Tired of this topic yet? Let’s have another look then!
Among the most divisive topics of current day, Gen-AI and its impact on various industries ranks pretty well. Some are welcoming the change, some are skeptical, but if there’s that one industry where no one is left neutral, it’s translation and localization.
As generative AI reshapes global communication, the translation industry faces its second automation wave, as translators wage another fight against progress. This article explores how machines like ChatGPT translation capabilities challenge traditional workflows, redefine specialization, and present new yet tendentious opportunities for human-AI collaboration, for both translators and language service providers, or LSPs.
The First Translators VS Robot War: Human VS Machine Translation
Understand this: as far as translators are concerned, that war, or rather, that forced revolution has already happened, and it was a decade ago. Many professions might be feeling that heat and fear today, but for professional translators, what’s happening now is merely a rematch, or a second phase to that era, but worse. Their sole comfort being to come ready with a hint of “we told you so” to the others.
The translation industry is a battered one, that has suffered a lot of hits right from the Internet revolution already. From the very moment those machine translation services – Google Translate and the likes – started to appear, and to sound appropriate, the discussions, the fears and resentment were exactly the same as they are today, way before the popularity of ChatGPT translation.
Google Translate was going to “destroy jobs”, or even “decimate the industry”, it would be the demise of “proper (your language here)”. Every translator and their mother would ravel at the first mistake, the first sign of weakness on the part of machines, as a proof of how they would never be “good enough”, never have that “creative edge”, because you see, “translation is an art”.
But it is valid that translators hated machine translation so much?
All of this while, of course, using machine translation when they could, if they could. Let me use this opportunity to bring some tough love to my fellows translators and LSPs:
TRANSLATION IS NOT AN ART. Sorry. It’s just not.
Let that sink in a minute if you need, then keep reading.
Now, for the two people still reading who couldn’t get infuriated enough, first, hey, thank you, and second, you probably know what I mean. If you’re lucky enough to be Antoine Galland, getting to translate One Thousand and One Nights to such a degree of derivation that you end up making up your own stories, and those stories in turn become the most well known part of the book itself (Aladdin was not part of the original book)…Well, all respect to you, you’re at a level I will never reach.
But let’s be honest: in this modern day, the large majority of translators don’t even work in anything remotely literary. And even then, communicating the jokes, puns, emotions, or ideas of another talent to a foreign audience is hardly an art. It’s an incredibly thought intensive process, that requires an enormous amount of talent sometimes. But still not an art. And it’s okay. You can still learn guitar on your off time.
Now, about that first Robot VS Translators war.
What did happen, who won? Did the industry get destroyed?
Yes and no. We’ve seen this before. We see this at every automation and every industry-changing progress of any kind. The debate is old, it never changes. Before the mid-2000s, translating anything would necessarily require the services of a human being, and a highly trained one. This would therefore require limited resources allocation, and as a result, companies and individuals would think twice before requiring any type of content to be translated. The cost was high, but demand was also limited.
Machine translation enters the equation. All of a sudden, anybody could get any type of text rendered in a language they can understand, at no cost, in a matter of seconds. Indeed, the immediate perception would be to think that this would destroy a lot of jobs.
First, let’s make an objective statement that everybody hates:
- The result is that the demand for translating services exploded.
- The problem is that it got fulfilled by machine translation. Multilingual communication solutions becomes cheaper, sometimes even for free (or so you think, but that’s another topic).
So, no, the market did not shrink, but it changed, and shifted. It became suddenly possible to get anything casually translated, so long as one wouldn’t expect the highest quality of human expression, but simply having something understandable. Like it or not, that’s a market share for business. A business with machine translation involved.
What this has brought on translators?
All of a sudden, translators would need to become much better at their jobs, in much tougher deadlines, for hardly the same level of prestige as before. They would also need to start to sell themselves differently – or at all for that matter. A very insulting dynamic would also start to pop-up where unscrupulous agencies would expect translators to proofread badly machine translated garbage for half the price, because “you know, it’s already translated, see there’s words and letters and all, how hard can it be“.
There’s no denying the transition has been hard, very hard. As I said, translator is a battered profession. So where is this all going? Who will win the Second Robot War with ChatGPT translation coming in place?
The Second Translators VS Robot War: Human VS AI Translation
Here comes the all-powerful artificial intelligence. AI is not an old concept. It was first introduced in 1956 by John McCarthy at the Dartmouth Conference, where the field was formally established. Recently, AI gained massive attention due to the overnight success of ChatGPT, a language model by OpenAI launched in November 2022. ChatGPT’s ability to generate human-like text and perform tasks like translation sparked global discussions about the potential of AI in global business communication to transform industries.
This has once again shaken the translation industry. Similar to the first wave of machine translation tools like Google Translate, translators now fear job losses and even the erosion of language itself with LLM-based translation. Language schools have seen a 20-30% drop in enrollment for translation courses in 2023, while freelance translators report a 15-25% decline in job opportunities as businesses turn to AI for faster, cheaper solutions.
The translator community is now reflecting on the nature of their work. With the trend of adopting generative AI for content creation, can AI replace human translators? While AI excels at speed and handling repetitive tasks, will all those benefits of AI in translation eventually add value to translators? Will ChatGPT translation become the new normal?
What makes translation so difficult then?
