A Brief Backward History of Automated Eloquence

Late in 2020, two SHL researchers collaborated with CDH to run a workshop series exploring the theory and practice of synthetic text. As a follow-on, in 2021 we will be creating a speculative object: a future textbook about the history of synthetic text, blending fact and fiction. Here’s a pre-print version of one of the sections (since it purports to be written in 2070, possibly by an AI, it is a very “pre” pre-print).


GPT-2 language model (via Max Woolf), inflected with training on feminist manifestos
Eureka machine beats poetry-writing bot by nearly 200 years - Arts and  Humanities Research Council
The Eureka. Image (c) The Alfred Gillett Trust-C&J Clark Ltd. See also Exeter’s Poetry by Numbers project.
A plate from Über den Schachspieler des Herrn von Kempelen und dessen Nachbildung (Leipzig, 1789)

“A (not ethically unproblematic) zombie resurrection of George”

SHL’s James Baker has a fascinating blog post about his use of AI to explore the curatorial voice of historian Mary Dorothy George.

I’ve written this post in the hope that it’ll help others with similar interests take a similar approach to automated text generation, not least as one of my challenges right now is how to read the outputs of simGeorge, how to grapple intellectually as a historian with fabricated catalogue entries in the style of Mary Dorothy George.

Elsewhere: Curatorial Voice: legacy descriptions of art objects and their contemporary use