Generative AI playing major role in translation

With generative artificial intelligence tools — such as ChatGPT — being widely embraced in foreign language education and translation tasks, the rapid development of such tools has also presented some challenges, reshaping the sectors.

Nevertheless, some educators, students and translators view generative AI as more of an opportunity rather than a threat.

Wang Huashu, a professor at the graduate school of translation and interpretation at Beijing Foreign Studies University, said generative AI is revolutionizing foreign language education in terms of teaching models, educational resources and language assessments.

Generative AI opens up new possibilities for personalized instruction. "Teachers can use AI to create realistic dialogue scenarios for immersive speaking practice, while students benefit from instant feedback on oral and written language skills, enhancing their communication abilities," said Wang, adding that many universities are producing multilingual learning resources using AI to enrich their curricula.

AI has also enabled innovative approaches in evaluating language proficiency as universities integrate intelligent scoring systems with traditional methods to make assessment more comprehensive and objective.

"We need to balance technological support with the professional judgment of educators to ensure both accuracy and empathy in evaluations," Wang said.

However, there is also an impact.

"The rise of AI necessitates a reevaluation of translators' roles. While AI excels at initial translation tasks, it struggles with nuances like cultural subtleties and contextual understanding. Educators need to focus on students' critical thinking and evaluative skills to ensure high-quality language output," he added.

Citing fears of AI replacing foreign language professionals, Wang advised against succumbing to such fears.

"Translation is not merely a mechanical task but also a process of deep cultural transmission and value reconstruction."

Wang said AI tools are invaluable for bridging linguistic, cultural and physical worlds, but they cannot replicate the poetic, creative and empathetic qualities that human translators possess.

Wang Siyuan, a graduate student in Vietnamese translation at Guangxi University in Nanning, Guangxi Zhuang autonomous region, said AI's current capabilities in Vietnamese translation are limited due to the language's unique qualities.

ChatGPT, which she described as her "cyber teacher", has improved translation efficiency and offered fresh perspectives for academic writing.

Wang Siyuan recalled an experience when the tool helped her understand workflows and terminology in an unfamiliar field during a factory interpretation assignment.

It can produce fluent and accurate translations in certain fields, such as political texts, but it struggles with creative content like literary works, she said.

"The final version of the translation tasks must go through human inspection. I would not directly submit an AI-generated assignment without personal review," she said, highlighting a cautious trust in AI tools and the impossibility of AI replacing human translators outright.

Professionally trained translators hold an edge over AI, particularly in scenarios requiring flexibility and real-time interaction, such as interpretation, Wang Siyuan said.

Human interpreters excel at understanding and adapting to the nuances of interpersonal communication, cultural context and speaker personality. "AI lacks these capabilities, as its answers, rigid and straightforward, are always based on data and algorithms," she said.

Yang Chao, who has been a freelance translator of French for 13 years, said she occasionally uses AI tools like Kimi for translation tasks and is currently in the early stages of exploring their potential.

While AI tools can enhance translation quality, Yang said their performance depends greatly on the clarity of the instructions provided. "Clear and detailed instructions can generate unexpectedly good results," she said, emphasizing the significant advantage of AI tools in information retrieval, which improves translators' research efficiency.

Machine translation post-editing (MTPE), which uses AI-generated translations as a base refined through human expertise, has seen a surge in demand and marked a shift in translators' work models in recent years, Yang said.

However, translation companies often offer price discounts for MTPE, and it has not yet reached the efficiency level that dramatically reduces time and effort, particularly for complex texts, she said.

Those without strong language skills may blindly follow AI-generated translations, hindering their professional development, she added.

As a veteran translator who is aware of the challenges brought by current changes in the sector, Yang said translators should adapt to the trend by learning new AI tools and integrating them into workflows.

Professor Wang said both teachers and students must stay updated with the latest educational technologies, and courses such as AI-assisted translation should be in place to prepare them for future challenges.

Shi Yudie contributed to this story.