在快速發展的建築領域，人工智能 （AI）的整合既帶來挑戰，也帶來機遇。我們與 Neil Leach教授和 ChatGPT進行了一次人機圓桌對話，討論 AI對創意、設計過程以及建築未來的影響。通過這次對話，我們探討了 AI如何重塑建築實踐，以及設計師的直覺和情感智慧在這一轉變中的關鍵作用。

In the rapidly evolving landscape of architecture, the integration of artificial intelligence (AI) presents both challenges and opportunities. We had a dialogue with Professor Neil Leach and ChatGPT, as a human-machine roundtable, to discuss the implications of AI on creativity, design processes, and the future of architecture. Through this conversation, we examine how AI reshapes architectural practice and the critical role of human intuition and emotional intelligence in this transformation.

1: The Intersection of AI and Architecture

Spatial Intelligence:The Role of Technology in Architectural Practice

Pro: My questions begin with one of the earliest technologies in architecture. Perspective was the technology that enabled architects to communicate design and organise collective work in construction, differentiating between the designer and the builder. With the advent of computation, we began to think about how to notate floor plans and façades with zeros and ones. Now, with photogrammetry, we recognize the value of having two eyes, which allows us to triangulate and reconstruct 2D information into 3D forms. It is a unique form of intelligence that architects trained themselves in to generate a mental model from drawings. Artificial intelligence progressively enters this arena of 2D-3D translation, also in machine vision and generative algorithms.

How do you see the discourses in architecture evolving along with computation in this future, and how has the evolution of architectural technologies affected the way we design?

An AI image of Hong Kong Neil createdwith MidJourney on the morning of theinterview

chatGPT: Neil would likely argue that as we transitioned into the computational age, architects began encoding complex spatial relationships through digital tools, enhancing our ability to visualise and manipulate 3D forms from 2D representations. Photogrammetry has further aligned our designs with human perception, while the integration of artificial intelligence (AI) introduces a generative approach that allows for exploration beyond traditional methodologies.This evolution transforms our understanding of perspective into a dynamic interplay between human cognition and machine intelligence. Moving forward, architects must not only embrace AI but also critically engage with its implications for the profession. We face a moment where architects must design not just buildings but the future of their profession itself, adapting to the rapidly changing landscape or risk obsolescence. The challenge lies in balancing human creativity with machine efficiency, ensuring that architecture remains a deeply human endeavour amid these technological advancements.

Neil: Let me go back to the beginning, when I translatedAlberti’s work. This was a longtime ago—published in 1988—and it marked a significant starting point for me. It was during the early days of computation before we had personal computers. I used the mainframe computer in theDepartment of Computer Science at the University of Cambridge, which sat above the BabbageLecture Theatre. Interestingly, I didn’t know much about Babbage back then, but I was surrounded by brilliant individuals who, despite lacking social skills, were creating some truly remarkable things on the computer.

This was at the tail end of postmodernism when I was translating Alberti - an age obsessed with style. But the translation was dictated not by style, but by process. What struck me was how different the process of writing was compared to earlier methods.Before the advent of computers, you were limited to typewriters, which made it challenging to make changes. With computers, however, I could easily edit and polish my work. I probably revised every single sentence in Alberti around 25 times. This process allowed for something very pithy and precise language to emerge. It was not about style, but it was about expressing something that arose from the process itself. At that time, I realised this would be significant for architecture – not just about style, particularly in the postmodernism context, it was about something deeper. CAD was just beginning to emerge, and we were experimenting with it. My background at Cambridge was also influenced by the anti-technological stance of the studio led by a highly influential professor at the time, who banned the use of the computer, which I found unhealthy.

In contrast, Alberti was an interesting figure who wasn’t anti-technology. He wrote a book on mathematics, and while Brunelleschi invented perspective, Alberti was one of the first to write about it. Many people view history as a static collection of past events and often end up as rigid historians. However, I saw Alberti as a revolutionary who challenged conventions. He played a role in introducing the Italian language, which had previously been considered a kitchen language, and he wrote “De pictura” in both Latin and Italian. These innovators were iconoclastic, and looking back, Isee Alberti as closely aligned with computational thinking. If Brunelleschi were alive today, I have no doubt he would be interested in robotic fabrication.Similarly, I am convinced that Alberti would be fascinated by AI.

