On GANs, NLP and Architecture: Combining Human and Machine Intelligences for the Generation and Evaluation of Meaningful Designs
Jeffrey Huang, Mikhael Johanes, Frederick Chando Kim, Christina Doumpioti, Georg-Christoph Holz
Abstract
Recent advances in Generative Adversarial Networks (GANs) hold considerable promise in architecture, especially in the early, creative stages of design. However, while GANs are capable of producing infinite numbers of new designs based on a given dataset, the architectural relevance and meaningfulness of the results have been questionable. This paper presents an experimental research method to examine how human and artificial intelligences can inform each other to generate new designs that are culturally and architecturally meaningful. The paper contributes to our understanding of GANs in architecture by describing the nuances of different GAN models (SAGAN vs DCGAN) for the generation of new designs, and the use of Natural Language Processing (NLP) for the conceptual analysis of results.