Language Machines

Generative AI systems’ disruptive effects include their capacity to generate plausible linguistic content (AI imagined as a ‘writer’) and restrictive modelling of reading and constituent activities as meaning-making, interpretation, and understanding (AI as ‘reader’).

Whether in Generative AIs’ production of health misinformation via inaccurate summarisation techniques, amplification of bias or cliche in text production, machine mistranslations for minority languages, or public misunderstandings about Generative AIs’ facilities, the potential societal harms enacted via such restricted readings are of growing public concern. Pushing back against “technochauvinism”, this project addresses emerging challenges and opportunities associated with Generative AI.

The project looks both forward and back. The project’s historical component aims to recover the central (but occluded) role of the Arts & Humanities in creating and analysing AI. It will map three innovative networks of experimenters working with language, meaning, reading, and machines at crucial moments before, during, and immediately after the advent of digital computing:

1. The network around I. A. Richards and C. K. Ogden who transformed the theory and practice of literary criticism in the 1920s-30s, and pioneered algorithmic thinking about language and semantics, and about language in relation to machines.
2. Bletchley Park’s heterogeneous assemblage of intellectuals, many Arts & Humanities-trained, working on cryptanalysis and language machines, who, under Alan Turing, pioneered the first programmable digital computer and, later, the theory of AI.
3. The groups around Margaret Masterman, the Cambridge Language Research Unit and the Epiphany Philosophers, who were among the earliest adopters of computers to develop machine translation, and whose theoretical advances paved the way for key developments in computing and AI from search engines to vector semantics.

The contemporary component utilises these historical discoveries as the basis:

1. For a series of experiments that analyse Generative AIs’ current reading practices, both to a) develop key findings which articulate and champion the methods, expertise, and practical innovations that the A&H provide, and b) to utilise such innovation in the creative design and deployment of Generative AIs in future
2. To create a glossary of 'AI Keywords' that aims to support better public understanding and evaluation of AI technologies. To the end of improving our understanding, deployment, regulation, and innovation of GAIs, Language Machines anticipates a broad range of beneficiaries: R&D funders and policymakers across industry, higher education, and government; tech sector leaders and innovators; Arts & Humanities academics; archivists; the general public.

The project benefits from robust interdisciplinary approaches to the study of AI and Digital Cultures and their intersections with literature at University of Birmingham via the Centre for Digital Cultures, the FutureThinking Network (housed in the Institute for STEMM in Culture and Society), the Centre for Modernist Cultures, the Institute for Data and AI, and the Transformative Humanities Lab.