Disrupting Disruptive Innovation with / in Large Language Models

This is a critical (and somewhat parodic) design experiment about the biases and fixations in large language models and how creative practices intersect with AI. I created a set of completion prompts that forces GPT-3 to come up with new hypes and innovations that are not in its dataset. I then created a simple web interface which enables you to enter an industry or field, after which it returns an idea for a disruptive innovation. This idea and its description will be nonsensical, but still reflect the vocabulary and ideas of trends and innovations. The experiment raises many questions (and hopefully some awareness) about creativity, surprise, technical progress and subversion. I discussed these questions in a workshop I led at ThingsCon 2023 and in a workshop we organized at DIS 2023.

Technically, the back-end is a Python-script that calls the OpenAI API. The interface is created using gradio and some HTML-styling. The full code is hosted on Gitlab, so you can try it out for yourself.

Van Der Maden, W., Van Beek, E., Nicenboim, I., Van Der Burg, V., Kun, P., Lomas, J. D., & Kang, E. (2023, July). Towards a Design (Research) Framework with Generative AI (workshop), in Companion Publication of the 2023 ACM Designing Interactive Systems Conference (pp. 107-109)