For several years now, the translation world has been undergoing a transformation driven by the development of artificial intelligence and its integration into the translation process. Machine learning algorithms and neural networks now help us translate texts with surprising precision and speed. These innovations have also raised a number of questions and challenges. How will AI affect traditional translation methods? What ethical and quality issues should be considered when AI is used for translations? These questions and more are crucial for understanding current and future developments in translation technology. David Možnar, CFO at the Zelenka translation agency - a member of the Czech Society for Artificial Intelligence - has some answers. This interview provides unique insight into the dynamic world of translation in the age of artificial intelligence.

what’s the current role of artificial intelligence in the translation industry?

The development of AI is significantly speeding up the translation process and making work more efficient, thus reducing costs. We primarily use machine translation based on neural networks. This technology enables large volumes of text to be translated quickly and with relative accuracy. However, artificial intelligence cannot substitute human translators—they have an irreplaceable role in preserving context, specifics, subtleties, and linguistic creativity. So, like other translation tools, artificial intelligence is more a professional tool and helper than a competitor to translators.

How the skills and competences for professional translators have changed with the arrival of large language models?

We’re seeing an evolution of their main activity. With the use of machine translation tools, they are no longer producing a new translation, but instead “post-editing,” that is, modifying the machine-translation output. Translators must be able to critically evaluate and adjust machine translation outputs to ensure the final results are correct, understandable, and natural.

In addition, translators need to continually develop their language and topical skills so that they can tackle terminologically or stylistically challenging texts where machine translation cannot provide high quality. And because the technologies are advancing swiftly, they need to be flexible and adaptable so that they can react to market demands and changes.

How can AI support translators’ education and skill development?

Artificial intelligence can support the training and development of translators in various ways - for example by offering feedback, tips, recommendations, and corrections as they go about learning languages or translating.

In the translation environment, it can develop them by pointing out the kinds of flaws in translations that humans are less likely to spot - such as inconsistencies, or bad sentence structure in long sentences. AI can also speed up the research process that is so integral to the work of every good translator. Often this can mean e.g. finding specific information to make a translation fit its intended purpose, audience, field, and the facts on the ground in the target region (country). This activity is often time-consuming, so the help of AI can be very useful.

Can AI help in any way to preserve linguistic nuances and cultural context in translation, or does it “fall short” in this area?

Language models are trained on a large amount of data, containing a variety of language styles, idioms, allusions, metaphors, and other linguistic phenomena. Machine translation is gradually being improved as to its understanding and reproduction of linguistic rules, structures, vocabulary, and meanings. However, it is still not able to fully understand and express linguistic subtleties and nuances, which often depend on human experience, emotions, intuition, and creativity. Machine translation also does not guarantee that texts will not contain errors. For now, artificial intelligence falls short in this area, and human verification and review are always needed.

Where does AI run into limits in sensitive areas like legal or medical translations?

As regards the protection of personal data and intellectual property to avert data leaks and abuse, legal and medical translations are just like any other texts where machine translation is used. Total security for all data must be ensured; this is an absolute priority. There is also no difference in the matter of qualitative standards and the suitability of the texts for machine translation, because high precision, comprehensibility, and consistency in the translated texts are emphasized here as well. In general, we can say that legal documents are well suited to machine translation because they are highly standardised texts. For medical texts, it depends on the specific type of document (e.g. instructions for medical devices, clinical studies, or instruction leaflets). In general - AI is not a good fit for highly terminologically challenging texts.

What are the most frequent misunderstandings that AI causes in the perception of translation services, and how do you deal with them?

The first misunderstanding is that AI is able to fully replace human translators. That is not the case, because machine translations still need reviewing and adjustments to handle precision, context, and linguistic nuance. Translators still play an important role in adaptation, creativity, and reflecting cultural context in translations.

Another example is the assertion that AI is able to translate every text error-free. AI in the form of machine translation still runs into several challenges, such as distinguishing irony, humour, sarcasm, idioms, figurative meaning, and other linguistic phenomena that require a deeper understanding. It can also have difficulties with translating unusual, professional, or new words. Texts that lack context, such as UI (user interface) texts, are a chapter unto themselves. In these cases, the use of machine translation is entirely impractical.

It’s also a myth that AI is able to translate any language you choose. This is because there are languages with limited availability of data and resources. Some languages are better supported than others, and some language pairs are more difficult than others. AI can also have trouble with translating between languages that differ in their syntax, morphology, alphabet or writing direction.

This misunderstanding can be eliminated by fairly informing customers of the possibilities and limitations of AI in translation services and by offering professional machine translation that uses AI and human translators together to achieve optimal results.

What are the greatest prejudices towards AI in the translation sector, and how do you grapple with them?

The main one is that AI is a threat to the profession of translation. Actually, AI is not a competitor, but more a partner that can help translators to increase the productivity, quality, and efficiency of their work, and also enable them to focus on the more complex and creative aspects of translation.

Another false supposition is that AI is simple and unchanging. That is not the case, because neural network-based machine translation is complex and dynamic and constantly changing and learning from new data and feedback. Professional machine translation can also adapt to various client and user needs and requirements and offer personalized and innovative solutions.

To eliminate these misconceptions, it is necessary to raise awareness and trust in AI in the translation sector. It’s important to highlight its benefits and value for translators, clients, and society. This includes for example sharing successful use cases, working with clients clearly and fairly, and supporting dialogue and cooperation among the involved parties.

What new opportunities does AI open up for translation agencies?

The use of AI lets us offer faster and cheaper translation services, and it also allows us to broaden their spectrum. This enables us to meet the needs and expectations of customers from a variety of sectors and regions.