Natural Language Processing for Creatives

We look at different ways the output of generative text models can be developed by humans before they are presented as art / performance. We cover instances where humans don’t just generate lyrics, they record the songs (AI Song contest), where they don’t just generate scripts, they get actors to perform it (Sunspring, Date Night), they don’t just generate stories, they do readings (Algonory – Shardcore). We do a generative film club where we watch two films written with different models (CharRNN and GPT-3) and compare their strengths and weaknesses as scriptwriters.

This is about encouraging students to take the stance of Human as Collaborator, rather than Human as Curator, and explore ways to take the extra step beyond just presenting the text (or image, or audio) from a model as the final outcome.