Can the Human Translator be Replaced with AI?

Let’s say you have a product description, an instruction, or a website that you need to translate into a few different languages. Hiring a translation agency to do it would cost time and money. But there’s an easy shortcut that we’re all too familiar with: Google Translate. A couple of clicks and, voila, international versions of your document are now available. If you don't like Google (who are you??), there's Microsoft, Yandex, or DeepL translator. What connects them? They’re all free and automated, praising themselves for using the newest technologies, the largest databases, and the most sophisticated algorithms.

What could go wrong? Well, the Internet has plenty of examples. [ https://www.babbel.com/en/magazine/15-best-google-translate-fails/]. Even simple phrases can get mistranslated. For example, “Fall Protection” becomes "Otoño Protección" in Spanish, ie “Autumn Protection.” Apparently, we need protection from foliage.

More surprisingly, the machine can even have problems with counting. A chef for the Norwegian Olympic team was trying to order 1,500 eggs for his athletes. Who knows what Google Translator was thinking (was it?) when it translated his message into Korean as 15,000 eggs. [ https://www.thrillist.com/news/nation/olympic-chefs-eggs-google-translate-fail# ]

There’s a straightforward reason for these fails: the machine doesn’t try to understand the context, it just decodes. Every good translator will tell you that to make a good translation they need to understand what is being written. They’ll read the text first and then move on to translating. Otherwise, the translation, even if it is grammatically correct, will end up being inappropriate in tone, potentially offensive, or at the very least will sound unnatural.

Google and other Internet companies are well aware of the shortcomings of their translation services, and they are continually improving them, recently focusing on “machine deep learning.” What does that actually mean? To put it simply, the computer’s understanding of the language improves with the amount of information it processes, and the mistakes become less.

So yes, automated translations are getting better. You can decide they’re good enough for you, especially if you only need an approximate idea of what a document is saying. However, besides still being far from perfect (or even as imperfect as human translators,) machine deep learning has another very serious caveat.

Companies like Google and Microsoft note in their terms of use that they have the right to use your every entry to improve future searches. They also retain the right to use the information you provide in very different ways. That means that entering sensitive legal, medical, or HR information might violate non-disclosure agreements and result in serious fines. And even if this worst case scenario doesn’t come to pass, you will never be able to know exactly where the information you entered is available or to whom it might have been sold. Are you ready to take that risk?

At the end of the day, deep learning, no matter how well it is designed, is probably not on the side of your business.

We still don’t talk to machines as much as we do to other people. And another person can get your message across most accurately. An experienced translator who understands context, style and tone can better represent your brand. Additionally, you will need a trustworthy agency who will keep your information safe and be able to deliver translations in different languages, all of them at a high standard.

That’s exactly what we offer. We have years of experience translating for different industries and into different languages. We gained trust of multiple clients big and small and can deliver the product adapted to your specific needs. To learn more, please contact us at 636 530 1010.