Dr Michał Dawid Żmuda from the Institute of Polish Philology and Journalism has received a grant from the National Science Centre (NCN) under the Sonata 20 competition. The awarded project, titled “Machine Techniques of Reading and Writing: How They Create Discourses, Shape Human Subjectivities, and Reconfigure the Human-Machine Divide”, will be carried out at the University of Rzeszów. The total project budget amounts to PLN 283,336.

Dr Żmuda’s research offers a critical perspective on the current debate surrounding chatbots and large language models such as ChatGPT.

“The discussion on this phenomenon often centres on the question of whether the ability to write makes machines more human-like. This comparison is based on the false assumption that our understanding of what it means to be human is culturally fixed. Likewise, writing tends to be treated in this debate as a universal phenomenon. Yet, both the category of the human and the ways of reading and writing are in constant flux. Since language skills have long been used in culture to define human uniqueness, the emergence of writing algorithms will inevitably lead to changes in how we read and write—and ultimately, to a shift in our definition of humanity,” explains the researcher.

Dr Żmuda notes that the process of machines beginning to read and write began as early as the 19th century. His project therefore connects contemporary algorithms such as ChatGPT to earlier inventions, technologies, and machines that engaged in the processing of written text.

“Machines have long influenced both the material and cultural dimensions of writing. Algorithms are part of a broader phenomenon,” Żmuda explains. “I investigate, for instance, how machines slice up material presented to them by humans for processing. This is linked to techniques of translating human writing into machine logic. Such processes were particularly evident in telegraphic practices and in the use of punched cards. I also study office practices involving copying machines. Faxes, scanners, and photocopiers are not typically considered writing machines. But that’s a misconception—humans also read and write by copying texts. Machines do the same, albeit without attempting to understand the content being copied,” emphasises Dr Michał Dawid Żmuda.

In his research, Dr Żmuda also draws attention to the element of randomness in how machines operate, and its influence on human labour.

“Large language models operate by predicting sentence sequences, but this prediction process includes an element of randomness. I compare this situation to the workings of gambling machines—slot machines. After all, a slot machine also generates a ‘text’ that a person then responds to. I also seek out techniques that made people read and write like machines—without thinking about what they were doing. This occurred, for example, in psychomotor experiments using a tachistoscope,” he adds.

The project will explore the cultural aspects of past and present machines in order to address key questions:

- How do these machines alter cultural practices of reading and writing?
- How do they influence the writing practices of people who use them?
- And how do they affect the redefinition of what is considered human and non-human?