Microsoft's translation treason

What's at stake when robots can translate language as well as humans

AI vs. humans.
(Image credit: iStock)

Years ago, when I was still in school, a professor said something that I never forgot: All translation is treason. I remember it because it sounded so inflammatory. The point, however, was not that one betrays someone or something in translating from one language to another — it's that, inevitably, all translation entails a loss, as meaning in one tongue never quite makes it over intact to another. The treason is the impossibility of translating perfectly.

What are we to make of the fact, then, that a team of Microsoft researchers claim they can now use artificial intelligence to translate from Chinese to English as well as a human? In the company's labs in Asia and North America, the group tested their translation tools on a standard set of news stories, and then compared the results to bilingual human evaluators. The AI matched the humans in accuracy.

What is so surprising is not that artificial intelligence has gotten better — it's the pace at which it has improved. It can now even match humans. And with that, we have to confront the reality that things like translation which we once considered to be deeply human may soon be done as well or even better by computers.

Subscribe to The Week

Escape your echo chamber. Get the facts behind the news, plus analysis from multiple perspectives.

SUBSCRIBE & SAVE
https://cdn.mos.cms.futurecdn.net/flexiimages/jacafc5zvs1692883516.jpg

Sign up for The Week's Free Newsletters

From our morning news briefing to a weekly Good News Newsletter, get the best of The Week delivered directly to your inbox.

From our morning news briefing to a weekly Good News Newsletter, get the best of The Week delivered directly to your inbox.

Sign up

The dream of computerized translation is an old one, but it's perhaps most famously captured in fiction by the babelfish, a translation "device" in The Hitchhiker's Guide to the Galaxy that is actually a small creature that sits in the ear. The ideal, however, is one predicated less on a desire for fancy tech than it is about understanding: skipping over the difficulty of translating and just allowing people to speak to each other. It's a naïve dream, of course; the treason of translation is that even when we comprehend the words we use with each other, mutual understanding can still elude us. But it's also that fuzzy, difficult, subjective part of being human that we always assumed would elude computers and AI.

Perhaps that isn't the case, however. When an AI can, as Microsoft claims, translate a straightforward news story as well a human, it means it can only get better. Because AI and machine learning work by analyzing patterns in enormous amounts of data, the capacity to translate will improve, and may well begin to include context, or jokes, or proverbs, and more.

Naturally, Microsoft's claim should still be treated with some skepticism. The test only used news, which are assertions of facts, rather than something more poetic or idiosyncratic. It's hard enough for humans to translate something as language- and context-dependent as poetry, and it seems unlikely AI will be able to do something like that soon, if ever. This represents an optimistic first step, not the obsolescence of translators.

But that we are talking about it all means something significant. Translation is just one area of human work that is now subject to replacement by AI. Another may well be photography. Just recently, The Verge's Vlad Savov used a Google Pixel 2 to take pictures at an auto show, leaving his high-end DSLR at home. It would have been unthinkable to use a phone for professional purposes a few years ago, but the machine learning, AI-driven photography of the Pixel 2 allows for quality that is beginning to rival not just the point and shoot camera, but the high-end, expensive ones, too.

There are also bots that can write basic news stories, AI that can compose music comparatively well, and even some attempts at that most ineffable of arts, poetry.

It all feels a bit strange, perhaps even dystopian. We like to think that the one way that machines cannot replace or better us is in those fields that are reliant not just on data, but feeling. But perhaps AI will force us to confront the fact that the meaning we get out of art isn't inherent, but contextual, something we create through inventing patterns. Maybe AI will, counterintuitively, liberate us from the idea that art is the product of a brilliant artist, and is instead, an arrangement of aesthetic elements that helps us see things in new, oblique ways.

That said, it's not that there isn't cause for concern in a world in which AI can translate, or write poetry or music. We know now that the algorithms that drive AI aren't neutral things, but are instead shaped by the differing interests and biases of the people designing them. Who decides which translation of a speech is definitive or correct: Microsoft or a human translator? It's just one of a series of questions that will occur as AI starts to creep into creative sectors. It's worth considering.

After all, translation is just one form of treason — AI that is allowed to run without thought for its consequences may well be another.

To continue reading this article...
Continue reading this article and get limited website access each month.
Get unlimited website access, exclusive newsletters plus much more.
Cancel or pause at any time.
Already a subscriber to The Week?
Not sure which email you used for your subscription? Contact us
Navneet Alang

Navneet Alang is a technology and culture writer based out of Toronto. His work has appeared in The Atlantic, New Republic, Globe and Mail, and Hazlitt.