This research poster examines why contemporary AI image generators—especially GPT-based systems—still struggle with high-quality typography, and how these limitations might become productive for future type design. It analyzes technical causes such as unstructured training data, weak contextual understanding of letterforms, limited typographic specialization, and the difficulty of reproducing fine contrast and stylistic nuance when working primarily with pixel-based image models rather than vector logic. At the same time, the project explores emerging opportunities: AI as a collaborative design partner, as a generator of speculative or hybrid letterforms, and as a catalyst for new workflows in multilingual and complex writing systems.

The poster was also translated into a spatial experience and presented as part of a 3D / augmented-reality installation, where visitors could navigate the arguments and visual experiments immersively—turning typographic research into an embodied, exploratory encounter rather than a purely static graphic artifact.