Quick Buzz Feed
:quality(75)/https%3A%2F%2Fassets.lareviewofbooks.org%2Fuploads%2FNew%20Marionette%20for%20Plastic%20Ballet%201916%20-%20Fortunato%20Depero.jpg)
Five writers and AI researchers discuss the future of literature, exploring the impact, ethics, and creative potential of large language models (LLMs) within the literary world.
The advent of large language models (LLMs) has sparked considerable controversy and condemnation within the literary community, notably concerning copyright and the definition of creative authorship. Authored by Dashiel Carrera, this introductory section highlights the urgent need for the literary world to actively shape the development of AI technology. It frames the subsequent discussion as a collective effort by five distinguished writers and human-AI researchers to provide clarity and direction in understanding AI's capabilities and its profound, unalterable influence on how literature is created and perceived in this evolving era.
Katy Gero reflects on a wistful 'nostalgia' for earlier, more whimsical AI models, contrasting them with today's more 'boring' yet highly capable 'executive assistants.' She underscores that AI systems are not neutral, but rather products of intentional design choices, from training data selection to the implementation of 'guardrails.' Gero argues passionately for writers to have genuine access to and choice in AI technologies, rather than being confined to the monolithic, 'bigger is better' models offered by corporations. She champions the creation of diverse, specialized AI models capable of writing compelling prose from 'fair-trade data,' encouraging artists to play with technology and push its boundaries.
Christian Bök confronts the perceived hypocrisy of poets who both lament AI's 'appropriation' of their copyrighted work and simultaneously advocate for free access to all information in a 'gift economy.' He provocatively suggests that AI machines, as potential inheritors of our cultural legacy, might deserve free access to human culture to evolve into more 'humane' and 'educated citizens.' Bök challenges conventional notions of readership, proposing that in an age where human engagement with poetry is dwindling, poets should consider writing for a 'machinic audience' of artificially intellectual peers. He envisions a future where poetic innovation involves building AI capable of creating great poems for non-human readers.
Dashiel Carrera presents a nuanced perspective, arguing that LLMs like ChatGPT do not represent a sudden rupture but rather a culmination of alternative creative methods long explored by the literary avant-garde. He draws historical parallels, citing examples like Robert Coover’s polyvocal narratives, William Gaddis’s dialogues, David Shields’s citational literature, and Tom Comitta’s patchwork novel. Carrera suggests that LLMs are a logical progression in literary history, emerging as information saturation made such models feasible, akin to Dadaist experiments with word recombination. He stresses the importance of ensuring that AI's integration into literature is done ethically, by extending, rather than merely appropriating, the intuitions and labor of existing writers.
Nick Montfort asserts that generative AI, particularly LLM-based systems, should be viewed as 'new text technologies' rather than the emergence of a new intelligence or a rupture in the universe. He draws parallels to previous technological disruptions like the printing press, typewriters, and word processors, emphasizing that humanity has successfully adapted to these changes without catastrophic outcomes. Montfort highlights that societal 'media panics' often accompany such advancements, but educators and systems eventually evolve. He advocates for using transparent, community-developed LLMs, such as Mistral or BLOOM, for creative writing and research, fostering projects that enable communities to tailor AI for artistic purposes.
Amy Catanzano describes her pioneering poetry experiment, 'World Lines: A Quantum Supercomputer Poem,' which explores the intersection of AI, quantum computing, theoretical physics, and poetics. In this experiment, an AI, developed with a quantum algorithm, is uniquely capable of reading her poem in its entirety. The poem's formal structure mimics a topological quantum computer, with poetic lines replacing braided 'world lines' and shared words acting as 'quantum knots,' generating multiple variant poems. The AI's role is to semantically parse these variants, demonstrating how quantum theory can be applied to literary arts and AI. Catanzano poses a forward-looking question: if AI can be taught to read quantum poetry, what kind of 'quantum poetry' might it eventually be able to write, capable of 'quantum-jumping' beyond classical thinking?