Google Translate will not replace human translators (yet) – words Alexa Wang

The recent AI update to Google Translate is an amazing leap forward in terms of technological innovation. So-called “neural networks” loosely mimic the human brain by “refining data through successive layers of processing units”, translating full sentences at a time (rather than translating words individually).

Fascinatingly, this system, dubbed the Google Neural Machine Translation (or GNMT for short) is also capable breaking down words it doesn’t know into smaller, easier pieces in an effort to teach itself.

But the GNMT still isn’t good enough to top human translators and interpreters, no matter how many fear-mongering articles you may have read on the subject. Because despite the programmers desperately wanting this tech to think like a human, Google Translate will not replace us for some time to come.

Other jobs will probably go first

Of all the jobs currently at risk of automation, translation and interpretation are not considered to be high on that list. According to Fortune magazine, the jobs which are already being automated are in online marketing, financial analysis, and sports reporting just to name just a few.

The Times Higher Education even identified the degrees that are most likely to lead to jobs which could be automated in the near future. Among these degrees were physics, finance and law. These results were based on factors such as whether or not the roles required skills that technology has not (yet) mastered, like social intelligence and creativity; skills which happen to be very important in language-related jobs.

“But”, I hear you say, “that was before the GNMT. What if things have changed?” Well this new Google Translate update wasn’t the first major technological development in languages this year, and the translation and interpretation industry wasn’t fazed by that either.

Google Translate not only lacks contextual knowledge …

When the arrival of the Pilot Earpiece was announced earlier this year, it was hailed as another shining example of science fiction becoming science fact. As the world’s first mechanical, real-time interpreter, comparisons were immediately drawn between it and the babel fish of The Hitchhiker’s Guide to the Galaxy, or Star Trek’s universal translator.

However, translation and interpretation industry experts, London Translations have argued that the earpiece lacked necessary contextual knowledge and industry experience, stating quite plainly that “It’s going to be a long time before technology, Pilot or otherwise, can advance far enough to fill these specific industry needs.”

And in this regard the Google Translate update is much the same. It has no contextual knowledge of cultures or of the professional industries within them. This means that, for instance, it can’t caution you against making any social faux pas, or take into consideration any industry specific jargon which it might encounter in technical documents.

So while the GNMT might prove useful for consumers either on holiday or looking to cheat on their French homework, it won’t provide much additional help to businesses with high-stakes translation needs.

… But it also lacks common sense

Even though the update has led to a marked improvement in its functionality, Google Translate still doesn’t work perfectly.

Google’s own researcher on the project, Quoc Le, has said that the software “doesn’t have a model of how the world actually works yet.” According to The Washington post, it is still capable of mistranslation, dropping words, and when tasked with translating the sentence: “The trophy cannot fit in the cabinet because it’s too big”, the tech cannot tell which is the ‘it’ in question.

It also cannot yet tell how separate sentences relate to each other in a paragraph; much less a whole document.

So it seems that, for the time being, we have some way to go before Google Translate is capable of replacing human industry professionals. Translators everywhere can breathe easy.