LIMITATIONS OF THE USE OF CHATGPT TECHNOLOGY FOR LEARNING TECHNICAL SUBJECTS IN ARCHITECTURE

The integration of digital technologies into education has long sparked debate among educators, often dividing opinion between those who embrace innovation and those who warn against its unintended consequences.The recent emergence of Artificial Intelligence (AI) marks a significant turning point in this ongoing transformation.Following the foundational principles introduced at the 1956 Dartmouth Conference, AI has developed into a set of powerful tools, among which ChatGPT-released by OpenAI in 2022-stands out for its capacity to generate natural language responses, simulate reasoning, and assist in problem-solving tasks across disciplines.This research evaluates the application of ChatGPT as a pedagogical support tool in the context of a technical subject within the Bachelor's Degree in Fundamentals of Architecture at the University of Seville.Conducted within the course "Construccin 5: Rehabilitacin", the study involved 32 students who used ChatGPT to analyze building damage based on photographic evidence and generate construction detail proposals for intervention.The activity was designed to assess AI's performance in three core learning objectives: (1) identifying typical pathologies in existing buildings;(2) formulating a diagnosis based on boundary and material conditions; and (3) proposing technical solutions through graphic and descriptive representation.Methodologically, the study was structured in two phases: first, students submitted building damage images and prompted ChatGPT for a diagnostic explanation; second, they requested the generation of constructive repair details.Student outputs were compared with AI-generated content and evaluated in terms of accuracy, graphical validity, and pedagogical value.The results show that while ChatGPT can support textual analysis and help students frame initial hypotheses, it lacks the precision and graphic reasoning required in construction-based assignments.Most notably, its inability to generate reliable technical drawings underscores its current limitations in architecture education.The findings affirm that, although generative AI offers valuable support for idea generation and autonomous learning, it cannot replace the expertise and judgment of educators in teaching technical subjects related to building rehabilitation.