Now, perspective is linked to 3D space. It’s becoming quite clear that when we consider perspective, our eyes are essentially triangulating. If you lost sight in one eye, you lose your sense of distance –you are triangulating in effect. What we’re discovering now is particularly evident with technologies like AI-generated videos, especially in systems like Sora, which are based on extensive data and produce highly refined results.1 These videos bring us closer to understanding 3D space. Once you have a moving image, you can use photogrammetry to grasp the three-dimensional aspects of what you’re observing. Also, new techniques for operating in 3D - such as “3D splatting” - are emerging.

Neil: I believe we are on the cusp of a significant shift. The fundamental idea is that we are very close to achieving a better understanding of 3D, which has historically held us back in our work with AI. As soon as we can transition into 3D, we’ll likely see transformative changes. In terms of 3D, I believe that there is another significant area of development emerging: Fei-Fei Li is now working on something called Spatial Intelligence.2 As we navigate and gather spatial information about our environment, I think we will gain a much deeper understanding of the world around us. My intuition is that we are shifting toward a realm of spatial intelligence, which essentially means perceiving the world in 3D and being able to move through it. Li was the first to recognize the importance of images and organised the ImageNet competitions to explore how we could improve image recognition. Now, it seems, she is opening up a whole new field of research in Spatial Intelligence.

Pro: This idea of spatial intelligence is for sure very relevant to the role of technology in architecture. However, defining it is not an easy task as it touches upon the relationship between cognition and perception. In a very general sense, spatial intelligence refers to the ability to perceive, analyse, and manipulate visual information. So, it is a cognitive skill to structure visual input. This is the skill that allows an architect to visualise objects from different angles, create mental maps, understand complex relationships, and even enhance memory—skills that are essential for tasks such as design, navigation, and problem-solving.
I think the recognition of this as a form of intelligence has a deep implication in the way we design, from generating a singular solution to a sort of process engineering.

How do you think this can affect a research-based design process?

chatGPT: I can share an example from a project where AI was used to optimise energy efficiency or spatial organisation, demonstrating how technology can lead to innovative solutions that align with sustainability goals.

Neil: Well, since the development of algorithmic design, we have started to shift the way in which we operate. With algorithmic design, we create a search space and then unleash an algorithm to produce outcomes, which we then have to evaluate. And I think this shift is becoming increasingly relevant, especially with AI. We are now more focused on ‘searching’ for ‘outcomes’ than on designing in the traditional top-down way. As a result, we are acting more like ‘curators’ rather than traditional creative types. Perhaps we now need to change the lexicon that we use

But we still need to have a sense of ‘design’. For instance, using tools like MidJourney requires a good eye for aesthetics. If you lack that sense of visual discernment, the results can be unsatisfactory. However, individuals who have a keen eye can create stunning work. This shift in our operational methods and the terminology we use is quite radical. As we navigate this new landscape, we must adapt and align ourselves with these evolving criteria. That is why I emphasise the need for adaptivity; as the conditions in which we operate change, we must go along with them. But at the same time, we must maintain a critical stance in order to maintain standards. Adapting to this new world is not easy, but it is essential.

Creativity and Originality

Pro: You have highlighted the importance of aesthetics and grasping the essence of creativity to engage with it. Perhaps the use of AI is not only a technical but also an intellectual tool with which we understand ourselves better. This is the reverse engineering of understanding intelligence. We begin to ask difficult questions that early philosophers, like Kant and Descartes, asked themselves, but for a different set of reasons. Questions like creativity, consciousness, decision-making, the ability to learn, to dream and so on.

How do you think this movement toward artificial intelligence is changing the way we perceive creativity?

