Architecture in the age of artificial intelligence: challenges and limitations to training AI and teaching
Puig Costa, Janina,Navarro Delgado, Isidro,Christodoulou, Marilena,Avellaneda, Omar,Ianni, Manuela
- Published
- 2024
- Language
- sp
Abstract
This article explores the utilization of Al within the field of Architecture education. Al has demonstrated favorable outcomes in information management and computational acceleration. Various studies indicate that generative Al also exhibits potential in the design stage. To investigate this assertion, an activity involving students pursuing a Building Engineering degree was devised, employing diverse Al applications to conceptualize an interior renovation. Through surveys assessing satisfaction levels with each application's outcomes and its subsequent analysis, it is observed that while Al enjoys widespread acceptance and considerable potential, it remains deficient in aspects such as perspective and lighting. This underscores the necessity for students to cultivate discernment in evaluating Al outputs, alongside the ongoing imperative of training machines with high-quality data.
Full text
si..gradL202Y biodigita l_ intelligent_systems iBAG UIC BCN Architecture in the age of artificial intelligence: challenges and limitations to training Al and teaching Janina Puig, Isidro Navarro, Manuela lanni, Marilena Christodoulou, Ornar Fabrisio Avellaneda Universidad Politécnica de Cataluña, España [email protected], [email protected], [email protected], mari lena. christotoulou@u pc.ed u, ornar. fabrisio.avellaneda@u pc.ed u Abstract. This article explores the utilization of Al within the field of Architecture education. Al has demonstrated favorable outcomes in information management and computational acceleration. Various studies indicate that generative Al also exhibits potential in the design stage. To investigate this assertion, an activity involving students pursuing a Building Engineering degree was devised, employing diverse Al applications to conceptualize an interior renovation. Through surveys assessing satisfaction levels with each application's outcomes and its subsequent analysis, it is observed that while Al enjoys widespread acceptance and considerable potential, it remains deficient in aspects such as perspective and lighting. This underscores the necessity for students to cultivate discernment in evaluating Al outputs, alongside the ongoing imperative of training machines with high-quality data. Keywords: Generative Artificial lntelligence, Architecture, Deep Learning, Design, Style 1 Introducción El empleo de la inteligencia artificial (en adelante, IA) en arquitectura ha dado buenos resultados en la automatización de procesos y búsquedas, tarea en la que destacan las aplicaciones conversacionales, de texto a texto, que facilitan la gestión y consulta de datos constructivos (Doukari, 2022). El artículo de Ali, Alhajlah y Kassem {Ali et al., 2022) ya pronosticaba la buena simbiosis entre los modelos tridimensionales informados, los denominados BIM (Building lnformation Modeling), y la emergente inteligencia artificial a la hora de acelerar 61