Header: Courtesy of Artificial-Architecture, Singapore University of Technology and Design
AI Sampling Singapore is an award-winning project by Artificial-Architecture at the Singapore University of Technology and Design (SUTD) that explores whether an AI model can generate high-rise buildings that align with local architectural patterns without relying on explicit regulatory control inputs. To achieve this, the project employs a redesigned 3D generative adversarial network (3D-GAN) to explore how deep generative neural networks learn and reinterpret architectural forms.
Trained on a large dataset of 3D digital models of high-rise buildings in Singapore, the AI model generates a range of structures, from recognizable housing typologies to more unconventional compositions that do not conform strictly to existing design norms. The process involves interpolating and extrapolating standard typologies, such as slab and point blocks, to create new spatial configurations that challenge traditional approaches to high-density urban development.
The results were presented at Singapore’s Arts House in multiple formats to show the potential of AI-driven architectural design, from 3D-printed architectural models to large-format digital prints and even in video animations that show the AI’s latent space transformations. The project has also been previously exhibited at the 17th Venice Architecture Biennale’s CITYX Venice Italian Virtual Pavilion, having deeply contributed to the ongoing discussion regarding AI’s role in shaping architecture and architectural discourse.
AI Sampling Singapore has consistently received recognition for its innovative approach, recently winning the Architectural Design – Speculative and Visionary category at the BLT Built Design Awards. The uniqueness lies in how the project demonstrates an alternative method for generating new building designs at scale, which raises questions about how AI can influence architecture: its production, regulation, and creative autonomy.
