Abstract: In theory, Large Language Models (LLM) ought to be excellent at Latin translation: Latin is a dead language with clear grammatical rules exemplified in Classical texts that are broadly imitated in the Early Modern Period. Yet LLMs, like the computational methods and tools that precede it, find humour, satire, and the nuance of language challenging. Join us as we use a seventeenth-century Latin Problemata to explore the value of expert human translation and the potential (and potential failings) of AI-powered translation, all while drawing attention to the most interesting features of our early modern texts and the challenges they pose.

Please see 'Notes' below for any suggested/official citation information added by the author(s) of this work. Otherwise, you can cite the HSS Commons instance of this publication as follows:

• Dase, K., (2026), "“To Translate is Not to Steal”:Translation, LLMs, and Lodewijk Rouzee’s Latin Problemata", HSSCommons: (DOI: )

  Copy Citation | Export metadata as... | JSON | CSV | RO-Crate | BibTex | EndNote

1. large language models

This talk took place on Thursday, 2 April 2026 online.

Kyle Dase is a researcher working with the Electronic Textual Cultures Lab at the University of Victoria (BC). He was recently a postdoctoral researcher on the STEMMA Project at the University of Galway and a former SSHRC Postdoctoral Fellow at the University of Victoria. His research engages early modern poetry and materiality studies, digital humanities, and the reception of classical friendship. Kyle’s work can be found in Huntington Library Quarterly, Digital Studies, and Digital Medievalist, among other places.