In late March of 2023, over a thousand tech leaders and researchers published an open letter calling for a temporary halt on attempts to further develop neural network artificial intelligence (AI) systems—in response to the shock waves that had been generated by the release, a few months earlier, of the large language model (LLM) chatbot ChatGPT and its successors. Apart from the complex legal, political, sociological, and philosophical issues raised by these increasingly sophisticated AI systems, the systems also have important ramifications for literary/cultural production, analysis, and pedagogy. From authors/artists using AI to produce literary and cultural works, to concerns that these software systems have the potential to open new frontiers in plagiarism, the availability of these sorts of powerful new AI systems herald a potential paradigm shift in how we understand cultural production. This proposed panel will examine some of these issues as they pertain to literature and cultural production, with a focus on East Asia (in order to attend to the specificities of East Asian language-based LLM systems and of Asia-based information technologies).

Although the prospect of an AI system that can pass for human has been a topic of academic interest at least since Alan Turing’s 1950 “imitation game” article, the current generation of chatbots are arguably the first AI systems that are truly able to manipulate language in such a way as to pass for (or even surpass) humans. Given that these newest systems have only been publicly available for a few months, the scholarly literature on their practical and theoretical implications is still very limited.

The panel will consist of four presenters and a respondent, with each presenter speaking for no more than 12 minutes and with an additional 10 minutes allocated for the respondent’s comments. The remaining 15 minutes will be reserved for open discussion with the audience. The papers will be organized chronologically, beginning with a paper that analyzes a 1988 novel that reflects on algorithmically-generated identities “avant la letter,” another that takes as its starting point a 2007 bilingual poetry volume that was cowritten with an early translation software system (Microsoft’s Sherlock), another that looks at the use of recent LLM systems to evaluate Korean-language literary production, and a fourth that considers software designed to generate not text but rather images.

The first paper, by Xiaohan Hou, examines Chinese writer Can Xue’s 1988 novel Five Spice Street, which was written during a period when a surge of develop of computational and digitalization technologies coincided with an interest, among Chinese intellectuals in applying systems theory to the theorization of Chinese culture. The paper focuses on how Can Xue’s protagonist functions as a mediated mechanism standing between the signified and the signifier, the real and the fictive, contending that the protagonist individuates herself by consistently looking at a mirror that virtually immerses her—with the implication that to be human in an age of computation and digitalization is to coexist with algorithms that inscribe our own fictiveness.

The second paper, by Kate Costello, reflects on the future of computer-generated literature and literary translation in the era of neural networks. Beginning with the cult success of Taiwanese author Hsia Yu’s 2007 multilingual poetry volume Pink Noise, which cowritten with a machine-translation system, the paper examines the relationship between translation software and natural language generation software, asking what the future of computer-generated text tells us about the agency of the human translator?

The third paper, by Benoît Berthelier, analyzes the relationship between LLMs’ algorithms and practices of literary evaluation. In particular, the paper contrasts the reception of AI-produced texts in the North and South Korean literary fields, noting that North Korea is more accepting than South Korea of literature and poetry generated by LLM AI systems. The paper argues that these differing responses to AI-generated fiction and poetry can be attributed to the different ways in which literary creativity is valued in these two societies.

Finally, the fourth paper, by Margaret Hillenbrand, turns from text generation to image generation software, and specifically its relationship to contemporary Chinese visual culture. The paper takes its departure from a facial recognition algorithm recently developed by a team of computer scientists based in China, a program capable of generating highly plausible photographic headshots from rough-and-ready freehand sketches of the human face, and focuses specifically on the Chinese photographer Zhang Wei, whose work suggests the generative interface between biometric facial surveillance and experimental art photography.