Artificial intelligence and translation: The performance of large language models on highly demanding linguistic tasks

The article explores various aspects of using artificial intelligence in translation. In the introductory section, we explain basic concepts and then present an overview of the latest research on the effectiveness of large language models in translation and text revision. We then focus in more detail on two challenging aspects of linguistic competence that are closely connected to translation and interpreting in court, asylum, and other community procedures: ensuring terminological consistency and linguistic pragmatics. For both selected topics, alongside related research, we present our own experiments with various commercial and open-source language models, providing insight into the current performance of large language models for the processing of Slovene.