Maybe Translators' Days Are Numbered After All

But content creators and copy writers had better watch their backs, too!

You hear a lot of talk around the translation industry about the pros and cons of “machine translation” (MT) and, naturally, some speculation that MT may one day put human translators out of a job. You get the usual batch of detractors and believers, and yada yada yada, but I'm starting to think that this whole debate is misguided. It seems to me that automated content generators may be the real disruptive force in the industry.

Machine translation is certainly great for getting the gist of a message, and I’m sure that it’s here to stay and will keep getting better, but I believe the most important, highest quality output will always necessarily need to be human-only. The question is how best to achieve a hybrid solution for content of mid-range importance, where time and cost are a greater factor, and I'm not sure that translation, for multilingual content, is the right way to go.

Everybody focuses on translation because that's been the only feasible way to go for millennia, so it's natural to try to replicate that process with technology. But even the best human translator always has to make choices about what to “lose in translation”. It’s virtually impossible for a person to take a text written by one individual in one language and convert it 100% faithfully into another language and convey exactly the same message with the exact same impact to the audience of that second culture. There's simply too much cultural and psychological and experiential baggage throughout the process for that to happen. All you can ever hope for is the best approximation possible.

And that’s human-to-human-to-human (writer-translator-audience) in a sort of Chinese whispers process. What can the value be in a human-to-machine-to-human-to-human process? Sure, maybe you cut translation time down a bit if the post-editor isn't worried too much about creating the same quality output that could be achieved without the machine middleman, but are the marginal time and cost savings really worth the inevitable drop in quality and the risk of major error?


What I'm starting to think makes more sense is to focus on a machine-to-human-to-human (machine-to-expert-to-audience) process whereby a machine outputs the basic information to be conveyed in each of the languages that it needs to be conveyed in. Then a human, if necessary, gives it that human touch to make it convey that message to the audience even more effectively. It seems to me that it's much easier - and much more effective - for a machine to output decent-quality writing on its own than to try to convert human output from one human language to another.

I don't think translators have any reason to fear machine translation, but I think, in the long run, we do need to keep our eyes on machine-generated content and learn how to adapt our skill sets accordingly. But so should content and copy writers!

In short, I think that content that calls for a skilled, human writer in one language should call for an equally skilled human writer in the other languages. And content that can be output by a machine alone or by a hybrid, machine-first process should be done by the same process in all languages. There will always be exceptions to this, of course, but I can see how they may become outlier cases in the not-so-distant future.


An Analysis Model for Multilingual Content Management


Agile Translating and the Death of Proofreading?