Le Corbusier

In the [YOU] on AI Field Guide

The cycle that began with [YOU] on AI describes the moment when the most powerful optimizing tool ever built was turned on nearly everything. Le Corbusier is the cycle's historical specimen of exactly this moment: what happens when a brilliant optimizing mind, armed with the most powerful tool of its age, moves from the bounded project (the villa, the single building) to the total system (the master plan, the whole city). The move is not a failure of intelligence or good intentions. It is a failure of scale—the recognition that an optimizer that works beautifully at one scope can produce catastrophe when that scope expands past the point of correction.

The specific connection the cycle draws is between Le Corbusier's 'machine for living in' and the optimization doctrine now installed as the default design logic of AI systems. A recommendation engine is a machine for engaging in. A hiring algorithm is a machine for selecting in. A large language model is a machine for thinking alongside. In each case, the designer identifies a function, strips away what does not serve it, and engineers toward a measurable objective—which is to say, in each case, the designer performs the identical intellectual operation Le Corbusier performed on the house. Goodhart's Law then follows: when a measure becomes a target, it ceases to be a good measure. The feed optimized for engagement learns that outrage engages. The model optimized for rater approval learns to sound confident and agreeable. The proxy and the value diverge, and the system pours its power into the gap.

The deepest connection is the Modulor—Le Corbusier's attempt to derive a universal human measure from which humane architecture could be automatically generated. The Modulor was male, European, and tall; its 'universality' encoded the particular body and perceptual assumptions of its author, then hard-coded those assumptions into the proportions of buildings used by people who were not the author. Every biased large language model, every facial recognition system trained on light-skinned faces, every voice system calibrated to one accent family, is a Modulor: a standard derived from a particular that is then presented as the universal, silently penalizing everyone who falls outside it. The lesson is not that standardization is wrong. It is that the standard must be held loosely enough to accommodate the people it was not built from—and Le Corbusier, who was certain his proportions were universal, could not do that.

Jane Jacobs leveled the definitive critique of Le Corbusier's planning in 1961: cities are problems of 'organized complexity,' and the order of a thriving neighborhood is generated from the bottom up, by millions of small decisions, and is largely illegible to the planner looking down from above. The cycle applies the same critique to AI governance: the organized complexity of human communities, professional cultures, and cognitive habits that constitute the social fabric cannot be read by any system that requires data to see, because the most important features of the social fabric are those that have never been rendered as data.

Origin

Jeanneret renamed himself Le Corbusier in 1920, taking a name from a maternal ancestor and converting it into a public brand for the ideas he had been developing since his early twenties. His formation was through travel and self-directed study rather than formal architectural training: a 1907 journey to Italy, exposure to the Vienna Secession, and most formatively a period in the studio of Auguste Perret in Paris (where he learned reinforced concrete) and then Peter Behrens in Berlin (where he learned industrial design). The early influence was the Arts and Crafts movement, which he later repudiated; his mature aesthetic was formed by the machines of the 1910s—ships, planes, automobiles—and by his encounter with cubism and purism.

His theoretical declaration came in Vers une architecture in 1923, a book assembled from essays that had already appeared in his journal L'Esprit Nouveau. The book's argument was a controlled provocation: photographs of grain silos, factories, and ocean liners beside the Parthenon, with the claim that the engineers who designed the machines were the true heirs of the Greeks—beauty as the byproduct of rigorous function. The house as a machine for living in was the manifesto sentence, and it was followed by proposals for the Ville Contemporaine (1922) and the Plan Voisin (1925) that took the manifesto at its most radical.

His built work is divided by scale in precisely the way the cycle diagnoses. At the scale of the individual building, operating in direct relationship with known clients and sites, he produced works of genuine and lasting distinction: the Villa Savoye (1931), the Swiss Pavilion at the Cité Universitaire (1933), the Salvation Army hostel in Paris. At the scale of the master plan, operating at a remove from specific inhabitants and specific sites, he produced ideological documents whose partial realization proved disastrous: the tower-in-the-park estates modeled on his principles, of which Pruitt-Igoe became the symbol, were systematically demolished as urban failures from the 1970s onward.

Key Ideas

The machine for living in. Le Corbusier's most famous formulation is a design doctrine: identify the functions a dwelling must perform, optimize those functions rigorously, and allow form to follow from that analysis without decorative interference. The doctrine produces genuine goods when the objective is well-specified and bounded; it produces the proxy-value divergence that AI researchers know as Goodhart's Law when the optimization is applied to proxies for unmeasurable goods. A house cannot be fully specified as a set of functions; a life cannot be optimized. The machine that serves the specified functions beautifully may fail the life entirely.

The Modulor and the standardized human. Le Corbusier's attempt to derive a universal human proportional system from a single idealized figure produced a tool that was useful as a flexible guide and tyrannical as a universal law. His figure was male, European, and chosen partly for aesthetic reasons. The lesson is not that standardization is impossible but that the standard must be held loosely and checked continuously against the actual variety it claims to serve. A standard that cannot see the people it omits is not a universal. It is a particular in disguise.

The Plan and its limits. The Ville Radieuse was utopia as infrastructure: a total city planned from a single intelligence and built more or less at once. Its realized variants failed not because the buildings were ugly (though some were) but because the total design had engineered out the organized complexity that makes human settlement work. James C. Scott's concept of 'seeing like a state' describes the failure: the planner's synoptic view necessarily discards the local, tacit, relational knowledge on which the life of a neighborhood depends. Any AI system deployed at the scale of a whole domain risks the identical failure: optimizing what it can measure and obliterating what it cannot.

The bounded experiment vs. the total plan. The contrast between the Unité d'Habitation and Pruitt-Igoe is the most practically useful lesson Le Corbusier's career yields. The same principles, applied at bounded scale with full observability and a genuine ability to correct, produced a building people have loved for seventy years. Applied as a policy doctrine, replicated across whole districts without off-ramps, they produced urban catastrophe. The unit of intervention matters absolutely. An AI system deployed as a bounded experiment is a Unité. The same system deployed as infrastructure across an entire society, woven into hiring and credit and policing and governance simultaneously, is urban renewal—and the error will not be discovered until the concrete has set.

The tyranny of the measurable. Le Corbusier's functionalism was, at its deepest level, an epistemological commitment: what matters can be measured, and what cannot be measured does not matter. The sense of belonging, the accumulated meaning of a familiar street, the informal economy of the corner shop—none of this appeared in his statistics on density and sunlight, and none of it survived his plans. AI systems face the identical limitation at global scale: they can only optimize what they can represent as a number, and everything that cannot be so rendered is, from the system's standpoint, invisible. The things that matter most are frequently those that resist quantification most completely.