Abstract
The current architectural design process is a largely manual task completed by a single or group of architects over an extended period of time. Typically, designs may take months or even years to complete and properly prepare for construction. Though software does exist to reduce the most tedious tasks, the majority of the process remains intensively hands-on and time consuming. As a result, there is a huge potential to harness AI’s ability to automate traditionally manual tasks to help reduce production time, increase efficiency, optimize designs, and open doors to new types of architectural creativity.
The primary goal of this research paper is to investigate AI’s capacity to generate new building designs in a specific and assigned architectural style. Instead of taking hours to create a handful of building designs via the traditional method, we intend use the latest neural-network-based algorithms to create thousands within a fraction of the time.
For this study, we used powerful algorithms called W-GANs (Wasserstein Generative Adversarial Networks) to create new architectural designs. In this case, we wanted to generate new designs in the style of Zaha Hadid, one of the most renowned parametric-design oriented architects of recent time. As shown in this paper, her buildings are easily recognizable by her signature sweeping wave-like gestures, elegant curves, and lightweight and playful masses. On the following pages you’ll both see and learn how WGANS have the ability to accurately learn then mimic her style. Once trained, these algorithms are capable of generating thousands of new building designs within extremely short spans of time. In addition, we will demonstrate how our algorithmic approach creates new and emergent styles of its own through the uncovering of previously hidden patterns and phenomena found within her original work.
These powerful deep neural network algorithms begin to challenge the status-quo of current architectural design processes and enhance our own understanding of past architectural work. My research suggests that a closer and more rigorous scientific rationality between architect and artificial intelligent system can be pursued in order to empower the architectural community and reveal the next wave of architectural creativity, analysis and discourse.
