Architecture in the Age of Intelligence

Prologue

I graduated in 2026. In the past five years, architecture school taught me how to organize a project and produce drawings that make sense, training me as a professional architect. Somewhere during those same five years, machines learned to do many of these, too. I noticed it gradually. First the rendering, and then the plans. Then the entire scheme, generated in an afternoon.

Where should we go from here? I don’t have the answer yet. I want to ask it properly.

What the Machine Takes

Let me be honest about how most of the designs are made. We look for precedents, save images, pin them to a board, and combine something new with familiar pieces. Learning from precedent is not a problem because most of us do that. But at some points, it becomes a loop — projects borrow from projects that borrowed from other projects.

That loop is exactly what machine learning is good at. It learns patterns from existing works and generates new combinations that look plausible. When a machine produces a convincing plan, it has just shown how much of design was recombination in the first place.

I learned this through my own work. For a research center on the Moon, I defined rules for how rooms could connect and let an algorithm arrange them. For a design in flooded Tokyo, I used a slime mold, a creature living without a brain, to explore paths between towers. In both cases, forms emerged quickly, and both were shaped by what I had fed in, including rules, entities, functions, conditions, clients, and every detail that I imagined.

Machines are becoming capable of performing more tasks that once required human designers. But let’s do a “subtraction”: What remains of architecture if we take away everything that machines can do? This question has been asked before, even if never quite in this form.

History Repeats Itself

The emergence of the architect as a distinct professional role was itself part of a larger transformation. In many traditions, design and construction were closely intertwined in the work of master builders. During the Renaissance and the centuries that followed, drawings and printed books made it possible to separate the idea from the making, and the architect became the person who represented a building without building it personally.

In the nineteenth century, iron and glass came along and transformed construction, and engineers took on specialized roles in structural design. One of the most famous buildings, the Crystal Palace, was designed by Joseph Paxton, a gardener whose experience with greenhouses informed its design. Then modernism arrived. Architects started working with industrialization instead of resisting it, and housing for ordinary people became a central architectural concern.

Not long before I was born, Chinese design institutes underwent a transformation known as “Shuai Tu Ban” (甩图板), which means throwing away the drawing boards. People drawing by hand gave way to computational drafting within a few years. That was only one generation ago.

With each shift, some skills were taken away. What grew afterwards was a changing set of questions about what buildings should be. These transitions were not painless. People lost jobs, and professional roles were redefined.

Our current “subtraction” might be more radical. Many skills emerge at once, including not only technical production but also design itself. History cannot promise this transition will end well. It can only give us a place to begin asking what comes next.

If the Machine Can Do Everything

Let’s consider the hardest case. Imagine a machine that can do everything an architect does today, including site analysis, plans, details, drawings, even client meetings. If AI can replace us, why would we still do this work ourselves?

Painting faced a similar challenge when photography emerged. The camera took over much of the work of recording, but painters did not disappear. They started to explore the experience of light, subjective perception, and abstraction. Painting did not beat the camera. It changed what painting meant.

Sixty years ago, the architect Cedric Price was already wary of technology that arrives before anyone asked what it is for, and he believed buildings should leave room for doubt and change. It is undeniable that machines are remarkably good at answering questions. Give it a goal, and it will find a form that satisfies it. But choosing the goal is not merely a technical problem. It is a decision about what we value and what consequences we are willing to accept.

“Technology is the answer, but what was the question?” (Cedric Price, lecture, 1966)

So the question for this generation is not whether the machine can design. The question is what we choose to do ourselves, and what we are willing to take responsibility for.

IV. Where I Want to Go

Buildings are gradually becoming “intelligent”. They are able to sense conditions and adjust their operation. Sensors can track occupancy, movement, temperature, and energy usage. In a way, buildings are becoming environments where machines can observe human behaviors and learn from data about how spaces are used.

This changes what architects design. Not only walls and rooms, but also the relationships between people and the intelligence around them. What should a building notice, and what should it ignore? When should it adapt to us? When should it leave us alone? Whom should it serve when the goal of its users and its owners are different?

In the 1960s, Gordon Pask imagined a building as a system in conversation with its users. Today, technologies gradually make responsive environments possible. However, the question of how such a conversation should work remains open, and it still needs someone to design its rules and its manners.

When I designed the research base on the Moon, the most important rule was not about form, but about a boundary: the airlock connects the pressurized interior to the exterior. It seems everything could be generated by the algorithm, except this boundary. This is the kind of work that remains — deciding where the interior is, and what it should feel like to be there.

What I Believe for Now

1. Much of what we called design was recombination. The machine just made it visible.
2. I will not build my career on what machines cannot do. That line keeps moving.
3. Every new technology has taken a skill from architecture. None has taken the question of what buildings are for.
4. The machine is good at answers. Choosing the question is still our work.
5. As buildings become intelligent, the architect designs not only spaces, but the relationship between people and the intelligence around them.
6. Someone still has to decide where the inside is, and answer for it.

And one thing I still don’t know:

If one day the machine can choose the question, define the boundaries, or even challenge the goals we give it, where will we go then?