隨著數碼時代的來臨，人工智能技術逐漸融入建築設計，重新定義建築師的角色。本文以香港大學的建築設計課為例，介紹設計教學如何透過結合不同的AI 程式，讓學生不僅能迅速生成設計概念，還能從AI 意外生成的結果中獲得靈感，展示AI 協助人類設計的可能性。另一方面，AI 可以作為與居民溝通的橋樑，幫助建築師即時且準確地回應使用者需求。

The field of architectural design is experiencing transformative changes with the rise of artificialintelligence (AI). This is an essential topic in the design studio co-taught by Mono Tung and Professor Christian J. Lange at the University of Hong Kong, which explores the potential of AI in reshaping Hong Kong’s public housing estates. As the government will redevelop Choi Hung Estate in Hong Kong, ourstudio investigated how AI can help us reimagine the old housing developments with a new perspective. This process of merging human intelligence with AI assistance offers possibilities for architects to rethink their role in shaping the built environment.
‍
The Choi Hung Estate redevelopment will have 9,200 units, 1,800 more than the current number, with an increase of approximately 24%. Taking inspiration from the French architects Lacaton & Vassal’s work in revitalizing typical residential areas from the 1960s and 70s in Europe, we aim to upcycle the public housing estates in Hong Kong. For example, their architectural design for Grand Parc in Bordeaux involved adding three meters of space to the existing building facades, providing residents with more usable space, improving ventilation and sunlight and thus creating a more livable environment. (01)
‍
The Choi Hung Estate redevelopment will have 9,200 units, 1,800 more than the current number, with an increase of approximately 24%. This studio aims to upcycle public housing estates in Hong Kong, takinginspiration from the work of French architects Lacaton & Vassal, who revitalised typical residential areas from the 1960s and 70s in Europe. An example is the Grand Parc in Bordeaux, which adds three meters of space to the existing building façades, providing residents with more usable space, improving ventilation and sunlight and thus creating a more livable environment. [01]

[01] Before and after the transformation of the Grand Parc, Bordeaux.

The Studio
Our design studio is the first at the University of Hong Kong’s Department of Architecture to apply AI to architectural design. We chose Choi Hung Estate in Kowloon as the research subject to explore how AI can be used to upcycle housing, including AI image generation (such as Midjourney) and text generation (such as ChatGPT). Choi Hung Estate consists of 11 slab-typed blocks with a central elevator lobby that extends into single units on either side. The block has long, narrow corridors with play spaces on each floor. [02] After conducting preliminary research on Choi Hung Estate, students are required to propose design strategies, communicate with AI, and use image generation technology to pursue transformation solutionsfor Choi Hung Estate to increase usable space by an additional 30%.

Traditionally, architects lead the design process. However, the introduction of AI transforms this role, positioning the architect as a curator in a collaborative relationship with technology after it generates thousands of outcomes. In this design studio, we used AI tools to help students envision new possibilities for the redevelopment of Choi Hung Estate, which not only helps speed up the design process but also opens up unexpected creative directions.

[02] Analysis diagram of Choi Hung Estate’s current situation.

The AI-assisted Design Process
The future of architectural practice is undoubtedly intertwined with AI's possibilities. Our students explored how AI could imagine expanding the GFA by 30% of building blocks in Choi Hung Estate. In every tutorial, they pin up new and inspiring ideas with the help of image-generating AI, producing four different design outcomes in less than a minute. The more time invested, the greater the number of outcomes that can be explored. It has shifted the role of architects from designers to decision-makers - selecting the most optimal design out of thousands of generations to present to clients. While modifications and detailed design work remain necessary, AI accelerates design progress significantly. As a result, architectural practice may potentially scale down the design team required for a project. This demonstrates that image-generating AI serves not only as a tool for visualizing concepts but also contributes to problem-solving in practical terms.
‍
One memorable moment occurred when a student asked AI to design a sectional perspective with circulation running through different levels. Rather than 118 HKIA Journal Issue 80 interpreting “circulation” as the movement of people, the AI-generated sectional drawing is spiral-like blood vessels running through different levels of the building. This surprising outcome made the student consider human movement in architecture through a biological perspective, revealing a new way of approaching spatial design. Such moments illustrate that AI can sometimes generate unexpected but meaningful results based on the database sets to interpret their understanding of the same issue. [03]

[03] Design process with input from AI to create a section perspective of vertical circulation through the building.

Challenges and Opportunities
Despite the opportunities, there are still challenges in using AI. One major issue is ensuring that AI-generated outputs are relevant to the context and needs of those who will inhabit these spaces. Architects need to understand the limitations of AI databases, which might not fully capture local cultural or human experiences. For example, when students ask AI to generate Hong Kong public housing, Tong Lau always pops out in the generation. This is due to the lack of images of Hong Kong public housing in the AI database. This is where the architect’s critical role comes in—guiding AI to produce meaningful and culturally appropriate designs. AI does not replace human creativity, but architects must use it with discernment and intentionality.
‍
Our experience with AI at HKU demonstrates that it can go beyond simple design assistance, engaging communities to form a bottom-up architectural design approach. For example, one student in our studio, Kelly Pang, engaged elderly residents of Choi Hung Estate in the design process. She interviewed the residents, transcribed the conversation with AI and asked AI to tidy up the useful material in the transcript. Ultimately, they used AI to create visual representations of potential futures for their homes and asked the residents to choose which images resonated with their desires. Kelly could further converse with them by translating abstract concepts into visualised images. (04, 05)

[04] Over 50 images were created by AI image generation tools as the material for engagement with residents.

[05] Student design of 2D AI-generated images into 3D spaces.

This participatory approach highlights one of AI’s potential strengths: the ability to enhance communication between architects and end-users. AI becomes the bridge between those gaps in an industry where misunderstandings between clients and designers can sometimes lead to dissatisfaction. The residents of Choi Hung Estate, who might otherwise struggle to express how they want their living spaces to be improved, were able to engage more deeply with the design process through AI-generated visuals. Looking ahead, integrating AI into architectural practice presents exciting possibilities and complex questions. On the one hand, AI can help architects rapidly generate design options, allowing more time for creative and strategic thinking. On the other hand, relying on AI tools requires architects to develop new skills—particularly in critically evaluating AI outputs to ensure that the final designs serve the needs of the users and the community. As AI becomes more helpful in the design process, it will challenge architects to stay engaged and adaptable without losing sight of the human-centred aspects fundamental to the discipline.

In conclusion, the ongoing digital transformation of architecture offers a wealth of opportunities for future design. Our work with AI at the University of Hong Kong illustrates how these tools can enhance creativity, drive efficiency, and promote inclusivity in the design process. By rethinking their role and engaging with AI thoughtfully, architects can play a pivotal part in shaping more sustainable, innovative, and liveable urban environments in the digital and AI era.

Mono TUNG is an assistant lecturer in the Department of Architecture at HKU, specializes in computational architectural design, digital modelling and robotic fabrication and writes as a columnist about the latest technologies in architecture.

童昭安從事建築設計，現為香港大學建築系助理講師，專注數碼建築設計，數碼建模及機械製造。同時亦是明報《開眼》建築版的專欄作家之一，關注科技如何影響建築界。

Christian J. LANGE is a registered German architect and Associate Professor (Teaching) in the Department of Architecture at HKU, where he teaches architectural design and advanced digital modelling and robotic fabrication classes.