n°31 Artifacts of Intelligence and Artificial Intelligence

Architecture is the expression of an intelligence that can henceforth be viewed in a new light. Artificial intelligence—source of anxiety for some, of enthusiasm for others—is now on everyone’s lips. What will happen to today’s professions and what will be the value of acquired knowledge once these expanding systems have transformed our practices and our means of production? Architecture cannot escape the ensuing upheaval. These days, software like Midjourney can generate images of buildings and sites based solely on one dictated sentence. Will plans and cross-sections soon be generated by a single click once functional guidelines and current regulations have been uploaded, along with a desired style? Let us make Spinoza’s maxim our own: Neither laugh, nor weep, nor scorn, but understand.[1] And in order to do so, let us briefly hark back to the way things evolved.

Architecture is an ancient discipline built on the tools of design. Square and compass, drawings on paper, and stone-cutting diagrams, followed by computer graphics and Building Information Modeling (BIM), have dotted a history in which technique was placed at the service of the mind. Might the tools now be replacing the master? The future offered by generative artificial intelligence remains uncertain, while the present is teeming with fundamental, unprecedented questions. Does the power of algorithmic computation threaten the work of architectural projects, extending from the conception of forms to the monitoring of execution? Are skill, acquired experience, and mastery of decision-making destined to fade behind machine learning, or will they, on the contrary, be precisely what is needed to fully harness robots? These questions prompt us to consider the imponderable dimensions of the work of the mind, dimensions foreign to codifiable criteria of production. We might wonder, for example, if the human mind is the only means of considering a site and its history, of designing places, of ennobling their use, of playing on rules and taming the laws of design in order to achieve unique forms.

The progressive delegation of intellectual production to machine systems calls for reflection on what the mind—and mind alone—possesses. Thus the contents of this issue opens with an article in which I question the sovereignty of the self when it comes to architectural invention in the West, going on to question the political meaning of the art of building. The immaterial capital of memory is deposited in a safe place, namely the brain, which is largely overlooked in debates on artificial intelligence. By exploring overlooked connections between mental space and the storage of data with no external support, Julien Gougeat recounts the invisible worksite to which we all have access. At a time of increasing dependence on smart phones, clouds, and soon on prostheses of “augmented humans,” a living, randomly accessed memory seems to be both an art and a political issue. Strangely, artificial intelligence obliges us to define what is human. Two authors evoke the humanistic age to place artificial intelligence in the perspective of the grand history of concepts and techniques. For Pierre Caye, there exists an intelligence specific to disegno, that quattrocento concept in which mental and instrumental functions are sharpened, achieving a high degree of power. Readers will find there the substance to nourish a critique of productive artificial intelligence whose unsurpassable efficiency is now being touted daily. Valérie Charolles, meanwhile, questions the binary categories derived from the Enlightenment, notably an overly strict opposition between humanity and nature; in sixteenth-century thinkers she rediscovers the spirit of nuances and multiplicities, so necessary to an “intelligent” grasp of artificial intelligence today. History remains in the forefront of Mario Carpo’s article, which relativizes the shock heralded by artificial intelligence, referring back to the tradition of imitation and, more generally, to the transfer of forms through which, ever since Antiquity, the new has been wrought from the old. This retrospective view offers an interpretation based on the status of precedents in art, and it also interrogates the now-forgotten notion of style.

These articles, which draw on the past to propose a sound welcome to the coming world, are followed by others anchored in current practice. Leda Dimitriadi explores the limits of processes designed to resolve ordinary problems of project management. She points to a set of considerations—some of which are subtle and non-codifiable—that are scarcely accessible to artificial intelligence in the current technical scheme. The process that emerges here is not so much one of tools as of society itself, which is free to recognize the importance of imponderables, and thus to define what it expects from architects. Laurent Lescop expects artificial intelligence not to replace architects, but to inevitably transform their activity, whereas Neil Leach more frankly claims that, as feared, humans will be surpassed. On a lighter yet equally scholarly note, François Frédéric Muller offers here the testimony of a practitioner, reflecting on the gradual easing of labor through digital tools—though not without consequences: he traces, thread by thread, the web that has been tightening around architectural imagination from computer graphics to BIM, then he surprises us by viewing from an archaeological perspective the relationship between words and things long prior to the magic of prompts.

