This contribution begins from the observation that large language models (LLMs) are commonly understood through two dominant metaphors. First, LLMs are frequently conceived as tools, a framing that mistakenly attributes to them a pre-industrial character. Second, they are anthropomorphized and described as ‘assistants’, ‘agents’, ‘co-pilots’. The point of departure for this contribution is that LLMs ought not to be understood merely as tools or assistants, but rather as machines.
A machine, however – following, among others, Marx – is not so much a technology as a technically mediated social relation within the production process. This characterization applies a fortiori to LLMs. Hence, just as the classical machine in industrial production generated new forms of inequality within the sphere of production, LLMs likewise engender novel asymmetries within contemporary knowledge production. This paper focuses on one feature of these relations that is crucial for the university and for academic knowledge production more broadly, namely epistemological alienation.
Epistemological alienation refers to the fact that the introduction of machines within the labour process is invariably accompanied by multiple forms of unequal transfers of knowledge between the various actors involved in production. I argue that the introduction of LLMs into academic knowledge production constitutes a radicalization of such transfers, and in doing so, further fuels different forms of epistemological alienation that began with the Industrial Revolution. One of the central conclusions that follows from this argument is that the use of LLMs in academic knowledge production should primarily prompt reflection on the political question of who controls knowledge and who is able to control it, rather than on the narrower question of whether uses of LLMs are ethically legitimate.
Speaker: Thomas Decreus