Abstract
Artificial Intelligence (AI) software has made dramatic improvements in the last few years, going from handy spelling and grammar checking and disturbingly accurate shopping recommendations, to something which can parse large amounts of information and generate reasonable and valuable analysis. Can architects and designers make use of these new tools? As an experiment, students were given a new, generally unfamiliar project type, and tasked with using AI to assist in developing pre-design research, analysis, and concept formation. Students were encouraged to have a “conversation” with the different AI interfaces in order to refine the results, and also to adapt those results using their own judgement. Primarily verbal tasks such as asking what sorts of spaces and activities are typically associated with their project type were accomplished quite successfully using ChatGPT. Students were able to develop their architectural programs easily and probably more thoroughly than if they’d worked on their own. Other activities, such as developing a visual concept using Adobe Firefly, were less successful, with results more unpredictable and difficult to manipulate. Analysis of student surveys and also the work produced leads to the conclusion that AI is a valuable tool for verbal design problems, but less useful for developing design solutions, even at the conceptual phase. This is probably due to the majority of sources used in AI “training” are verbal in nature. The design process, however, often takes verbal instructions to generate 2- and 3-dimensional elements, and relies strongly on aesthetic judgment and spatial ordering.
Presenters
Andrew BrodyProfessor, Architectural + Design Studies, Endicott College, Massachusetts, United States
Details
Presentation Type
Paper Presentation in a Themed Session
Theme
KEYWORDS
Artificial Intelligence,Studio,Programming,Concepts,Students,Analysis,Architectual Program
