Documents in the Age of Non-Human Agency

Abstract

As documents are inherently tied to humans and their expressions (Lund 2024), facilitating communication across time and space, we might have historically reflected on the implications of inserting a computer or information system between humans. If we envision the human as a handyman engaged in technical engineering, we can comprehend the development of AI, Artificial Intelligence, as part of a much longer history of creating instruments and machines to support human endeavors. These include the printing press, musical instruments, steam engines, typewriters, microphones, and progressing to digital musical instruments that create their own music.<p>

<p>Yet, it remains humans who craft the metal and plastic constituting large computers. The discourse thus far largely concludes that while inserting computers and information systems between humans may alter some characteristics of documents—particularly the means and modes—it does not fundamentally change the concept of a document. The current model sufficiently encompasses digital documents and digital "documentality."<p>

<p>The advent of Large Language Models (LLMs), a form of AI that has seen significant advancements over the past decade, has led to disruptive technological breakthroughs exemplified by ChatGPT and its competitors in the last two years. LLMs are based on a technical concept labelled Transformers, which saw light of day around 2017. The scale of modern computing power that emerged around the same time as Transformers has enabled these literally – large – language models since. These developments have matured to a point where they have been popularized and made accessible to the general public.<p>

<p>This article aims to explore how generative models like LLMs fit into the complementary documentation model and whether documents can still be considered exclusively human expressions.