This academic project investigates the interface between human architectural ideation and machine learning. By employing Generative Adversarial Networks (GANs), the project examines their utility in generating new visual content, a practice central to architectural design. This research scrutinises the potential of merging GANs with Natural Language Processing to create meaningful designs within a cultural and architectural context.

The project unfolds in two parts: first, a generative phase that includes data preparation and GAN training, aimed at producing diverse images. Second, an analytical phase involves interpreting these images, analyzing them through NLP, and ultimately translating them into concrete architectural designs. In the context of Swiss architecture, the project uses a feedback loop to refine the concept of "Swissness" in architectural design, adjusting the selection of images and design criteria as the machine learning process progresses.