Professor Mahmoud Kamel
Artificial intelligence is no longer only a topic of discussion for translation scholars. It is becoming part of the daily work of professional translators. ChatGPT can now help translators produce drafts, polish texts, explain unfamiliar concepts, search for information and suggest alternative formulations. But the more translators use it, the more another question becomes important: can ChatGPT really become a professional translator’s colleague, or does it also create new risks for the profession?
Chapter Six of Translation Studies in the Age of Artificial Intelligence, edited by Sanjun Sun, Kanglong Liu and Riccardo Moratto, examines this question from the perspective of professional translators. Written by Yining Shi, Han Xu, Ho Ling Kwok and Kanglong Liu, the chapter is titled “ChatGPT in Professional Translation: A Double-Edged Sword – Insights from Chinese Translators on Capabilities, Concerns, and Future Prospects.”
Unlike Chapter Five, which looks mainly at students and translation education, this chapter moves into the professional translation workplace. It asks translators themselves how they use ChatGPT, what they think about its quality, what problems they face and how they see its influence on the future of their profession.
The findings are interesting because they do not give a simple answer. Professional translators are clearly attracted to ChatGPT because of its convenience and efficiency. At the same time, they remain concerned about information security, reliability, creativity, cultural meaning and professional ethics. In this sense, ChatGPT is really a double-edged sword.
A study of translators, not students
The researchers used a mixed-methods approach. They combined an online questionnaire with semi-structured interviews in order to understand both the general attitudes of translators and the reasons behind these attitudes. The questionnaire contained 23 questions covering knowledge of ChatGPT, experience with the tool, translation quality, attitudes, concerns and expectations about the future.
The survey included 94 respondents. Most of them were relatively young: about 65 per cent were 30 years old or below. More than 60 per cent had between one and three years of translation experience, while half of the respondents were working in universities or colleges. Importantly, 71 translators, or 75.5 per cent, had already used ChatGPT for translation, while 23 had never used it for this purpose. Six translators were then interviewed in more detail.
This gives the study an important practical value. Instead of asking what AI might do to translators in the future, the researchers asked translators who are already working in the profession what ChatGPT is actually doing in their work.
From translation machine to working assistant
The first interesting finding is how frequently translators are using ChatGPT. More than 60 per cent of the users said that they used it either daily or weekly. About half had been using ChatGPT for more than six months. Some interviewees reported using it for almost all their translation tasks, while others used it mainly to produce a first draft and then carried out post-editing themselves.
Why do they use it?
The main reason is surprisingly simple: convenience. About 73 per cent of respondents identified convenience as a reason for using ChatGPT. Satisfactory output quality came next, at 63.4 per cent, followed by its human-machine interaction at 46.5 per cent.
This interactive element is one of the major differences between ChatGPT and traditional machine translation. A translator can tell ChatGPT to make a translation more formal, adapt it to a particular context or change the way information is presented. The translator can continue the conversation and modify the instructions until a more useful result is produced.
This means that ChatGPT is not being used only as a translation engine. It is becoming a multifunctional tool in the translation process.
What do translators actually ask ChatGPT to do?
The answer is also revealing.
The most common uses are not necessarily translating an entire document from beginning to end. The three most frequently reported purposes were proofreading and polishing translations, explaining unfamiliar concepts or theories to help understand the context, and generating a rough translation. Each of the first two uses was reported by 66.2 per cent of respondents, while 57.7 per cent mentioned generating a rough translation.
In other words, ChatGPT is useful before, during and after translation.
Before translation, a translator may ask it to explain a difficult concept, summarise information or identify terminology. During translation, it can generate a draft or suggest different formulations. After translation, it can proofread and polish the translator’s own work.
One translator explained that ChatGPT can be especially useful when dealing with long and difficult sentences. Under heavy workload or time pressure, it can also be used to create a first draft.
This suggests an important change in the professional workflow. The translator is no longer necessarily starting with a blank page.
Better than machine translation — but not always
The translators generally gave ChatGPT a positive evaluation of its translation quality. About 80 per cent were satisfied or very satisfied with its accuracy and clarity. More than 60 per cent expressed satisfaction with style and tone, consistency, cohesion and coherence, and its adaptability to different domains.
One of its strengths is its ability to break long sentences into shorter sentences that are more suitable for the target language. It can also deal with sentences without explicit subjects and change word order according to context. Some interviewees felt that it performed better in this respect than conventional machine translation systems.
This is important for professional translators because a translation is not simply a word-for-word transfer. A system that can understand enough of the context to reorganise a sentence can save the translator considerable time.
But there is a clear limit.
The lowest satisfaction was found in the handling of idioms and figurative language, where only about half of the respondents were satisfied. Interviewees also pointed out that ChatGPT may work well with business or specialised texts but struggle when the text contains culture-specific concepts or highly figurative language.
This is where the human translator continues to have an important advantage.
When culture becomes a problem
One of the strongest examples in the chapter concerns the Chinese concept of tianren heyi, usually understood as the unity or harmony between human beings and nature.
ChatGPT translated the Chinese character tian, literally “heaven”, in a way that did not capture its philosophical meaning in traditional Chinese culture. The problem was not grammar. The problem was culture.
Another interviewee noted that ChatGPT could face difficulties with political or sensitive texts and sometimes produce wording that does not correspond to official terminology.
This is a very important distinction. A translation may be grammatically correct and still be wrong.