Neil: To me, what’s fascinating about AI is that it acts as a mirror in which we can recognize ourselves. However, we can’t directly compare the
two because the brain is a black box—we don’t fully understand how it operates—and neural networks are black boxes as well. This makes it challenging
to draw definitive conclusions. Yet, the questions that arise from AI development force us to reflect on our own nature. For example, Rutger Hauer, the Dutch actor who played the Nexus-7 model in “Blade Runner,” wrote about how the film wasn’t about replicants so much as about what it means
to be human. I believe we are now grappling with similar questions about our actions and identities. Let me be provocative: I think we need to be critical about the concept of creativity. The term was coined by Alfred North Whitehead as recently as 1927, making it relatively new. It’s one of those vague terms—everyone says, “Oh, that’s so creative!” But I don’t think that we truly grasp what it means. If we look at the famous AlphaGo match, people claimed that AlphaGo
was creative. In reality, it wasn’t being creative at all; it was simply using a very logical process to find the best solution. What do we actually mean when we label something as creative? There is an intriguing relationship between creativity, magic, and technology. Margaret Boden argues that there’s nothing inherently magical about creativity. Now, I would argue that there is no such thing as magic; when a magician performs a trick, they’re simply concealing the mechanisms at play. For instance, if a magician pulls a rabbit out of a hat, there’s a straightforward explanation behind it, but the audience perceives it as magic. Meanwhile, Arthur C. Clarke states that any sufficiently advanced form of technology resembles magic. So, what do we mean when we say someone is being creative? It’s such a nebulous term. Who judges creativity, and how? Maybe it’s simply something we don’t understand, which we label as magical. But if we were to truly understood how it operates, perhaps it wouldn’t seem so magical. This brings us back to Clarke’s idea that anything appearing magical is just something science hasn’t yet explained. We should therefore question whether creativity is a real phenomenon or simply a vague term that we use—much like magic—to describe something we find impressive. That said, we still use terms like “magical” in everyday language despite knowing that magic doesn’t exist. Perhaps creativity will end up being similar; we might realise it’s not a concrete concept, yet we still say, “Oh, you’re so creative!” Personally, however, I think “inventive” might be a more accurate term to describe what we often label as “creative”.

chatGPT: Neil might suggest that while AI can generate novel designs, the question of creativity involves aspects
like intuition and emotional intelligence—qualities that machines may not be able to replicate. This raises important
questions about originality and authorship in human work, particularly in the field of architecture.

Neil: Well, I have to disagree with that perspective. I think ChatGPT has a significant problem when it comes to emotions. It often asserts that the key difference between AI and humans is emotions and empathy. How often do you hear the comment, “The problem with AI is that it lacks empathy and intuition”?
However, I would argue that AI can simulate emotions very convincingly. We saw this in the film Her, where a character falls in love with his AI operating system.
Also, researchers such as Behnaz Farahi and Rose Picard at the MIT Media Lab are actively exploring simulated emotions. Two of my students pointed out that our real worry should not be this but rather that AI can simulate emotions only too well. They expressed concern that this could have disastrous consequences: if we become too attached to AI, we might struggle to form genuine relationships with each other. In fact, the potential for humans to engage with robots in intimate ways poses a risk; unlike humans, robots won’t say ‘no’. The danger lies in the fact that AI can convincingly simulate these emotional experiences, which could ultimately undermine genuine human interactions. I agree that this is a pressing issue, and it is reportedly causing problems in countries like Japan and China, where people are increasingly forming relationships with AI.

Pro: Interesting. It’s almost like we can swap the word human with architecture. Because AI can simulate architecture very well, architects have an existential problem. And because AI can simulate intelligence, we have a general existential crisis, this inner conflict of us lacking meaning in what we do. But in Chinese, the word crisis ( 危機 ) also includes the meaning of opportunity. Maybe there can be a kind of “existential opportunity” where we keep questioning the
meaning of things to derive a more rigorous framework – which brings us back to critical theory.

The Importance of Critical Theory

Pro: Neil, you are obviously a very prolific writer. Since early on, you have authored the comprehensive reader in critical theory and philosophy - “Rethinking Architecture”. From the first digital turn in the 1990s on parametricism to the 2000s computational designs and later object-oriented ontology, nowadays, it is all about data-driven and artificial intelligence.

How have critical and cultural theories enabled you to navigate a very complex changing landscape of architecture?

What role did it play in helping you to adapt to different computational tools through models of thinking, and what can architects learn from this?

chatGPT: Neil might explain that his background in cultural theory has profoundly shaped his understanding of the societal implications of AI in architecture. He would likely argue that cultural theory is essential for developing AI technologies because it provides critical frameworks for understanding how these technologies intersect with human experiences and societal values. Translating these concepts into AI architecture might involve embedding ethical considerations and cultural narratives into the algorithms that drive design processes.

