Machine Interpreting and Misleading Marketing

Nowadays, it seems that most tech companies have produced a product that promises machine interpreting. These are then promoted and sold to a public desperate to be told that the language barrier has fallen and they can now go anywhere they like and talk to anyone, without ever seeing another interpreter again. Whatever we think of this publicity, there might be lessons in it for human interpreters.

What’s Really Being Sold?

To demonstrate how its device will “change the way you travel,” ili, a Japanese company that makes a pocket-sized one-way interpreting device, created lots of videos showing people using its product.1 The company’s flagship ad, however, simply features lots of women wandering around with an ili device in their back pocket. All the camera sees for the entire ad is precisely that device and those back pockets. It seems that ili has decided that the best way to sell machine interpreting is with butts.

What do women’s bottoms have to do with machine interpreting? The truthful answer is: not a lot. Yet, the fact that a company would commission an ad that places posteriors over performance says a lot about what it’s really selling.

If the main selling point for the device is its technological achievements and flawless performance, then it makes no sense whatsoever to create an ad that simply shows it in someone’s back pocket. This makes it clear that ili is not simply selling an interpreting device, but a lifestyle choice. For ili, and for its target users, the device is both a travel accessory and a fashion item.

There might also be a second reason why a company’s ad would emphasize the physical features of the people carrying its device over the actual results produced by the device: sex sells. In a much more controversial ad for ili’s device, entitled “Kisses in Tokyo,” a monolingual guy sets a goal for himself to get a kiss from as many Japanese women as possible.2 The ad might have been too controversial for ili to keep it on the company’s corporate YouTube channel, but it was clear in its intention. For ili, as well as a few other makers of machine interpreting devices, the key to a successful marketing campaign is not so much to show what the device does, but to provide the consumer with a vision of who the device allows them to become.

What ili is selling is not just the idea of suddenly becoming adept in another language, but the power that gaining that ability gives you. We interpreters and translators might happily talk about how our work connects cultures, enables business deals, or saves lives. However, the makers of the ili device would much rather show how their devices might get you a nicer holiday or even a date.

Using sexual attraction to sell machine interpreting is not exclusive to ili. Waverly Labs, the developer of a two-way interpreting device it claims is capable of interpreting between 15 languages and 42 dialects, uses a similar, albeit more subtle, strategy. In a video explaining why the people at Waverly decided to invent their device, the company’s founder admits that he came up with the idea when he met a French woman who didn’t speak English very well.3 After this grudging admission, the ad alternates between suitably impressive engineering diagrams alongside happy engineers and scenes of the company’s founder showing the French woman around a town.

It may not have the same amount of cheek as the posterior-filled ad from ili, but the message is almost identical—language ability gets you dates, so use these devices to give you that ability. In the Waverly Labs example, the difference is that the company is also happy to discuss the technical expertise that went into making the device in the first place. The overarching message is that Waverly has called in the best engineers to make the best product so you, the user, can date the best people.

We probably shouldn’t expect too much honesty and technical accuracy from ads. Their role is to sell products, not explain science, but the fact that two major manufacturers saw sexual attraction as the key selling point of their devices should give us pause. I have yet to see a professional interpreter or interpreting agency use the same tactics (and perhaps they shouldn’t), but we have to admit that the ads have been successful. Waverly Labs raised $2 million within the first hour of their crowd-funder being live, and, despite the $199 price tag of its device, ili has built a growing fanbase, including 600,000 followers on YouTube.

Selling Solutions to Real-Life Problems

The way that these companies are selling machine interpreting is gaining them credibility and sales at levels that surpass what we would expect in our industry. Before we look at the actual truth behind machine interpreting, there’s one more machine interpreting device manufacturer to mention. No discussion of machine interpreting would be complete without Google Pixel Buds, billed as the perfect match for a state-of-the-art Google Pixel phone.

Since Google is one of the world leaders in machine translation, which is the core of any machine interpreting device, it might be expected that the company would be the most brash in its sales pitch. Yet the opposite is true. While ili and Waverly Labs are keen to bill their devices as dating aids, Google’s pitch is much more sedate. For them, the greatest selling point of machine interpreting is pizza.

In a blog post that trumpets the power of its “Google Pixel Buds” to overcome language barriers, the Google product engineering team asks us to imagine being in Italy and not being able to order a pizza. Just one touch on your earbud, they say, and you can get the Italian for your favorite toppings.4 It’s a much more down-to-earth approach than the one adopted by ili and Waverly Labs. It also has the added advantage of being an application that sounds more plausible to anyone familiar with the current state of machine interpreting technology.

What all these marketing approaches have in common is that they present an everyday problem and push a piece of technology as the solution. Cheap airfares have opened up international travel to the masses, leading to inevitable language issues. Whether those language issues involve just getting around, ordering pizza, or meeting the love of your life, everyone can recognize the awkwardness that comes from realizing that you simply can’t communicate.

In targeting traveling consumers and promising to solve real-life issues, the manufacturers of machine interpreting have done a far better job at marketing their products than most of those who sell professional human interpreting. Check any interpreting agency and you will see the same old, hollow platitudes about “100% accuracy,” “24-hour availability,” and “expert linguists.” None of this comes close to the persuasive power of showing how access to an interpreting service can make your holiday easier or get you a nice, hot pizza.

The truth is that these machine interpreting manufacturers aren’t targeting the business-to-business, legal, or medical markets that most interpreters will focus on. Few of us could make a decent living by selling our time in short increments to people who want to be able to get kisses from strangers or order a meal. Yet, we can learn a lot from the way in which these manufacturers have used real-life problems and a bit of imagination to sell their wares. What if we marketed our services according to the problems we solve and the difference we make, rather than our list of qualifications and certifications?

