Inside the Mind of an AI: How to Read Machine-Generated Literature
A talk given by N. Katherine Hayles (Duke University)
With the development of large language models such as GPT-3 that can produce semantically coherent and syntactically correct text, deep questions arise about how we should understand these machine-generated texts. Are they merely stochastic recyclings from the gigabytes of human texts on which the programs have been trained, without intrinsic meanings of their own? Or are they evidence for AI sentience, as Blake Lemoine claimed for Google’s LaMDA? In addition to these techno-philosophical questions, other issues arise from a literary-theoretical perspective. What assumptions should underlie our engagements with these texts? Should we take into account their origin in machine reading and writing, or following the postmodern claims of theorists such as Foucault and Barthes, should we treat them as we would any other (human-produced) text? This talk explores these issues, arguing that we should indeed attend to their machinic origins. It offers four exemplary strategies of interpretation appropriate to machine texts, with samples from GPT-3 transcripts. It also muses on the future of writing, especially in composition classes, in an era when it is increasingly difficult to distinguish between human and machine-generated texts.
- Length: 1:33:04
- Date of Event: March 30, 2023