What’s Up With Machine Translation?

In his May 30, 2016 article “Who’s Afraid of Machine Translation?” Pieter Beens explores many of the questions, anxieties, and concerns surrounding MT, or machine translation. And, indeed, machine translation has long had the language services community aflutter. “Most translators and interpreters are…wary of machine translation technology,” writes Beens.

Does machine translation warrant wariness?

The bulk of his article details specific advances in machine translation, advances that, until fairly recently, were the stuff of science fiction. These include instant interpretation technologies (e.g., Skype Translator and the wearable device Pilot) as well as Xerox’s translation copy machine, Xerox Easy Translator. Despite the staggering accomplishments of these and other machine translation projects, Beens doubts their ability to supplant human interpreters and translators entirely.

A machine translation therefore can offer benefits, but it doesn’t replace the need for human services. An instant interpreting app is simply not too reliable at the moment, and people in the future will still seek refuge in professional translators and interpreters in critical situations.

Nataly Kelly shares Beens’s skepticism. In her 2014 Huffington Post article “Why Machines Alone Cannot Solve the World’s Translation Problem,”  she highlights the frequent over-optimism of machine translation enthusiasts, remembering how

Sixty years ago this week, scientists at Georgetown and IBM lauded their machine translation “brain,” known as the 701 computer. The “brain” had successfully translated multiple sentences from Russian into English, leading the researchers to confidently claim that translation would be fully handled by machines in “the next few years,”

and how, still,

every few months, starry-eyed and often misinformed journalists herald a new era for language translation, announcing a “groundbreaking milestone” related to a technology that has been around for 60 years.

Linguistic utopia or pipe dream?

As an example of such “starry-eyed…journalism,” Kelly references the 2013 article “How Google Converted Language Translation Into a Problem of Vector Space Mathematics.” Published in the MIT Technology Review, the article celebrates a “team of Google engineers” who uses vector math to

represent an entire language using the relationship between its words. The set of all the relationships, the so-called “language space”, can be thought of as a set of vectors that each point from one word to another. And in recent years, linguists have discovered that it is possible to handle these vectors mathematically. For example, the operation ‘king’ – ‘man’ + ‘woman’ results in a vector that is similar to ‘queen’.

 Despite the article’s jubilant tone, it ends with the Google team’s understated conclusion that “there is still much to be explored.”

Another article from 2013 (the year of linguistic utopianism?), The Economist‘s “Conquering Babel,” covers much of the same territory as MIT’s piece. It claims that we approach the era of “simultaneous translation by computer.”  It asks”how long, then, before automatic simultaneous translation becomes the norm, and all those tedious language lessons at school are declared redundant?” With this rhetorical setup, the author jabs at the language services community with a monologic response: “Not, perhaps, as long as language teachers, interpreters and others who make their living from mutual incomprehension might like.” Still, and again in parallel with MIT’s article, the Economist admits that “some problems remain.”

Machine translation problems

Nataly Kelly’s HuffPo piece narrows in on the “some problems remain” component. She offers six arguments for why “machine translation is not going to replace human translators anytime soon.” These range from the importance of context, to the subjective nature of the translation process. Her final conclusion, even more unequivocal than that of Pieter Beens: “Computers will never fully solve the translation problem” (emphasis mine).

The entire machine translation debate, of course, occurs within the larger debate about the future of artificial intelligence. And, whether it risks rendering human ingenuity obsolete. Some fear machines replacing us all, either in an absolute sense —causing the extinction of the human species— or an economic one (a few technocrats controlling sophisticated AI machines that produce vast abundance and leave the other 99.999% of us in a dystopian hell). Others imagine a future in which humans and machines work together to create a veritable Eden. Most likely, the truth lies somewhere in the middle. Or, off to the side, in some scenario not yet conceived.

The larger AI debate

For me, the most well-reasoned treatment of the artificial intelligence controversy was that of Erik Brynjolfsson and Andrew McAfee in their 2014 book The Second Machine Age. Tending more to the utopian than the dystopian, the authors posit that machines will always, or at least for a very long time, be extremely domain-specific (while far outperforming humans in their respective domains of specialization), a limitation which will always leave room for humans, with their generalist, cross-domain abilities, to work with machines. In this scenario, human-machine collaboration can produce results that far exceed those of either humans or machines.

In the realm of machine translation, the assessment of Brynjolfsson and McAfee holds true. At least for now. Machines translate with speed and accuracy — at least in specific technical genres — but humans serve the important role of quality control, contextual checks, and other post-machine editing functions. And, many industrial sectors (e.g., the life sciences), tend to avoid machine translation due to the highly sensitive and high-liability nature of their content.

We can speculate all we like about the future of machines, machine translation, and artificial intelligence. And the role we humans will play. For now, though,  translation and interpretation processes need our puny, fragile brains. For these brains, despite their shortcomings, have awesome abilities. They possess a dexterity, fluidity, and adaptability that machines have yet to approach.