Drawing inspiration from Noam Chomsky's seminal work 'Syntactic Structure,' this competition endeavors to explore the application of linguistic transformational generative grammar concepts within architectural design, facilitated by Artificial Intelligence. The methodology encompasses the selection of three distinct geometric frameworks: Euclidean geometry, representative of Frobel's building blocks; Topological geometry, derived from sophisticated mesh modeling; and traditional hand-drawn sketches. These foundational datasets are subjected to refinement within the Kohya GUI, a Python library tailored for precise adjustments of Stable Diffusion models. Through this process, LoRa (Low-Rank Adaptation) models are meticulously calibrated before integration into the Stable Diffusion pipeline, initiating the transformative process. Central to our investigation is the incorporation of natural materials—specifically mud and terracotta—into the computational framework. These materials infuse the geometric structures with organic characteristics, resulting in 'computational morphophonemic structures.' While inspired by traditional morphophonemics, this concept transcends linguistic boundaries, offering insights into the interplay between form and materiality in diverse physical contexts. Our objective extends beyond mere aesthetic exploration; it is a systematic inquiry into how artificial intelligence perceives and manipulates transformational natural variations. Each iteration represents a meticulous examination of the intricate relationship between technology and the natural world, underpinned by rigorous experimentation and analysis. (Note: All presented images are the outcome of Stable Diffusion processes, with final adjustments executed using Affinity Photo.)

Figure 01a: Multiple transformational structures from Froebel’s datasets generated by Stable Diffusion. The form highlighted in red advances to the subsequent computational morphophonemic phase.

Figure 01b: Computational morphophonemic transformation of Froebel’s dataset utilizing the mud dataset.

Figure 01c: Computational morphophonemic transformation of Froebel’s dataset utilizing the terracotta dataset.

Figure 01d: Computational morphophonemic transformation of Froebel’s dataset utilizing the combined terracotta and mud datasets.

Figure 02a: Dataset showcasing Topological Mesh Modeler (TopMod) transformational structures derived from Stable Diffusion. The form highlighted in red progresses to the subsequent computational morphophonemic phase.

Figure 02b: Computational morphophonemic transformation of the TopMod dataset utilizing the mud dataset.

Figure 02c: Computational morphophonemic transformation of the TopMod dataset utilizing the terracotta dataset.

Figure 02d: Computational morphophonemic transformation of the TopMod dataset utilizing the terracotta dataset.

Figure 03a: Multiple transformational structures from sketch datasets generated by Stable Diffusion. The form highlighted in red advances to the subsequent computational morphophonemic phase.

Figure 03b: Computational morphophonemic transformation of the sketch dataset utilizing the mud dataset.

Figure 03c: Computational morphophonemic transformation of the sketch dataset utilizing the terracotta dataset.

Figure 03d: Computational morphophonemic transformation of the sketch dataset utilizing the combined terracotta and mud datasets.