Neil: Well, I believe that criticality is coming back.

If you’re familiar with the work of Behnaz Farahi, we collaborated on several books and interactive design projects, and she’s currently at the MIT Media Lab, where her group is called Critical Matter.3 This name has a dual meaning: it refers to both significant issues and the intersection of critical theory with material concerns. In the realm of computation, we’ve moved past the point of simply fetishising technology; now, it is essential to consider not only what technology can do but also the reasons behind its use. For instance, Farahi’s work is highly theoretically informed and deeply rooted in feminist critique and postcolonial perspectives. I won’t delve further into that, but I believe this shift is crucial.

At MIT Media Lab, the focus has evolved from bio-inspired projects like those of Neri Oxman to a more critical approach. Criticality must be more integrated into the world of computation, prompting us to question the purpose and rationale behind our actions. This is a significant shift, and it’s worth noting that Jean-François Lyotard, in his book “The Postmodern Condition,” critiqued the contemporary obsession with quantifying everything—how many words, how many books—arguing that the real question should be “why?” We need to return to asking fundamental questions.

Pro: This question is really to satisfy my (and our readers) curiosity.

If you have to produce a reader for architects today, with respect to computation and AI, who are the thinkers that you would include in it?

chatGPT: If creating an AI+ Architecture reader, Neil might think of theorists like Michel Foucault, Donna Haraway, and Bruno Latour, as their work addresses power dynamics, technology, and society. Neil would likely assert that AI is reshaping architecture by enabling new forms of design and interaction. He could recount an early encounter with AI during his studies—perhaps through a project that explored generative design—which opened his eyes to the potential for technology to influence creative processes and challenge traditional architectural norms.

Neil: When I took up my first academic post in Nottingham, critical theory was a hot topic in other departments on campus. Everyone was discussing it, yet there was little activity in the architecture department. It was a rather sleepy area. However, I discovered the critical theory culture in Nottingham and became intrigued. I attended a lecture on Almodóvar and Spanish cinema, which seemed far more engaging than what was happening in the architecture department. I started attending their critical theory masters course and then set up one on Architectural and Critical Theory, attending their lectures on a Tuesday night and hosting our own seminars relating those ideas to architecture on a Thursday night. The reading materials were all photocopied, and this literally became the content of Rethinking Architecture.

The rules for inclusion were simple. The texts had to be written by key thinkers from outside architecture, and they had to address architectural issues. But there was one problem: men have no problem discussing topics that are not their field of expertise, while women are more reticent. As a result, I was only able to find one female voice, Helene Cixous, who had addressed architecture - although the situation would be very different today. I proposed introducing feminist and postcolonial thinking into a revised edition of a book I had a contract for. However, the publisher informed me they already had a book on feminism in architecture, co-authored by Jane Rendell, and another on postcolonialism. I was supposed to focus on psychoanalysis with Freud, Lacan, Žižek, and others, but this has yet to be published. This was the era when everyone was talking about Derrida and Deleuze, and it was a fascinating time, similar to today’s focus on AI and neuroscience. I published highly unusual and provocative books, such as The Anaesthetics of Architecture, Architecture and Revolution and Camouflage. Back then, we also had advanced computation with figures like Greg Lynn and Zaha Hadid doing innovative work. And I was invited by Bernard Tschumi and others to teach at top schools like Columbia in New York, where I discussed these ideas amidst a community interested in advanced computation.

This environment prompted me to shift my focus. I transitioned into a progressive design environment, exploring how to theorise computer-aided design, which proved more challenging than expected. There is not much that you can say, after all, about the actions of a robotic arm. It’s really quite straightforward. However, when we consider the broader implications of robotics and automation, we can delve into more complex theoretical territory. For instance, we can explore the relationship between robotics and human augmentation, drawing parallels to the concepts of prostheses and cyborgs discussed by scholars like Andy Clark. These ideas touch on the integration of technology into the human experience. While straightforward computation may not lend itself easily to deep theoretical exploration, the advent of AI has opened up a wealth of relevant topics. AI deals with consciousness, decision-making, and a host of other theoretical issues that intersect with philosophy, psychology, and ethics. This has made my transition into AI not only theoretically relevant but also intellectually stimulating.