Having spent most of his formative years as a design architect, creative programmer and research writer in some of the most unlikely interdisciplinary contexts, Immanuel became increasingly obsessed with artificially intelligent and creative design systems as well as their epistemological, cultural and aesthetic implications. He began experimenting with deep neural networks since 2016 and now directs Artificial-Architecture.
Immanuel holds a joint assistant professorship between the schools of Architecture & Sustainable Design (ASD) and Design & Artificial Intelligence (DAI) at the Singapore University of Technology and Design (SUTD). Current research focuses on architecting and applying machine learning and deep learning models for predictive urbanism, generative architecture and defence intelligence, alongside collaboration with industry, academia and governmental bodies. Courses taught at SUTD include Creative Machine Learning and Artificial & Architectural Intelligence in Design. Previously, he was based at École polytechnique fédérale de Lausanne (EPFL) in Switzerland, doing transdisciplinary research work between the School of Computer Sciences and the Institute of Architecture. His doctoral studies, which was nominated for the EPFL Best Thesis Prize, interrogated the epistemological and formal basis of architecture, by reformulating a new design theory through the conceptual and algorithmic lens of architectural sampling in machine learning. Since graduating from the Architectural Association (AA) London, he has taught at the AA, Royal College of Art (London), Tsinghua University (Beijing), Strelka (Moscow), Angewandte (Vienna), DIA (Bauhaus Dessau), Harvard GSD, UCL Bartlett and many others. His design work has been exhibited internationally, such as at London’s V&A Museum, Shanghai’s 3D Printing Museum and Taipei’s Tittot Glass Art Museum; and published widely, such as in Architectural Design (AD) and Design Computing & Cognition. Immanuel has also practiced as an architect at Zaha Hadid Architects (London), as a programmer at ARUP with Relational Urbanism (London), and as a creative coder at Convergeo (Lausanne) and anOtherArchitect (Berlin). His book Artificial & Architectural Intelligence in Design, published in 2020, interrogates the epistemological implications of AI on architecture, and vice versa.
3d gan HOUSING
Featured at the 17th Venice Architecture Biennale Italian Virtual Pavilion (CityX Venice), this project focuses on the semantic articulation of architectural interiority and exteriority with increased data granularity. Trained with a large dataset of 3D digital models of high-rise residential apartments in Singapore, novel housing configurations are sampled and interpolated from the latent spaces of the 3D generative adversarial networks (GANs).
3d gan chair-architecture
This project, recently exhibited at NeurIPS 2020 (Machine Learning for Creativity and Design), is an exploration of architectural interiority and exteriority using 3D generative adversarial networks (GANs) to produce spatial and formal prompts for creative computational 3D designs across scales and domains. Three different latent spaces (i.e. chairs, buildings and chair+buildings) are used for the sampling and interpolations. Through a seamless computational pipeline, instances of the generated design artefacts are either 3D-printed directly as a whole or assembled with 3D-printed parts.
An ongoing research in appropriating autoregressive machine learning models for discrete architecture. An alternative approach to generative adversarial networks (GANs) that leverages on the discrete conception of parts-to-whole relationship in architectural design. The Discrete-Bofill, Discrete-Corb, Discrete-Mies demonstrate the architectural sampling of Ricado Bofill’s La Muralla Roja (1973), Le Corbusier’s Maison Dom-ino (1914) and Ludwig Mies van der Rohe’s Barcelona Pavilion (1929) respectively.
The project forms part of a research strand in appropriating deep learning within digital humanities, as a critical design reflection on the restoration/augmentation/imagination of cultural heritage. The notion of ‘in-painting’ a presupposed past/future void in time and space is architecturalized as Addition & Alteration Works and Oldify Architecturally. The former samples from a parallel ‘before & after’ latent walk, while the latter samples from a custom generative adversarial network (GAN) by extending recent deep learning models to ‘oldify’/’youngify’ buildings, rather than faces.