Symposium abstract

With the latest developments of AI such as Large Language Models (LLMs), language returns to the center of the stage in critical humanities, history of science and technology, and political economy. Already in the 1990s, political economy advocated for a “linguistic turn” to grasp the transformations of social relations and labour in post-Fordism (Marazzi 1996), but nobody then could foresee the degree of “linguistic automation” that is taking place today through LMMs. Ultimately, in their latent space, LLMs appear to materialise the ‘machine interlingua’ (Liu 2023) that AI practitioners, linguists, and philosophers have envisioned and cultivated since the 1950. As at the times of information theory and cybernetics, a technical paradigm appears to impose a shift in the theoretical discourse. It is urgent, therefore, to investigate the postulates underlying this second ‘linguistic turn’ driven by AI, in which language reemerges at the core of both the technical composition and philosophical concerns.

The symposium Chat Token Vector addresses the new architecture of language, labour, and social relations in what we call AI today. In current AI, language is involved in the making of a complex technosocial scaffolding and a new variant of structuralism. Language is rewritten along the vectors of statistical models in order to become computer-readable. Under this regime of knowledge production, languages but also artefacts such as images become fictitious commodities. The labour that renders these phenomena possible has been, many pointed out, made invisible. Instead of seeing data centres, cable infrastructures, venture capital, and foremost workers cleaning data, maintaining servers, and repairing hardware, we see AI. What are the actual components of the hidden production pipeline of AI? In which way is language represented and mechanised along such a global assembly line?

The format of communication known as Chat has become an interface to access not only AI but also global communication and the labour market. In LLMs, chats orchestrate the new division of labour of the platform economy, in which “work as language” is deconstructed into “microtasks” and all users become “microworkers”. The unit of this formalisation of language is the Token, that grounds the new automation of linguistics and its mobilization for profit (something that French structuralists could not even remotely foresee possible). On AI platforms, atomised individuals talk, unknowingly, to the abyss of the multidimensional space, to the space of the Vector, a novel cultural technique in which collective knowledge and culture collapse into vast statistical manifolds. Given such a complex scaffolding of social and power relations, the symposium invites papers to question the role of language and labour in the new knowledge economy of AI.