With over 7,000 different active languages, the world is brimming with nuanced thoughts and feelings that most people can’t understand. Translation acts as a bridge that connects cultures and people through literature, politics or shared interests. But what happens when the medium for that translation moves away from being a human-led experience?
Thanks to the recent boom in artificial intelligence, the less-than-perfect Google Translate is a thing of the past, with well-trained models like ChatGPT or Meta’s MMT model being able to give accurate and real-time translations between languages. So, do we even need human translators for our books, films and conversation, anymore? In short, yes.
The biggest draw of using AI for translation is how quickly it can do so. Thousands of languages mean there is so much content yet to be translated and more coming every day that it is an uphill and impossible battle to get everything done. As a result, translators, and the companies they work with, must be selective, leading to unconscious bias limiting the work of one language to just what is wanted to be translated. This can be a myriad of reasons for this like too much context lost in translation or objections from the original writer, but the reality is money. It is expensive and time-consuming to translate a book. Publishers will pick books to translate based on how well they think it will sell, based on how the original sells and how well the core themes of the book work in the other language. Getting publishing rights in other countries requires lots of leg work, especially if it is popular in its original language, which is yet another reason why sometimes hard choices have to be made. There is also the bias of some languages being translated over others based on cultural interest, or arguably fetishism, with languages like Japanese having a large number of English translations. Again, this could be soley because of the talent of the original writer (Haruki Murakami), but is another thing that must be considered before commiting to a book translation.
AI can translate significantly faster than a human, and on multiple pieces of text at once, meaning we can gain access to a lot more work in other languages without being limited by the biases of other people.
Being able to generate large numbers of translations is great and proves that there is a place for AI in translation, there is a reason why human translation takes longer and that same reason is AI’s main shortcoming — attention to detail. The challenge with translation is that languages don’t always have a direct equivalent for a certain word or phrase; even if there is, some of the nuance and meaning behind the original will be lost, an issue made even worse with AI thanks to using its own, third language — binary. An AI model won’t really care for this as its translation will technically be correct, even if along the way it has lost a lot of its original meaning. Human-led translation is not as simple as writing the equivalent word down.
Translators will work with both the original writer and their understanding of both languages involved to create a piece of work that best represents the original. This could mean that they might rework entire sections of text as a direct translation might work technically, but lose the context and subtext of the original and thus the actual meaning. This problem is only compounded by the fact that no two people write in the same way. Each person has a specific tone that they have honed in their writing, be it short and snappy sentences, dry wit, or rich dialogue. A translator will tend to work on multiple projects and so must not only understand the tone of each writer, but how to convey the tone and style of that writer into a completely different language. A good translation requires not only a good understanding of the languages, but an understanding of the source material and the person who wrote it.
The biggest sticking issue with using artificial intelligence for translation is how comfortable we feel in it representing our voices. Already we use technology across almost every aspect of our lives, including communication. How much involvement do we have in our conversations if the other person is never actually reading your own words, but the words a computer thinks are the same thing? While it has gotten better, a lot of the time it is quite clear to spot when something is written by an AI, and that is just in one language. There is not enough trust in the current state of AI to write in one language well, so how can we trust it to best represent the work of another language?
At its fundamental level, translation is communication. Technically artificial intelligence can do this, but not to the degree of fluidity and understanding a human can have. It is hard to say just how nuanced artificial intelligence may get at performing translation, and while it is clear that it has broader applications, currently the written word remains best with us.
Words by George Bell
Want more Books content from The Indiependent? Click here