Generative AI, Language, and Authorship

Mark Coeckelbergh and David Gunkel discuss what happens to our understanding of language when texts are no longer only produced by human beings but also by large language models (LLMs) that generate impressive-sounding texts in response to prompts from human beings. As they highlight, this also raises the question of what it means to be an author in an age of LLMs, and Coeckelbergh and Gunkel suggest that LLMs should lead us to reconsider our understanding of authorship. In fact, they go further than that – they think that AI should lead us to rethink our understanding of the relation between humans and technology and the relation between human thought and the meaning of texts. They draw on the works of the literary theorist Roland Barthes (famous for his “death of the author” thesis), the continental philosopher Jacques Derrida (and his idea of “deconstruction”), and the historian of ideas Michel Foucault (and in particular his discussion of the “author function”) to paint a picture of the meaning of language and authorship in a world where texts cannot only be produced by human beings with communicative intentions but also by LLMs who may neither understand what they are “saying” nor have any communicative intentions that they desire to convey to the readers of their outputs.