Mario Carpo

Publication History:

“Computational Design, Artificial Intelligence, and the End of New.” Review of Architecture and Building Science of the Architectural Institute of Korea 558, (69, 11, November 2025): 30-33 (in English and Korean).

The text posted here is a preprint draft and it is significantly different from the published version. Please only cite from copy in print

What’s the most inflated, stereotyped, ubiquitous cultural trope of the 20th century—across languages and cultures?  I am certain that, if asked, we could all come up with a range of believable candidates, and there is no way to provide an objective, verifiable answer.  For what that is worth, the answer I would pick is: newness, or the urge to be new.  This is what, it would appear, almost everyone was permanently looking for and aiming at, throughout most of the last century.  In the first season of the American television series Mad Men (aired in 2007, but recounting a story that would have taken place in 1960) the fictional advertizing wizard Don Draper argues that “new” is the most important word in advertizing: new is “what creates an itch”; what invites consumption.  That may well have been the case in the America in the 1960s.  When I was student in Italy in the 1970s, we used to make fun of all the cultural fads coming from France at the time: the nouveau roman, the nouveau réalisme, the nouvelle vague and the nouvelle histoire—and even Art Nouveau, which started it all, at the end of the 19th century, and which wasn’t even French—it was Belgian.  But it still purported to be new—like almost everything else throughout most of the 20th century, and likely to this day. 

There is a good reason for the overwhelming importance that the term “new” acquired in the culture of industrial modernity.  The industrial revolution itself was new, evidently, and new were all the products and processes and technologies that came with it; new in the most banal and simple sense of being unprecedented—never seen before.  In architecture, steel, laminated glass and reinforced concrete were new; so were elevators, electric light, central heating and air conditioning.  The industrial metropolis was new.  As every aspect of daily life was new, a consensus soon was reached that art and culture should renew, too.  And so they did.  The theory of progress that was born with 19th century historicism implies a notion of permanent technological innovation, and while mechanical technologies have been mostly stagnant since at least the late 1970s, electronic technologies have now taken up the mantle, and computers have been the standard bearers and almost the symbols of innovation for the last thirty years or longer. They are, today, what’s new—the novelty.

At the same time, modern, Turing-complete computers have been around for long enough to allow us to look at the rise of electronic computation from some distance.  We can now assess the inner and outer workings of digital tools in a longer historical perspective, discerning the long-range trends hidden behind, or beneath, the short-run time of events.  And if we do so, we must come to the conclusion that every step forward of digital technologies for the last thirty years has been the negation, the reversal, or the nemesis of some basic tenets of modernist innovation: in short, the digital turn is not the continuation, but a turn away from industrial modernity; it is modernity in reverse.

I know this may sound odd, and hard to believe: how would a new technology negate the technologies from which it descends?  Computation does not come from another planet; it still uses, for the most part, the same transistors and electric circuitry we use to microwave potatoes, climb to the top of a skyscraper, or power a tramway.  But this is the thing: regardless of the technology we put into computers, the way computers work, and solve problems, is the opposite of the modern way of thinking, and the negation of the industrial way of making.  Let me illustrate this apparent paradox with a few examples.

The most disruptive idea that accompanied the rise of computer-aided design and manufacturing (CAD-CAM) in the 1990s was the idea of mass-customization.  In the industrial way of making (think of print with moveable types) once a mechanical matrix is made, it must be reused many times over in order to amortize its upfront cost, by spreading it onto many identical copies.  Thus mass-production achieves economies of scale; and this is why industrial products that are made this way must be all the same (standardized).  This is what economists call the law of decreasing marginal costs; and this is the technical logic underpinning the modern, industrial mode of production.  But digital fabrication processes, like CNC milling, or 3D printing, do not use casts, stamps, molds, or dies.  When we make things digitally (think of a laser printer), most of the time there is no need to reuse the same mechanical matrix, because there isn’t one in the first place.  So each copy can be different if needed—and in theory at the same unit cost.  In the old mechanical-industrial system, the more identical copies we make, the cheaper each copy will be; in the new, post-industrial mode of production, making more identical copies of the same item will not make any of them cheaper.  In short, if we make things digitally, standardization no longer saves money, and customization no longer entail additional costs.

The socio-economic consequences of the shift from mechanical mass-production to digital mass-customization are epoch-making.  In the old, industrial system bigger factories meant lower costs, and bigger markets meant cheaper goods.  Hence it made sense to concentrate manufacturing in a few convenient locations, regardless of the distance between factories and their markets; regardless of the costs of transportation; and regardless of custom duties and tariffs, which were low and were expected to get lower.  But with digital manufacturing, in theory, we can make one car, or one million identical cars, or one million different cars, at the same cost per car; hence we could make almost anything we need where we need it, when we need it, as we need it—on time, on site, to specs, and on demand.  A global factory in another continent could now be replaced by a network of distributed, local robotic micro-factories.  To some extent this is already happening: the cost of robotic labor is the same in Zurich or in Vietnam; the cost of a kWh generated by the same solar panel is the same wherever the sun shines—regardless of how many solar panels we install next to one another.  Production can now be decoupled from cheap human labor, and, importantly, from scale.  The permanent quest for scale that accompanied the rise of modernity is over; bigger is no longer cheaper.

