Cancel your children’s Spanish lessons. Put down your phrasebook. Free up the last of your grey matter that knows how to conjugate etre. Google has rendered it all obsolete.
This week, the company unveiled an “interpreter mode” for its smart speakers, which can translate 27 languages in real time.
The concept is simple: You speak in your native tongue, and the machine will transcribe your words, and read it back in perfect Italian (or Mandarin, or Hindi). Your counterpart then responds, with the speaker returning the favour in English.
Google, of course, has a running start in developing this. It’s Translate tool is now 13 years old and according to some tests is almost as accurate as a human interpreter. The company’s artificial intelligence prowess also makes it a world leader in understanding voices. Marrying the two seemed like an obvious choice.
The technology, which features on Google’s Home Hub speaker and is due to be on others that run its Home voice assistant software, is meant to become a mainstay of hotel check in desks, train stations, restaurants, and other places in which the language barrier can present a problem.
I tried it out at Google’s base at the Consumer Electronics Show in Las Vegas, in which I was put in front of an Italian speaker. On first impressions it’s hard to fault the technology, but at the same time, I’m not convinced that this is the future.
From there, it works as intended, chewing up one person’s language and spitting out the other’s. Usefully, if the device has a screen it will also display the words in both languages, so the speaker can tell if they’ve been misinterpreted.
The main ingredients work almost flawlessly. Yes, Google understands what you’re saying; yes, it translated my counterpart well enough (as far as I can tell). But oh boy, this is a strange way to hold a conversation.
For one thing, the process takes a long time. Having a sentence understood, interpreted and then read out by a machine takes a lot longer than just… saying something. For any meaningful exchange of information, you’re going to have to dedicate a few minutes.
Both conversationalists naturally end up staring at the screen to view their own words, and then to see the translation. This generally means any nuances of conversation – eye contact, gesturing and emphasis – are lost in translation.
And fundamentally, conversations just don’t work like the way Google is imagining, in which one person speaks for a few seconds, and the other person replies. One person may have to talk for a long time to get a point across (Google seemed to stop listening after I had said a few words), or another will interject, or there may be long pauses between people speaking – the systems stopped listening after long pauses.
Gogle’s interpreter seems best suited for the sort of fake conversations you find in a French textbook (“Hello, where is the video store?”, “The video store is two miles northwest, you can take the tram”) than how they tend to work in reality.
So no, it’s not quite a substitute for being able to speak in a foreign tongue. But for anyone who’s ever been thwarted in their attempts to check into a hotel room by the language barrier, it may just get them over the line.