Übersetzen von Latein und Altgriechisch mit ChatGPT – Reloaded

Zur Struktur interpretativer Entscheidungen

Abstract

Building on an earlier exploratory study of AI-assisted translation from Latin and Ancient Greek, this article examines how current versions of ChatGPT reshape translation as a guided, dialogical workflow. Focusing on the role of instruction and interaction settings, the paper revisits the same test passages used in the initial study and compares a text-close version with a fluent, reader-oriented translation across different model modes. The case studies – drawn from ancient texts (Cicero, Plutarch) as well as modern scholarly texts written in Latin and Ancient Greek – reveal a stable structural pattern: while text-close translations preserve semantic openness, reader-oriented versions enhance coherence and readability but frequently resolve ambiguity and introduce implicit interpretative commitments. Variations between model modes affect the degree of semantic specification, yet do not alter this underlying functional distinction. AI-assisted translation thus emerges as a heuristic instrument that does not replace philological control, but shifts and reconfigures interpretative decisions already at the stage of translation itself.