Is this “new”?  Well, it is new with regard to the history of industrialization, because it reverses its course.  But it is not new at all if seen in the longer duration of the history of manufacturing.  Digital making is simply reviving the technical logic of artisanship—and with it, its mentality, its scale, its aesthetics, and even its politics—if not yet its modes of social organization.  The consequences of distributed manufacturing for urban planning are equally momentous.  If we can make stuff wherever we can install and service a couple of robots, manufacturing needs not be kept apart from other functions of cities. 

Moreover, we have known for a long time—and the COVID pandemic has proven—that just like today’s blue-collar work doesn’t need big factories, today’s internet based white- collar work may not need big offices—and it certainly needs less office space than we thought.  Likewise, retail now needs less floor space, entertainment fewer movie theatres, etc.; in short, due to computer-driven “despatialization,” the modernist theory of zoning, aimed at the separation of land uses in cities, is defunct.  Today living and working can happen in the same block, in the same building, even in the same apartment.  Which is exactly what happened before the industrial age.  This is the way pre-industrial artisans lived and worked and sold their wares—all of it often under the same roof.  The digital turn is simply bringing back the functional organization of pre-modern cities.

Let’s look at another aspect of the digital turn, well known to designers.  Before the rise of CAD-CAM, modern architects could only build forms that they could notate in orthographic projections—plans, elevations, and sections, proportionally drawn to scale.  This limited the range of buildable forms to the kind of forms designers could draw and measure on paper; as a result, built architecture tended to be made of the repetition, or combination, of simple geometrical shapes.  But computation has removed this notational straitjacket; using digital tools, architects can now measure and build very complex geometrical shapes, as well as non-geometrical ones (free form).  This is how, starting from early 1990s, Frank Gehry, to name one, could build irregular volumes that were originally crafted in three dimensions by the free, unimpeded movement of his own hand—hand-sculpted, in fact.  Computation eliminated the mediation of geometrical notations, and the limits and constraints that came with them.  Again, this is the way most buildings were built before Renaissance humanists came up with the novel—and, at the time, perfectly outlandish—idea that buildings should be fully designed on paper prior to being built. 

Other, important and starkly anti-modern features of the digital turn became apparent in the course of the last twenty years.  Let me just mention, briefly, the eminently collaborative nature of digital creation—which, for designers, is embodied chiefly by a family of software known, generically, as building information modeling.  BIM has reversed the modernist idea of building as the creation of a single, humanistic author—a staple of Western art theory since the Renaissance (but perfectly irrelevant in parts of the world that were not subjected to Eurocentric ideas).  And, last, let’s have a look at the recent and utterly unexpected rise of Generative AI—which has fascinated and flabbergasted scientists, artists, and the general public alike since the spring of 2022. 

As we now know full well, Generative AI is a prodigious imitation machine: if we use it to make images, it creates images that are derived, somewhat mysteriously, from the dataset (or corpus) of images on which the system was “trained.”  Thus Generative AI reminds us that nothing is created out of nothing, but everything derives, to some extent, from things that already exist.  Whether dealing with human or artificial intelligence, the logic appears to be the same: there is no creation without some degree of creative imitation (meaning the reuse, assimilation, and transformation) of relevant precedents.  The latest trend in artificial intelligence, often called “machine learning,” means that machines can learn from observing what happens around them; by creatively imitating the models they are shown.  But this is neither new nor particularly original.  Is this not the way we all learn one or more languages (also known as “mother tongues”) before we go to school and we are taught to read and write?  We learn to speak by imitating the voice of our parents.  And this is also the way apprentices of all times and places learned their artisan trades: not from mathematical formulas, nor from any formal schooling, but by observing, and intelligently repurposing, the skills of their masters.  Artisans do not calculate: they imitate a craft perfected through—and warranted by—centuries of expert tradition.  Well, oddly enough: this is also the way today’s Generative AI makes stuff, and solves problems: not by applying rules out of a handbook, but by imitating well-chosen examples—models that are already out there, and which are known to work.

This is the technical logic of today’s newest technology.  “New” is now irrelevant.  There is no “new” creation without and outside the invocation, selection, and acknowledgment of a tradition we refer to, and into which we inscribe ourselves.  New is the awareness of being last in a long line, which made us what we are—dwarves on the shoulders of giants.  This is the teaching of the newest of today’s new technologies—which can make pretty much everything, but will never make anything “new.”  And this also reminds us that—for better or worse—we may not even need that kind of newness any more.  The mandate of newness—the imperative to be new; the obligation to always create something new—well, that was itself a brand-new idea: a revolutionary novelty when the first moderns came up with it.  But in the grand scheme of history, that was just a blip—a flash in the pan: a straw fire, soon kindled and soon burnt.