Semiotic Machines: Artificial Text and the Praxis of Reading

Recent discussions about the significance of generative AI technologies, and particularly of Large Language Models like ChatGPT, have tended to revolve around the question of how the new language-generating machines relate to traditionally human-centered categories like cognition, reasoning, and consciousness. This conference will instead take its point of departure from the observation that what these machines actually produce is text. Their activity is therefore inherently semiotic, both at the level of its material-technological conditions (the LLMs are trained on the world’s largest database of social signs) and at the level of its reception by us, the general public of linguistically-capable users. The conference will explore the consequences of this perspectival switch towards semiotics—from questions about the general nature of the organic-mechanical interface to questions about the particular structures of the social sign systems within which this interface is embedded—for our contemporary understanding of text as a linguistic and possibly literary object. It asks about what the new reality of semiotic machines does to inherited definitions of textual artefacts and established reading practices, but also about how humanists, as text-trained scholars, might be uniquely poised to contribute to the analysis of our text-driven present. Areas to be explored include the literary and technological history of artificial text production, intersections of computing technologies and theories of signification, theories and practices of LLM poetics, and the social and political implications of automated writing.

Fig. above: Neural network parameters, visualization generated with Midjourney