We are only at the dawn of this new era. Artificial intelligence is growing at such a speed that even the time to produce this issue has already resulted in a lag behind current developments, while everyone has a different grasp of the impact it will have on our profession and our lives. Some architects immediately see artificial intelligence as a threat; others are eager to embrace it in order to interact with the machine, so that humanity ultimately benefits from turning to what exceeds certain of its capacities.

Architectural intelligence is measured above all against projects that are actually built, so we are keeping our promise to readers to “pay a visit” whenever a site is worth the trip. Laurent Salomon follows in the footsteps of Zhu Pei, revealing the evocative power of Pei’s refined oeuvre in a landscape where the raw power of industrial construction usually prevails.

Translated from the French by Deke Dusinberre

What is architectural inventiveness? It might be defined as the art of destabilizing a state of knowledge with the very tools of that knowledge. In the Vitruvian tradition, architects first played upon the grammar, then critiqued it, and ultimately contradicted it in the Baroque era, liberating a repertoire of architectural figures and original approaches down through the centuries. In the early twentieth century, the decline of the classical idiom triggered new imaginative faculties explored by masters both great and minor. The playing field then shifted from the classical orders to wider possibilities whose borders were vague but structured by a number of doctrines. The postmodern era, meanwhile, is marked by a profound solitude of the individual self: no major voice has made itself heard, no treatise or manifesto is guiding our actions. For the first time, arbitrariness is given free rein, prey to the whims of fashion. In this period of intellectual fragility, an alleged interlocutor has appeared on the scene, in the form of software. It promises to end the toil of design, to save energy, to facilitate production—in short, a panacea for the construction industry. Faced with what might well triumph due to its incredible efficiency, hiding from it would be as trite a response as swallowing it whole. The challenge is to define what ultimately counts, and what must be saved: style and sovereignty. At stake is not just the future of project design, but also everything that architecture represents and its contribution to a self-governing society.

All intelligence is artificial, that is to say cultivated, practiced, and indeed equipped with technical prostheses of all kinds, failing which nothing would be comprehensible. In this regard, the finest example of artificial intelligence can be found in the arts, notably Renaissance art through its development of disegno (“design” and/or “drawing”). Disegno is not merely the core of the arts of the Renaissance, in a broader way it constitutes an episteme, namely a structure or operation that defines the potential conditions for all knowledge and determines the way scientific fields are constituted. It futhermore structures relationships between the various types of knowledge in a given era, thus profoundly influencing the overall organization of society and all its practices. This episteme was not limited to the Renaissance, but governed Western culture with great coherence down through the centuries, up to the early twentieth century.

Since 2014, Generative Adversarial Networks (GANs: initially a general machine learning technology) have been trained to recognize similarities in a corpus of images labeled as instances of the same term or concept. In reverse mode, this same technology can create new sets of artificial images (including realistic images of non-existent originals) that illustrate certain terms or ideas. The technology has also been modified to merge aspects of two existing sets of images, and in 2015–2016, computer scientists (probably unaware of the major historical implications of the words they were using) dubbed this process “style transfer”—a term that has stuck and is now in common use. Although the practical applications of AI-based image creation technologies are negligible at this point, the use of generative AI in the visual arts has already prompted a reassessment of certain critical categories that have long been dormant, absent from artistic practice and critical discourse, or deliberately expunged. A technology that offers every artist or designer a quick and easy way to imitate the style of any master, or several masters combined, is bound to raise questions about the nature and definition of style, as well as imitation itself. This is where the history of the classical tradition could be of some help to today’s computational designers, for imitation has been at the heart of classical art theory from its Greek origins to its modern and neoclassical revivals.

Some forms of artificial intelligence are already present and used in architecture: complex geometries, generative algorithmic processes, and enhanced models like BIM. Rather than dissipating the creative act in the realm of algorithms, the “computational” or “non-standard” approach—embraced by such figures as Bernard Cache, Zaha Hadid—aimed to boost the architect’s directive role and give greater luster to the figure of the master. The shift toward machine learning signals a new relationship between humans and technoscience, based on artificial neural networks and the use of vast datasets. Whereas some think that this development will lead to a future where machines take over the profession, I offer a different hypothesis: that certain fundamental choices made by the architect cannot be reduced to the power and logic of the computer. Identification of these choices and their significance is essential, as the threat facing architects derives less from the machine than from society itself.

