Exploration of Conceptual Design Generation Based on the Deep Learning Model – Discussing the Application of AI Generator to the Preliminary Architectural Design Process

The rapid advancement and heightened awareness of Artificial Intelligence (AI) have stimulated a substantial surge in the growth of deep learning models. This has led to an increasingly prevalent debate surrounding the possibility of AI replacing human architects. Nevertheless, it is foreseeable that architects will explore an AI-aided design process at this stage. Several deep learning models, including NovelAi, DELL-E·2, Midjourney, and Stable Diffusion, have emerged, making it easier to generate images efficiently without the need for multidisciplinary knowledge related to algorithms and programming. During the conceptual design phase of an architectural project, it is crucial for architects to present several massing proposals in various styles within a limited timeframe. It means a huge amount of modelling and drawing works. This paper focuses on Stable Diffusion, Midjourney, and DALL-E 2, as primary examples to discuss the approaches used in the preliminary design process as a smart assistant. These AI platforms are expected to optimize the conceptual design work by reducing the time of transforming hand-drawn sketches into rendering photos and enhancing the visualization of massing diagrams. This article analyzes the impact of AI work-related activities carried out by architects and architectural students based on a designed survey containing various images generated by AI programs. The survey aims to investigate the performance of several popular AI programs and architects’ perspectives towards AI. The findings indicate that AI has the potential to assist human architects to some extent with satisfactory performance. The effective application of AI generators can significantly optimize the design process, allowing architects to explore more creative and aesthetic aspects.