ChatGPT and Language Translation

ChatGPT and other Large Language Models (LLMs) have garnered immense attention since the first versions of ChatGPT were released. Language translation, meanwhile, has a storied history of evolving in response to ever-improving Machine Translation (MT). In the interest of comparing this new tool to existing human and Neural Machine Translation (NMT) tools, this study presents a focused examination of translation, comparing the ability of ChatGPT, Google Translate, and professional human translators to translate short English passages into Mandarin. This language pair was chosen to present maximum difficulty, with two very different languages, each with many native speakers and thus plenty of training data. The study’s methodology was designed to ensure a comprehensive and unbiased comparison of methods. Five 250-word English passages were translated into Mandarin by three professional translators, Google Translate, and ChatGPT. Each of the translators were subsequently asked to rate the quality of the other translations on a scale of 0–100 and predict whether the translation was done by a human, ChatGPT, or Google Translate without knowing the origin of the translation. The results indicated that there is no statistically significant difference in the quality of translations among the three methods, underscoring the advanced capabilities of ChatGPT in providing translations comparable to that of professional human translators and Google Translate. Though this study was proof-of-concept in nature due to being limited by its small size (and thus it should not be taken as absolute evidence for the conclusions), it does highlight potential trends in the ever-increasing quality of machine translation. The methodology of the study offers a blueprint for more extensive research with more translators and more language combinations, presenting findings that suggest a narrowing gap between machine and human translation quality. This paper presents the findings of this study, outlines the process for more extensive research, and it hopes to offer insights for translators looking to respond to the ever-increasing abilities of AI translation tools.