Ever since November of 2022 the irruption of artificial intelligence on the world scene has profoundly transformed the creative professions and architecture in particular. While the profession of architect itself is not at risk of disappearing, the acceleration of its processes, the surge in volume of data, and the use of natural language to interact with software systems, modify the ranking of worth and redefine the link between billable hours and added value. Quasi-real-time, now a factor in the execution of many of a project’s phases, shifts professional effort toward feeding, fitting parameters to, and verifying input. This calls for reevaluating the different roles within a firm: strategy, analysis, and resolution of complicated problems take precedence over repetitive tasks. This new timeframe also alters techniques of conceptualization. No longer does the process aim only at refining an initial idea, but also to explore a wide spectrum of solutions in order to discard the less relevant. Parallel integration of climatic, mechanical or economics-related input must improve the quality of decisions. This kind of processing speed can create the illusion that the profession is getting simpler; on the contrary, it demands an increase in the expertise needed to manage it. The financial models of architectural firms must evolve in response. Jobs such as analyzing calls for tenders, feasibility studies or generating plans take less time thanks to generative tools but require even stricter oversight. Thus does AI redefine the contours and contents of projects assigned to architects.

Whether reticent or impatient to dive into the vast sea of artificially intelligent design, architecture hardly recalls that for centuries it was the applied partner of a singular form of memory: artificial memory. In a situation in which all of us blithely buy dematerialized memory storage but can hardly remember the last two lines of a novel, this article attempts to shed light on the deep links between architecture, memory, and conceptualization. It invokes Frances Yates’ Art of Memory to explain ancient and medieval systems of memorization, seeking to explore the development of the fertile relationship between architecture of stone and mnemonic spaces.

Having only just recovered from compulsory BIM, architects are now being urged to jump on the AI revolution bandwagon. But what revolution are we talking about here? The automatic generation of images that is pushing perspective artists toward the exit? Or the parametric generation of plans, which has developers drooling with desire, thrilled at the prospect of milking the value of their land? The fact is, there is something vertiginous about the prospect of algorithms that are more capable, more knowledgeable, and more adept at synthesis than we are. Architecture schools are rushing to adapt to the new fad, RE 2020 environmental training programs are suddenly looking like yesterday’s thing, and the whole architectural world is trembling at the idea of being left behind. Architects survived the advent of CAD in the 1990s and embraced BIM even before project owners turned away from it. Why should AI be any different? What is so different about the fear it inspires? Perhaps it is the literary nature of AI that takes us by surprise. Whereas CAD and BIM were presented as simple tools—digital versions of our ancient drawing instruments—AI stands apart from mechanics and challenges our ability to write good “prompts.” No machine has come so close to the rhetorical aspect of our profession. But is a project the sole answer to well-formulated questions? Or is it, still, that imperfect and brilliant thing, the answer to the question no one asked?

Everyone is talking about AI these days. But what exactly is AI ? How did it evolve ? And what potential does it have to influence the future ? This lecture takes you on a roller coaster ride looking at the extraordinary—but often somewhat terrifying—potential of what is arguably the most significant invention of humankind. The lecture concludes that we are about to face a radically different form of intelligence—an “alien intelligence”—that will far exceed human intelligence, and completely transform the discipline of architecture.

The development of artificial intelligence usually provokes reactions of either extreme expectation or outright rejection. While AI has already become a reality in the realm of spatial design, a philosophical examination of machine learning offers a more nuanced understanding, revealing such illusions as the notion that these systems are able to operate without human agency. Regardless of its autonomy or intelligence, AI has arisen during a technological moment defined by networked screens, one that in many ways marks the end of naturalness. That is to say, there is no longer any part of the Earth’s surface untouched by human activity or technology. The current design of AI is shaped by mechanisms of echo and reflection seen in social media platforms, search engines, and other interfaces, which raises critical questions about the economic models they promote, the societal structures they reinforce, and their long-term sustainability. Moreover, this design is built on a binary language (0/1) rooted in Enlightenment rationality and a worldview that recognizes only truth or falsehood, leaving little room for nuance, gradation, or neutrality. For this reason, this article argues that we turn to the humanist tradition as a foundation for both the design and use of learning machines.

Here, many people today either reduce architecture to facades or else indulge in technical and commercial “greenwashing.” Faced with this situation, a well-meaning ecological trend seeks a return to natural materials to replace the products of an all-powerful industry. Yet far from here, in an extravagantly productivist China, where reality crushes philosophy, there has emerged an ambitious policy of major facilities designed by Pritzer-Prize celebrities yet also by local figures spurred by such ambitions, leading to the blossoming of architects such as Zhu Pei. They are developing an oeuvre critical of the way we inhabit things, sketching an outline of a better future.