In the early 19th century, the mathematician Charles Babbage argued that ‘the unerring certainty of machinery’ would solve the problem of human fallibility. He was troubled by the many calculation errors in mathematical tables and dreamed of a future in which calculations would be performed flawlessly by machines. The idea of the computer thus arose from a frustration with human failure.
Just 200 years later, his dream seems closer than ever. Artificial intelligence and Large Language Models (LLMs) are performing incredibly complex calculations at a dizzying pace – calculations that are beyond human capability – and appear to have mathematically cracked the code of language itself. Artificial brains attempting to match and even surpass the human brain and all its potential. AI tools that compete with writers, programmers, graphic designers, advertising agencies, animation studios and even ministerial cabinets in Albania.
So the question arises: if AI is becoming ever better at both knowledge and creation, can machines ever take over the role of humans in shaping the built environment? In other words: is an artificial architect conceivable? For me, as an architect, but also for you as a user, that question is crucial. For the built environment is shaped one way or another, yet at the same time it shapes our lives: how we live, work and coexist. It is precisely therein that its significance lies.
We are not quite there yet. Today, people mainly use AI as a smart assistant for small tasks: drafting a text, creating a summary, understanding a difficult email, ... Even at this very early stage, pressing questions arise. What is the ecological cost of a prompt, and when is it responsible to use AI? What exactly is authenticity, and which things are best left for us to write ourselves? How do we remain competent in writing and thinking, and capable of assessing the work of an AI, if we increasingly outsource our thinking to AIs? How can we be experienced ourselves, without having experienced it ourselves?
Meanwhile, AI is helping not only with texts, but also with programming code and automation. Increasingly, processes are simply left to AI. In the tech world, people even talk about ‘vibe coding’. But who will bear responsibility in a world built on ‘vibes’? Who will still understand how something was designed and can be held to account if things go wrong, to fix it? Perhaps it is time to consciously slow down where AI is speeding everything up. Not to hold technology back, but so that a human with ultimate responsibility can remain involved. Slow AI, then: slowness as a conscious choice, to learn to appreciate our inefficiencies and humanity once again.
On the visual front, AI today is primarily capable of generating images. In our Belgian context too, architects such as Valérie Codesido use AI to create rich, stimulating images, completely different from the push-button AI drivel that floods the internet. But here too, in the skilful use of generative AI, all sorts of questions arise. Whose intellectual property is the final generated image? Who is the author: the architect, the AI developer or the owner of the source data? And are we not in danger of ending up in an echo chamber of pseudo-creativity? But more importantly: is it really okay to upload hundreds of images of architects’ public works to fine-tune an AI? Nobody loses any sleep over the fact that we use those images in our architecture courses. After all, we are shaping unique individuals like ourselves, who will operate on our scale in relation to us and become valuable members of society. AI, on the other hand, is trained for competence expansion — potential without a subject. The system gets better, but no one within the system becomes a person.
At the same time, design is becoming increasingly complex. More and more rules, expectations and constraints are making the search for good solutions more difficult. Any help is therefore welcome. What, then, is still missing from such a potent assistant other than the ability to experience for itself in the world how space feels and what it means to move within it and interact with people? In robotics, too, huge strides are being made today that were unthinkable just a short while ago. Soon, embodied AIs may well experience a shared space with us, thereby achieving an even more complete context — and thus intelligence. Can we then imagine that one day an artificial entity, grounded in sensory experience, might reason about spatiality, order and sequence, design and logic, feeling and beauty, meaning and expectation?
Whatever it becomes: we must draw a line here. An intelligence built on an artificial neural network of artificial neurons and embodied on a robotic platform of plastic and aluminium, sensors and servomotors is, by definition, alien to us and therefore not like us. They are not like us. The intelligence generated within them is therefore not so much an artificial intelligence as an alien intelligence and thus radically different, as the writer Yuval Harari so aptly puts it.
However, the built world has always been intended for humans, not for AIs. Can we really imagine it being designed by non-human entities? Even if the simulation of humanity seems increasingly complete, it remains asymptotic. It approaches, but never touches, the fullness of human experience and meaning.
What is truly lived, experienced and embodied by a human being cannot be achieved through simulation. The imitation can never be the original. That is precisely why it is a dangerous idea to hand over the care of our spaces and buildings.
Architecture is more than a calculation. The architect, as the Greek word suggests, is the ‘arche’ of the ‘tecton’: the origin of what is designed or built. An ‘artificial architect’ is therefore always a contradiction. For what is artificial cannot be original. People fail, that much is certain, but to conclude from this that we must therefore place our trust in ever more comprehensive, non-human systems is a bridge too far and a step we would take only to our detriment.
In Nerdland, Jeroen Baert once dismissed my statement “AI calculates, but a human interprets” as utter nonsense — until, two minutes later, to the hilarity of the entire panel, he found himself reaching exactly the same conclusion. AI generates, but we choose what matters.
Let me be clear: I want to believe in people, in authorship and in powerful AI tools that support us, but not in handing over our responsibility or stewardship of this wonderful world entrusted to us. It is precisely in that care for one another and the world we live in that we will be able to find lasting meaning in every square metre. For an AI calculates, but a human gives meaning.
Paulus Present
28/01/2026