Before we understand the value of AI translation for businesses, let’s take a step back and look at what adds value to a translator. A translators’ duty is to understand and/or empathize with the author or speaker, in order to be true to the original tone. But then, why is this also considered a technical job? That is because, sometimes, the ground work of researching, preparing, and building a knowledge base is more time-consuming than outputting words.
Translation is not about forming the perfect text over and over. This’s the author’s job. It is instead about leveraging two-sided comprehension skills, along with industry knowledge to work effectively. This is why the industry needs “specialized professionals” and “industry experts.”
But these are rare. Generalist translators are driven by a love for the language and passion, but that is not where the money nor industry focus is. Specialized technical translators, however occupy a completely different seat. They are the ones excelling at one field. The focus is an efficient transmission of knowledge, with less attention given to linguistics.
In view of the difficulties involved, why persist, then?
The Role of Technology
A few years back, in 2021, the technological side of translation as a field remained relatively niche among language students, save for the nerdiest, or those with a real drive in technology. Most students were there by lack of other options. At the time, CAT tools were as far as it got for the general public. Trados was king. Memsource had not yet become Phrase, and Smartcat was not prominently advocating AI. ChatGPT translation hasn’t dominate. Custom trained Neural Machine Translation (NMT) was where cutting-edge pioneering was.
While LLM wasn’t yet a buzzword, it was evident that technology was the future. I once jokingly told a friend that we should let ChatGPT translation take over the jobs. I hoped to become an experienced linguist capable of training the world’s best linguistic machines, creating the most valuable AI-powered translation tools, exploring with them the nuances of translation patterns.
Only by letting machines get better can we increase our own output. Technology is here to improve our productivity, free us to focus on what matters. Technology would happen either along with us, thus driven by a balanced approach taking in consideration the limitations of these new tools. Or it would happen despite of us, thus driven by the will of tech-bros to automate anything and everything regardless of common sense and zero regard for the human input. But this will happen.
What makes LLMs, Deepseek, and ChatGPT translation effective?
Fast forward to today, and machine translation with LLM are omnipresent and prominent. With the right approach and clever prompting, LLM’s capabilities can surpass our expectations: A simple sentence like “What’s your name, dear frog?” can, with minimal prompting, be translated into an output that not only is accurate, but also follows the style expected with authentic expressions. Truly impressive.
Despite rumors that LLMs have hit a bottleneck and that high-quality training material will reach exhaustion by 2026, there is no reason to doubt that models will continue to improve, and with lower cost. Just like how the world is surprised with the rise of Deepseek that nobody predicted. Returning to the earlier question: why do a lot of translators still persist in this field? Why do we even try integrating ChatGPT into workflows?
Because we see the potential of combining human wisdom with technological power.
Imagine if, when using ChatGPT for translation, prompts included not only text type and style but also the necessary background knowledge, context, and even term-base and memory matches. With such comprehensive context, wouldn’t we make major gains in time? Of course, some may argue that writing prompts and preparing context takes much longer than the actual translation.
Remember what we said about preparation often taking more time than translation? Proper preparation is essential for translation to be viable. Similarly, AI translation requires “preparation” with proper context to be effective.
The old IBM motto “Garbage In Garbage Out” stands stronger than ever in this age. And those who discard it are about to learn a bitter lesson. The “context” here includes all the key information of the project, such as glossary and translation memory, familiar to our translator friends. And to say the least, preparing context is a unskippable step in ChatGPT translation workflows.
So who will win the war this time?
Machine, or any kind of AI like ChatGPT translation is not flawless; it struggles in understanding and outputting highly complex or intricate texts, especially with highly creative writings. They are, after all, machines, and will never be able to recreate the beauty of human mistakes or emotions. Their ability to make an audience tick on the same emotions in a given era, is and will always be limited. Let alone two audiences from two cultures. The best matching algorithms will only go so far as matching the best sequence in a long database.
It takes a human soul to feel enough of the heat of summer, the horror of war, the beauty of love or the sweet bitterness of a long journey ending. And then to write it down or express it in a new, unseen manner. Regardless of the all-around praises on ChatGPT translation accuracy, one has to take a careful look especially if they’re dealing with highly complicated and nuanced contextual translation with ChatGPT.
The good news is, high-quality linguistic assets can be effectively reused if managed correctly. And it’s applicable for a lot of use cases, such as AI-assisted localization, and translating business documents with AI. Imagine if we have a way to help us effortlessly prepare and leverage those assets along with the eye of linguists, then we will find the ideal balance, and the way to properly exploit new technologies while keeping the proper human touch.
Final Thoughts
True, the rumor remains strong about the replacement of human by generative AI for multilingual communication entirely some time not far in the future. But just by observing how all human revolutions have gone so far, from tractors, to telegrams, cars, planes, Internet… We’ll likely be fine. Stuff will get faster and cheaper, new opportunities will appear where nobody expected them. Maybe.
As for translators and LSPs, I doubt the angle “I Am Human Therefore I Do Better” is going to work this time like it did last time. Fortunately, there’s still a lot to do to improve the profession. Not to mention there’s a lot of ways not discovered yet of generative AI applications in translation. Maybe if human are to co-evolve with technology, then we will all win this time. Let the machines do what they do well, so we can focus on what they can’t.
