Architecture's Future In Europe: Deep Learning and Generative Ai

Abstract. Applications based on large language models show a preference for North American cultural references due to the texts and images they were trained on. This can pose a barrier to AI adoption by European architects. Faculty and administrative staff from various departments of the Architecture Degree participated in a workshop where images were generated using textual prompts and sketches. They later completed a survey assessing AI's potential in architecture and higher education. The language used was analyzed, and discussions on cultural biases and European regulations were conducted. This evaluation highlights the importance of emulation in image generation and the semantic limitations of AI. Contrary to previous studies, the faculty did not intend to use AI in the classroom during the creative process. The need to train AI with culturally relevant references to the European context, respecting intellectual property, was emphasized.