Making What AI Cannot: Generative AI and Knowledge Through Making

Abstract

Contemporary industrial designers have increasing access to Generative AI, computer-aided design (CAD), and fused deposition modelling (FDM) 3D printing. These tools can expand the possibilities available to individual designers, but their value cannot be understood only through the outputs they produce. This thesis examines how transferable knowledge can emerge through a practice-based design process that uses these tools, with particular emphasis on Generative AI. The research uses the design and fabrication of a free-standing loudspeaker as a vehicle for inquiry. Grounded in Research through Design and structured through Programmatic Design Research, the project was organised as a sequence of connected engagements involving visual and technical exploration, AI-assisted inquiry, CAD development, material experiments, comparative testing, practitioner feedback, fabrication, assembly, and reflection. Each engagement produced artefacts, observations, and takeaways that informed the direction of those that followed. The project demonstrates that transferable design knowledge was generated through the multiple interactions among these activities and shows how designer judgement emerged as a crucial part of the process, shaping how outputs were interpreted, which assumptions required proving, and which findings were rejected, revised, or carried forward. Generative AI supported exploration, technical questioning, comparison, and reflection, but its outputs were uneven and required further investigation through CAD, physical making, testing, and external feedback. CAD allowed possibilities to become controlled geometry while exposing material and assembly requirements. FDM printing brought these decisions into reality, revealing constraints involving tolerances, print behaviour, and material behaviour. The thesis argues that Programmatic Design Research provides a useful structure for turning these situated experiences into communicable and documented knowledge. By showing how ideas were generated, tested, rejected, revised, and carried between engagements, the research makes the design process open to reflection and extension by others. Its contribution is both the loudspeaker artefact and a transferable account of how contemporary tools, particularly Generative AI, can be investigated through contemporary industrial design practice.