Bridging the Gap Between AI and Reality • Rhodes, Greece
Time: Wednesday, 5.11
Room: Room C
Authors: Ciara Breathnach, Rachel Murphy, Tiziana Margaria
Abstract: In early August 2025 a team of employees at Microsoft and Microsoft Research published an article about the impact of AI as the ‘next candidate general purpose technology’ on various occupations [1]. Tomlinson et al analysed 200k de-identified American chats across the main Gen AI platforms over a nine month period in 2024 to examine how they were used to assist in various work-related tasks commonly associated with particular occupations. |In this paper we adopt a case study approach to explain in greater detail what research-active historians actually do, and how difficult it would be for AI to fully displace or replace them. We use examples from Optical Character and Handwritten Text Recognition (OCR and HTR) technologies to illustrate our points. As agile researchers, teachers and public intellectuals, historians have embraced the digital age over the past forty years and, we contend, are critical to ensuring that any AI training or adoption must have, not only humans, but trained historians in the loop [4], to ensure accuracy, quality outputs and limit potential harms from inherent past gender and racial bias being replicated in and by LLMs. We conclude that certain tasks performed by the historian, like large-scale transcription, could undoubtedly be automated using AI (with the caveat that they are created openly to mitigate against inbuilt biases). Others, like identifying and analysing non-digitised sources, reading against the grain, and interpreting layers of meaning will likely always require the intervention and critical awareness of the human mind. We suggest that of most interest to historians and computer scientists is a third category of tasks that might be augmented but will involve close interdisciplinary co-operation in the design and development of suitable AI tools to ensure successful implementation.