Blog | A collaboration with an A.I. (GPT-2) | 2020
The blog www.furor-poeticus.ch is an experiment and an instrument for self-study. However, it also shows processes of artistic research that have a phenomenon as the object of investigation, which, depending on the field of practice and / or cultural epoch, can be described with the terms “furor poeticus“, “inspiration“, “Dionysian intoxication“, “kairos“, “literary creativity“ , “ [hypo]mania“ or “Schreibrausch“.
Although I know this state from my own experience, I am skeptical of many ideological and cultural-historical as well as psychological explanatory models. Nonetheless, I can safely assume that “my own definition” of “writing intoxication” was ultimately influenced by these concepts. Since I cannot say exactly to what extent the prevailing discourses shape my perception and my self-image about the
topos at the moment, I decided to write about it.
The posts in May 2020 were written togheter with an artificial inteligence. It was the text generator Generative Pretrained Transformer 2 (GPT-2). Similar to the AI chatbot Meena from Google, Facebook‘s chatbot BlenderBot and GPT-3 from OpenAI, the predecessor model of GPT-3 was also able to imitate human language remarkably well. GPT-2 was also trained with huge sample amounts of text from the Internet. There, like all other text generators, GPT-2 learned to imitate anti-semitism, racism, sexism, homophobia and transphobia, etc.
By feeding GPT-2 with scentences from mostly dead authors I tried to write about writing with an automatic text generator but also against its bias learned from texts from the Internet.