During the course of the last two years, we heard numerous claims that machine translation has reached or nearly reached human parity, meaning it got so good that it is essentially indistinguishable from human translation. Those very bold claims might prompt questions about the role of linguists in the future. But let’s not forget that it is still humans that machine translation serves and not the other way round.
It is an easy conclusion to jump to and assume that artificial intelligence (AI), or machine translation (MT), specifically in the context of the localisation industry, puts humans’ status into question. However, upon taking a closer look, it becomes apparent that the AI toolkit, including MT, came into being to support and enhance humans, not to threaten them.
One of the top industry quotes from 2018 cited by Slator, reads:
“Closing the gap between humans and machines is what the market demands, and what the market demands is always what the market gets. Continuing to deny that computers and humans complement one another is regressive and nonsensical. Battling it is futile” — Simon Klys, Freelance Translator / Editor.
However blunt the above words might sound, there are several great examples of how MT can really alleviate some of the pains and stresses in different aspects of human life.
Think for a moment of a time when you were in a country where you did not speak the language at all. How did it feel not to understand a word of what was being said? Probably not very nice.
Free public machine translation platforms like Google Translate come in handy in those moments.
Machine translation, although often imperfect, can help disseminate information and lower the language barriers. It’s not only helpful when you are on holiday and want to know how to get to the nearest train station, but also in more critical contexts such as humanitarian crises, where spreading the right information in the relevant language can literally save lives.
Machine translation certainly contributes to the knowledge economy and often helps to reduce inequality. Lilt, an interactive platform combining machine translation with predictive typing, has built its entire business around the idea of language not being a barrier or a limitation to one’s “ability to learn, grow or support themselves” (https://lilt.com/about).
In a less localisation-focused context, MT allows ordinary people to pursue the things they are passionate about in a more meaningful way. The whole Machine Translation Stories blog talks about the ways MT is applied in everyday life, among others by Internet users who want to understand interviews with their favourite artists, who happen to be native in a language that is not known to the reader.
On the other hand, there are always going to be contexts where human translators will be irreplaceable. The Senior Director of Globalization at GoPro, Sonia Oliveira, said “Our content does not lend itself well to machine translation”, which perfectly reflects the fact that it is always going to be humans deciding when or if to apply MT. The deciding power is in the human hands.
What’s more, let’s not forget that at the very base of MT models lay translations done by humans – the training data on which machine translation systems are built are essentially translation memories, where human translations are stored.
Linguist input and feedback is invaluable not only in training but also improving MT systems. It is from post-editors’ corrections that the MT models learn and improve over time.
Therefore it is not the machine translation technology leading humans; it is the MT technology helping humans to achieve more.
If we approach AI and MT with a healthy mix of optimism and critical thinking, apply the toolkit that they offer where it is relevant and not in a blind or pervasive fashion, then great things can happen.