In collaboration with Paul Eßer, Maika Dieterich, Julia-Jasmin Bold, Francesco Scheffczyk, Jean Böhm

The un/learn ai publication series emerged from the KITeGG grant-funded project, which explored the use of AI in design education. It documents research, student projects, and symposia from this initiative.

Several factors made a hybrid format the logical choice. First, articles were to be published both online and in print, managed through a central content management system. This also enabled the creation of a pipeline in which LLMs could analyze the articles. This is particularly evident in the cross-references that link semantically similar text sections throughout the entire series. In the second part of the series, we also used a vector embedding model to analyze the occurrence of specific topics in the articles and mapped these onto a graph underlying the text.

The features embedded using LLMs thus do not directly interfere with the text, but rather offer an additional reading perspective and improved searchability of the articles.

Software:

• nuxt
• pubpub
• qdrant
• LlamaIndex

Fonts:

• Bagoss by Displaay
• WhoIs Mono by Raphaël Bastide
• 1. issue: AI generated fonts by typetable.xyz
• 3. issue: Roumald by Erkin Karamemet

Inspiration

• arena channel
• GDP-DA
• article on conceptualizing the series in un/learn 3