Marketing to An Ideal—Understanding the Capabilities

To grow our businesses, we also need to understand something about the accuracy, or lack thereof, of machine interpreting marketing. While we could debate the merits of multilingualism as a factor in attracting a mate, it’s probably more important for us to understand what machine interpreting is currently capable of and how far its marketing represents an ideal, rather than actual deliverable results.

Let’s start with Google’s pizza example. Assuming that you have an accent that’s compatible with speech recognition (as a Scotsman, mine isn’t), you can run a simple test on how well Google Pixel Buds would perform, even if you haven’t bought them. Open an app that allows you to use speech recognition on your phone, such as Google Docs, and simply say “Can I have a pizza with [insert your favorite toppings], please?” Then copy and paste whatever your phone comes up with into Google Translate and ask it to translate it into another language. For the big European languages, and perhaps even Chinese and Japanese, the results will probably be pretty good. Simple sentences will most likely be handled pretty well in languages where Google Translate has a large enough corpus from which to work.

There is a slight issue, however. My personal favorite pizza topping is ham and pineapple. Now, it’s my understanding that asking for pineapple on pizza in Italy isn’t a good way to get into the good graces of the pizzaiolo. No matter how accurate the interpretation, Google Pixel Buds would be no good in the ensuing debate over the merits of my pizza choice. However, a good human interpreter, who values the continued health of the client, might be tempted to suggest that pineapple and pizza are not classically seen together in Italy.

Beyond Pizza Toppings

What happens when we ask machine interpreting to do more than deal with simple cases of people asking for unpopular pizza toppings? While I don’t have an ili pocket device or a set of Waverly Pilot Buds on hand to try, there have been two recent trials that provide insightful data points.

Tencent might not be very well known in the English-speaking world, but it is a Chinese company with the financial and social clout of Google. Like many tech companies, Tencent enjoys releasing new products with flashy demonstrations. In April 2018, the company felt that there was no better opportunity to launch its new interpreting system than to let it take over from human professional interpreters at the Boao Forum, a showcase aimed at senior leaders of government and industry.

The results were far from perfect. The system, which was also asked to transcribe its output onto a big screen, produced a stream of garbled Chinese characters, repeated words, and, at one point turned a line about China’s “Belt and Road Initiative” to build infrastructure around the globe into “a road and a waistband.”5

Recently, there was a more hotly debated example. In a story that easily deserves an article of its own, iFlyTek, a software company based in China, was accused of trying to pretend that interpreting produced by a professional human interpreter had actually been produced by its artificial intelligence system.6 The twists, turns, claims, and counter-claims involved in the story are complex, but the basic upshot was that iFlyTek was faced with little option but to admit publicly that its system was not even close to matching the capabilities of human professionals.

Time to Change Marketing Tactics

That the actual performance of machine interpreting is far from the results advertised should surprise no one. In fact, most professional interpreters would expect the technology to be flawed. However, just because machine interpreting marketing is misleading doesn’t mean we can dismiss it.

We may be aiming at different markets than Google or Waverly Labs, but we can learn from them. Few, if any, of us will have a marketing budget the size of ili’s or the connections of Tencent, but what we have on our side is our track record and stories. If nothing else, the success of machine interpreting marketing in the face of reality should be a wake-up call to all professional interpreters. We need to start using our imaginations to market our services according to the difference we really make. If we don’t shift our own marketing, we’re handing over the future to well-marketed but flawed machines. It’s our call.

Remember, if you have any ideas and/or suggestions regarding helpful resources or tools you would like to see featured, please email Jost Zetzsche at jzetzsche@internationalwriters.com.

Notes
  1. Watch “ili—It Will Change the Way You Travel,” http://bit.ly/ili-YouTube.
  2. Watch “Kisses in Tokyo,” http://bit.ly/Kisses-Tokyo.
  3. Watch, “Waverly Labs—A World Without Language Barriers,” http://bit.ly/Waverly-language.
  4. Champy, Adam. “Google Pixel Buds—Wireless Headphones that Help You Do More,” Pixel Blog (October 2017), http://bit.ly/buds-Pixel.
  5. Chen, Celia. “AI-Powered Translation Still Needs Work After Errors Mar Debut at Boao Forum,” South China Morning Post (April 16, 2018), http://bit.ly/Tencent-Boao-Forum.
  6. Ng, Lance. “China Tech Leader Accused of Faking AI-Generated Simultaneous Interpreting,” Slator (September 26, 2018), http://bit.ly/iFlyTek-Slator.

Jonathan Downie is the owner of Integrity Languages in the U.K. He provides French<>English conference and business interpreting services, public speaking, and content writing. After two years as a columnist for industry publications, including the ITI Bulletin and VKD Kurier, he published Being a Successful Interpreter: Adding Value and Delivering Excellence (Routledge, 2016). His work has appeared in magazines for senior administrators, the U.K. events industry, and Flight Time, the in-flight magazine of the European regional airline, Flybe. He is a board member of the European Society for Translation Studies and the Institute of Translation and Interpreting. He has a PhD in interpreting from Heriot-Watt University, Edinburgh, U.K. (thesis topic: client expectations of interpreters). You can find his blog at www.integritylanguages.co.uk/blog. Contact: jonathan@integritylanguages.co.uk.

The ATA Chronicle © 2019 All rights reserved.