Sorry ChatGPT, it’s nothing personal – but localisation should be done by people, not machines. Cultural sensitivity, a deep understanding of context and finely tuned linguistic expertise are required for high-quality localisation. And chatbots are far from having these capabilities. But will they ever?
An article by Debbie, a human copywriter with a great deal of respect for the machines in her life.
It’s one thing to have ChatGPT translate a text and quite another to localise content for a specific target audience. As the term implies, localisation is the process of rendering content “local”. With optimally localised content, a user would never guess that it was originally created for another market or geography, or in another language. The task requires a nuanced understanding of the culture, context and subtleties in the language of the target audience.
What can AI translators do well?
Like any tool or technology, there is a suitable place for Artificial Intelligence (AI) within the translation process. The key is to recognise its limitations while taking advantage of its powerful capabilities and increasing accuracy.
Machine translation and AI are by no means new in 2023. They have already helped make the process of translation (and of writing copy) much more efficient. Especially where large volumes of repetitive text are involved, the use of AI technology can significantly speed things up and improve consistency across different pieces of content. As a result, reviewers don’t need to spend as much time checking translated work. For straightforward tasks, such as translating general texts and basic documents, AI translators can save a great deal of time and resources.
AI translators also excel at handling multiple languages. As such, they are a versatile tool for multilingual projects. And some AI translators even offer real-time translation capabilities, making them useful for live conversations or quick language assistance.
Progress is also being made in making AI sound more human. Tools with natural language processing (NLP) and deep learning use historical data to approximate tone based on the content type, industry and domain. They rely on neural networks and algorithms to learn what a word means and how to structure sentences in a way that feels natural to readers.
What are some popular AI translation tools?
By now, casual users are familiar with some of the commonly available translation tools on the market. The use of machine translation and AI resources is also widespread in the localisation industry. Here are a few of the more popular AI translators:
Google Translate
Google Translate offers support in a wide range of languages, including many less commonly spoken languages, and can translate both text and speech.
DeepL
Like Google Translate, DeepL offers broad language coverage. It is known for its high-quality translations in European languages and continues to expand its language support over time.
Microsoft Translator
A translation tool with a simple and intuitive interface, Microsoft Translator is easy to use for both text and speech translation. It is integrated into various Microsoft products, which can be convenient if you already work with Microsoft tools.
IBM Watson Language Translator
IBM Watson lets users build and train their own models for specific industries or translation needs. This high degree of customisation allows for better quality translations in specialised areas.
Amazon Translate
Amazon Translate is one of the AI translation tools offering robust API and integration options for developers. This is essential for businesses and developers looking to seamlessly incorporate AI language translation for use in their applications and workflows.
ChatGPT translation
With its language understanding and generation capabilities, ChatGPT is a tool suitable for dynamic, natural-sounding conversations. While there is no ChatGPT translator, the chatbot is good at performing translation tasks in informal or interactive settings.
In addition, there are many easy-to-use apps like Translate AI with multiple features, such as voice-to-voice, image and text translation between dozens of languages. With the ongoing advances in technology, we can expect to see new apps and tools in 2023 and beyond that support multimedia processing.
What is the difference between localisation and translation?
The main objective of translation is to convert text or content from one language to another while preserving the original meaning and intent as faithfully as possible. But localisation is about more than just language. While linguistic accuracy and message communication are important in a translation, localisation goes further to ensure that the content is adapted to the linguistic, cultural, and functional requirements of a specific target audience.
So, in addition to the accurate translation of content, localisation accounts for cultural references, humour and context. And it applies to content in any format, be it text, images or other media, symbols, currencies, units of measurement and so on. The goal here is to create an authentic local user experience for the target audience in their native language, wherever they are and whatever their culture.
Where do AI translators fall short when localising content?
Given AI’s potential to save time and costs in the localisation industry, the hype around it is understandable. But for many reasons, AI is not yet able to replace human resources at this time, in 2023. In fact, it still has a long way to go.
As a rule, AI-based language tools are only as good as the input they have received. The quality of the output depends on the data they have been trained on. AI’s intelligence is informed exclusively by this data. If the data are biased or incomplete, the translations produced by these tools will be similarly flawed. This is why we always need human oversight on translations that may have a high impact or when there is a greater potential risk or liability from mistranslation.
Human communication and speech are informed by intuitive knowledge acquired through universal and individual experiences. We tend to overlook the fact that a great deal of meaning is carried between the lines. In reality, it often has more to do with what has been left out than what’s actually been said. This is why data-driven machine “learning” is unlikely to bring AI localisation up to the level of human capability, no matter how large the language models become.
AI simply does not experience the world as we do and cannot acquire human experience through machine learning. It therefore has no real understanding of the messages we are trying to communicate.
We need to keep in mind that it is artificially “intelligent”, and even calling it intelligent is a stretch. Because AI lacks comprehension, it cannot be relied upon to convey our messages with all their nuances of meaning and subtext. This means that it cannot be entrusted with anything but low-stakes content. Any content that truly matters needs to be processed by a translator with a human mind and language skills.
Is AI, with its simulated competence, destined to languish in the shadow of our intuitive human expertise as a pale imitation of our complex intelligence? Not likely. It is already claiming its rightful and sensible place within our order, where we can continue to work with it and build a collaborative relationship.
