Adam Sebestyen (2025): Designing with Data: A Generative AI-Framework for Early-Stage Architectural Design Using Parametric Models as 3D Training Data, 1st reviewer: Urs Hirschberg, 2nd reviewer: Immanuel Koh
How can artificial intelligence support the early architectural design process? This dissertation focuses on the use of generative AI to develop new three-dimensional forms and spatial ideas that can assist architects in the early design phases.
The lack of suitable 3D datasets for architectural applications, which are essential for training generative AI models, is addressed by taking a new approach: parametric design models are automatically varied and their outputs systematically stored as 3D datasets. These form the basis for training the AI, enabling it to combine different design logics and to generate new architectural geometries. The result is a navigable, digital design tool that allows a wide range of architectural variants to be explored and further developed in a targeted manner. The project “Designing with Data” contributes to an expanded understanding of design, in which AI is a complementary tool rather than a substitute in the creative design process, and opens up new possibilities for exploration and experimentation, especially in the early planning phases.
