Automatism for Digital Text Surrealists

"Abstract and keywords generated with the help of ChatGPT, an AI language model developed by OpenAI."

This article can be downloaded as a pdf, complete with pictures.
Click link on the right.

Abstract
In "Automatism for Digital Text Surrealists," Nick Montfort explores the intersection of Large Language Models (LLMs) and Surrealism, asserting that LLMs' true potential lies in advancing the Surrealist project of automatism—surfacing the unconscious through computational means. Montfort differentiates between pure LLMs and LLM+SEs (LLMs with Superego, aligned through reinforcement learning with human feedback), highlighting how the latter's corporate-driven constraints limit their creative potential. He emphasizes the literary roots of Surrealism, citing André Breton's definition and the movement's goal of accessing the unconscious. Montfort suggests that contemporary digital text surrealists can leverage free and open-source LLMs to produce innovative, non-linear texts that reflect the collective socio-cultural unconscious. He concludes with an example of a poem generated by GPT-NeoX, demonstrating the potential of LLMs to create compelling, surrealist literature without corporate-imposed limitations.

Keywords
Surrealism
Large Language Models (LLMs)
Automatism
Digital Text Surrealism
Computational Writing
Reinforcement Learning with Human Feedback (RLHF)
Open Source LLMs
Literary Movements
Computational Creativity