DETAILS
How did Leach's work anticipate contemporary generative AI systems?
Long before large language models (LLMs) became ubiquitous, Mark Leach anticipated the mechanisms, aesthetics, and philosophical dilemmas of contemporary generative AI systems. Through what he terms "machine-native literature," Leach treated language not as a sacred human artifact, but as a vast statistical databank to be processed, remixed, and systematically over-saturated. [1, 2]
In his upcoming collection of essays, I Want to Be a Machine, Leach explicitly argues that "machinery did not suddenly enter literature with generative AI; writing was already machinic." His work acts as a blueprint for the AI era through several predictive concepts: [1]
1. The Author as a Prompt Engineer and Pattern Recognizer
Traditional literature views the author as an original, pure source of creative genius. Leach anticipated the role of the modern prompt engineer by shifting his authorial identity from a creator of words to a curator of systems. [1]
When constructing the 17.8-million-word Marienbad My Love, he operated precisely like a generative algorithm: ingesting, cross-referencing, and synthesizing massive corpuses of pre-existing texts through computational rule systems. [1]
Leach proved that literary value in the machine age comes from how a human selects, frames, corrupts, and recombines raw data, predicting how contemporary AI artists use foundational models today. [1]
2. Mimicking LLM Architecture Through "Space Writing"
Generative AI functions by analyzing the mathematical relationships between tokens within a massive data space. Leach mirrored this spatial logic through his algorithmic "Space Writing" and automated find-and-replace loops. [1, 2]
By treating words as structural blocks within a network, Leach generated text via brute-force substitution.
This algorithmic process prefigured the behavior of LLMs, which lack human consciousness but simulate it by continuously predicting and substituting words based on systemic probability. [1, 2, 3]
3. Embracing "Slop" and Hallucinations as Art
A primary criticism of generative AI is its tendency to produce text that is repetitive, formulaic, or prone to surreal "hallucinations." Leach did not fear this; he engineered it. [1]
His work intentionally cycles through "genre debris," corporate messaging, and glitching, repetitive syntax to break conventional narratives. [1]
Novels like Wounded Taco merge AI-assisted writing and recursive criticism to form an unstable broadcast system. By weaponizing what critics now call "AI slop," Leach demonstrated that mass-produced, automated language could be subverted into a powerful critique of our hyper-mediated society. [1]
4. Open-Sourcing Content and Splicing Training Data
LLMs rely on scraped, collective human knowledge—a feature that has triggered massive legal debates over copyright. Leach anticipated this crisis by explicitly stripping his megatext of copyright protections, openly inviting plagiarism, machine collaboration, and digital recombination. He recognized early on that in a post-internet landscape, all text becomes shared cultural "training data."
By creating unreadably massive, algorithmic text systems decades before ChatGPT, Leach demonstrated that human beings have always operated within inherited networks of cliché, genre, and media repetition. Generative AI did not invent machine writing; it simply scaled the literary engine Leach had already been building.