From Poet’s Pen to AI’s Algorithms: Can ChatGPT’s Translation Parallel Yu Kwang-chung’s Self-translation of Evening?

The rapid evolution of Generative Artificial Intelligence (GenAI) has raised concerns regarding its potential to replace human translators. Recent scholarship has demonstrated the competency of GenAI such as ChatGPT in literary translation, arguing that GenAI can generate high-quality translations of poems and classics with proper prompts (Guan 2024; Farghal and Haider 2024; Hu and Li 2023). However, the ability of GenAI to fully capture the intricacies of languages and cultures remains questionable. GenAI may lack the cultural understanding, creativity, and personal experiences that human translators bring to their work.
To explore this issue, it would be insightful to compare GenAI translation with selftranslation. Self-translation, understood as a process of rewriting (Bassnett 2013) or a hybrid of translation and rewriting (Cordingley 2019), demonstrates how authors/translators modulate their creative projects across different linguistic and cultural systems (Cordingley 2022). Self- translators are often considered to have great liberty and freedom compared to non-authorial translators (Santoyo 2013), allowing them to make deliberate choices to fulfill their translation motivations or purposes. To examine GenAI’s capabilities in handling the complexities of translating across linguistic and cultural boundaries as creatively as human translators, this study attempts to use translations generated by ChatGPT 4o to compare against Yu Kwang- chung’s self-translation of poem Evening. ChatGPT 4o’s translations are generated with three different prompts: a basic translation prompt, a prompt framing ChatGPT as the poetry translator, and a prompt framing ChatGPT as a self-translator.
The comparative reading and textual analysis reveal that while ChatGPT can generate acceptable translations, it cannot render and adjust the poem in a similar manner as the self- translator, even when prompted to do so. The findings underscore the unique characteristics of self-translation and the current limitations of GenAI in capturing the full depth and nuance involved in translating poetry across cultural and linguistic boundaries.
The study demonstrates that despite their impressive capabilities in literary translation, GenAI such as ChatGPT may still lack the cultural understanding, creativity, and personal experiences that human translators, particularly self-translators, bring to their work. The human translator’s ability to modulate and manipulate the translation process, drawing from their intimate knowledge of both languages and cultures, remains a distinct advantage over GenAI’s translations.