A professional translator does not work only with words. The translator also works with history, culture, institutions, political terminology, social conventions and the expectations of the target reader. ChatGPT may recognize patterns in language, but recognizing a pattern is not always the same as understanding its cultural meaning.
For this reason, the chapter suggests that human checking remains necessary, especially when the text is culturally sensitive, highly specialised or politically important.
The problem of hallucination
Another concern is reliability.
The translators reported that ChatGPT sometimes misunderstands a request, gives inaccurate information or produces inconsistent answers. One interviewee observed that ChatGPT can sometimes become “too intelligent” and translate according to its own understanding rather than the translator’s intended meaning. Another noted that the system can produce invented information.
This creates a special problem for translators.
A human translator usually knows when he or she does not know something. ChatGPT does not always behave in the same way. It may provide an answer even when the answer is uncertain or incorrect.
Therefore, the professional translator cannot simply ask ChatGPT for an answer and accept it. Verification becomes part of the translation task.
The study shows that when ChatGPT’s translation differs from their own, about 80 per cent of respondents prefer to check the answer through other tools or resources. More than 60 per cent also re-check their own translations before making a decision.
The translator, therefore, is becoming not only a translator but also a verifier of machine output.
The hidden cost:information security
Perhaps the most serious concern identified by the study is not translation quality at all. It is information security.
Among the users, 70.4 per cent were concerned about the possibility of information leakage. About 60.6 per cent questioned the reliability of information generated by ChatGPT, while 57.8 per cent were concerned that the use of AI could reduce originality and creativity. Professional ethics was a concern for 45 per cent.
Copyright and confidentiality were also major concerns among the interviewees.
This issue becomes especially important in professional translation because translators often work with confidential business documents, legal materials, government information or unpublished material. In such situations, putting the text into an online AI system may create a risk that is much more serious than a simple translation error.
One interviewee explained that when dealing with confidential documents, even accessing the internet may be forbidden. In contrast, AI tools may be considered acceptable for less sensitive materials, such as advertising or promotional content, depending on the client’s requirements.
So the question is not simply: “Can ChatGPT translate this?”
The professional question is: “Should I put this text into ChatGPT at all?”
A threat or an opportunity?
Here the study becomes particularly interesting.
When translators were asked whether ChatGPT represented a threat or an opportunity for their profession, the answers were almost divided. Among current users, 38 per cent saw it as a threat, while 36.6 per cent considered it an opportunity. About 25 per cent remained neutral.
This shows the complicated position of translators.
They can see that ChatGPT makes their work easier. But they can also see that the same technology may reduce the demand for human translation in some areas.
This concern is particularly serious for new translators. If AI can already produce acceptable translations for routine texts, inexperienced translators may have fewer opportunities to practise and develop their skills. At the same time, deadlines may become shorter and clients may expect more work in less time. Some interviewees even suggested that the widespread use of AI could contribute to increased workloads and pressure on translators.
The future, therefore, may not be equally difficult for all translators. Routine and low-risk translation may become increasingly automated, while human translators may need to move towards more specialised, sensitive and creative work.
The translator as the final gatekeeper
Despite these concerns, there is an important point on which the interviewees were almost united: human translators are still necessary. The reason is quality control.
AI may be able to process a much larger amount of text than a human translator. It may work faster and at a lower cost. But someone still needs to decide whether the final translation is correct, appropriate and suitable for its purpose.
This is why the chapter describes the human translator as a kind of gatekeeper. The machine can generate the text, but the human professional remains responsible for deciding whether that text should be accepted.
This may be one of the most important changes brought by AI.
The translator of the future may spend less time producing the first version and more time evaluating, correcting, adapting and approving it.
Learning to work with AI
The chapter does not suggest that translators should resist this change. Quite the opposite.
Among the current users, 60.6 per cent believed that ChatGPT could be a great help in translation work, while another 33.8 per cent believed it should be used as a supplementary tool rather than as the main resource. Only a very small percentage questioned its usefulness.
This creates a clear message for translator education.
Future translators will need technological skills in addition to linguistic skills. They will need to know how to write effective prompts, how to compare different AI outputs, how to recognise typical AI errors and how to verify information.
But they will also need something more important: professional judgement.
Knowing how to use ChatGPT is not the same as knowing when to trust ChatGPT.
A double-edged sword
The title of Chapter Six is, therefore, appropriate.
ChatGPT can increase efficiency, save time, improve productivity and support translators in many stages of their work. It can act as a first-draft generator, a proofreading assistant, a terminology helper, an information-search tool and even something like an “intelligent dictionary.”
But the same tool can produce inaccurate information, misunderstand cultural references, create inconsistent answers, threaten confidentiality and encourage translators to depend too much on machine-generated language.
The future described by the chapter is, therefore, not one in which ChatGPT replaces the translator. It is a future of human-machine collaboration.
The authors conclude that most translators are willing to integrate ChatGPT into their future work, but they also emphasise the continuing importance of human translators as gatekeepers. They argue that translators need to accept and learn new technologies rather than resist them.
The real lesson of Chapter Six may be very simple.
The translator who ignores AI may find it increasingly difficult to compete. But the translator who trusts AI without checking it may also create serious professional problems.
The successful translator of the future will probably be neither a person who works without technology nor a person who simply follows the machine. The successful translator will be the professional who knows how to make the machine work for him — and knows when to say no to the machine.
ChatGPT is becoming a colleague in the translation workplace. But it is a colleague that still needs supervision.
By Dr Mahmoud Kamel
Professor at the Academy of Arts
