Abstract
The rise of Large Language Models (LLMs) has profoundly transformed the landscape of terminology and translation, creating a complex interplay between terminology as a resource for artificial intelligence (AI) and AI as a tool for terminologists. This keynote critically examines this dual relationship, navigating the promises and pitfalls that come with the integration of AI into terminology work.
On one side, terminology plays a crucial role in AI applications, underpinning the quality of outputs in systems such as neural machine translation and automated content generation. At the same time, LLMs are useful tools for terminologists, offering assistance in term extraction, multilingual terminology research and database creation.
Yet, this shift also brings forth ethical questions. The cost of training and deploying LLMs is significant (both in terms of environmental cost and socio-economic impact). Furthermore, LLMs exert a transformative influence on language and terminology themselves, challenging established notions of authorship, originality and the concept of a ‘genuine’ text in an era increasingly shaped by machine-generated texts. This evolution also prompts critical reflection on the preservation of linguistic diversity (not only across languages but also within them), raising concerns about the homogenizing tendencies of AI-driven language production.
Therefore, a critical reflection on the opportunities, risks and ethical concerns surrounding LLM-based terminology work is needed to shape the future of terminology management and specialized translation in a sustainable and responsible manner.