don’t rewrite them in your own words you’re in for trouble.) Beyond Digital is here to help architects think through the subtle trade-offs that come along with automation technologies. It’s a rare technical primer that pairs theory with narrative and adds a dash of humor, all while managing to highlight the paradigm-shifting choices we’re now in the position to make as designers. So remember to pause and think before you shift—and spare a thought for the standard teapot.