Interpreters vs Machines

Can Interpreters Survive in an AI-Dominated World?

Jonathan Downie

• Publisher: Routledge
• Year: 2019
• ISBN:
  9781000753974
• Edition: 1
• Topic: Languages & Linguistics
• Subtopic: Linguistics
• Index: Languages & Linguistics

From tech giants to plucky startups, the world is full of companies boasting that they are on their way to replacing human interpreters, but are they right? Interpreters vs Machines offers a solid introduction to recent theory and research on human and machine interpreting, and then invites the reader to explore the future of interpreting. With a foreword by Dr Henry Liu, the 13th International Federation of Translators (FIT) President, and written by consultant interpreter and researcher Jonathan Downie, this book offers a unique combination of research and practical insight into the field of interpreting.

Written in an innovative, accessible style with humorous touches and real-life case studies, this book is structured around the metaphor of playing and winning a computer game. It takes interpreters of all experience levels on a journey to better understand their own work, learn how computers attempt to interpret and explore possible futures for human interpreters.

With five levels and split into 14 chapters, Interpreters vs Machines is key reading for all professional interpreters as well as students and researchers of Interpreting and Translation Studies, and those with an interest in machine interpreting.

Level 1

The fundamentals

I still remember it now, even though I must have been only about five or six years old when we got it. Family legend says that my dad saved up for months to buy it and, on the Christmas Eve when he set it up with one of his mates and a couple of my uncles, they missed the Watchnight service at church because they were busy “testing” it until 3:00 a.m.

Eventually, it got moved to the bedroom I shared with my older brother. My children are confused by how much simpler it was then. Its capabilities are laughably limited compared to anything you could buy for the same amount of money now. But at the time, the thought never occurred to me that within a few short years, it would be obsolete and within a decade or so, it would be all but forgotten, its stark green lines and tiny frame dwarfed by the white aluminium casing of our newer model.

You never forget the first computer you use. Ours was an Amstrad CPC 464. It couldn’t do much but it did enough to keep us busy and entertained for years.

Could it be that computers might soon be so capable that they make the final leap from green-screened noisy boxes with limited memory and clunky keys to becoming fully capable of replacing humans?

No single book can answer that question for all possible futures. Even answering the interpreting version of that question takes hard work and careful thinking. For a start, interpreting itself is a startlingly complicated profession. From sign language interpreting in universities to spoken language interpreting in hospices, there are no facets of human life that interpreting doesn’t touch in some way. Before we start predicting the future of interpreting, we need to know what interpreting is and what exactly the inventors of machine interpreting are trying to do.

Since technology is the theme of this book, it helps to break down the journey towards the future of interpreting as if you were going through different levels of a computer game. As I mentioned in the introduction, this level is where we learn the basics: what the aim of the game is, what the basic controls are, where the “game” might be taking us.

That’s what this section is about. As tempting as it might be to throw away the manual and wade into the game, hoping that you can learn on the go, that route leads to nothing but dead ends and frustration. Instead, let’s begin by answering the apparently simple and fundamental but actually fiendishly complex questions that are at the heart of any prediction of the future of interpreting: what is interpreting?, how do humans interpret? and how do computers try to interpret?

Chapter 1

What is interpreting?

So what is interpreting? What seems like a simple question is actually the source of controversy, confusion and misunderstanding. Are we dictionaries on legs, filled with excellent terminology but hardly a human emotion? Are we conduits, who just say what the original speaker said, no more, no less? Are we mediators, who find the middle ground between two opposing interests? Are we activists who use our considerable power to make the world a better place?

What actually is an interpreter? Rather than giving a philosophical answer, I think it makes more sense to start with a story.

The interpreter believes that there has been a misunderstanding

Every interpreter has a favourite interpreting story. This is mine:

It was a bracingly cold day somewhere in central Scotland. I had been asked by one of my agency clients to interpret at a “sales qualification meeting.” It sounds fancier than it was. Basically, a buyer had flown in from France to go through a questionnaire on company policies, finances and product quality with representatives of a British company. Their answers to the questionnaire and his impression from a tour of their facilities would decide whether they got a multi-year, multi-million-pound contract.

I didn’t know all of that. I knew the British company well and I had studied its target client. I had studied the French questionnaire and knew what was likely to come up. I didn’t know how big the deal was likely to be (the British company didn’t even know that!) and, most importantly, I didn’t know that the company had translated the questionnaire in-house and answered “as best as they could” with their “in-house French knowledge.”

That last little piece of information would prove to be more important than anyone could have imagined.

The tour went well. I got some free education from the Brits, who were very keen to show off their expertise on how to make those specific products, and from the French buyer, who was keen to show that he had once run a facility just like the one they were seeing. The rest of the first day was fairly standard stuff: the usual yawn-inducing presentation of the British company and its client book, some semi-technical presentations and some pretty standard chat about dashboards and KPIs.

Midway through the second day, the atmosphere changed dramatically. It started with an offhand remark from the French buyer. He wondered aloud why the British company was scoring very poorly for quality on its questionnaire when the tour had reassured him that it was actually producing excellent products.

And then it all kicked off. Within less than half an hour, what had been a congenial meeting turned into a stand-off. The French buyer felt that the British company was being deliberately evasive. The British company felt that the French buyer was asking the same questions again and again when they had already answered them. Chairs were pulled back from the table. First the British technical director, then the French buyer and then everyone else stood up and made for the door. We marched along corridors through sets of heavy double doors. Fingers poked charts on the walls. Heads shook. I am sure I heard a few exasperated sighs.

In one wondrous moment, I had a realisation:

“The interpreter believes there has been a misunderstanding.”

Those weren’t the exact words I used, but they are close enough. In under a minute, I explained what had suddenly become obvious to me. The French client wanted to know the proportion of pieces that the British company was producing with no defects. The British company was used to measuring the number they produced with defects.

The relief was almost as tangible as the lunchtime sandwiches had been. A very relieved technical director smiled “Och, is that all he wants? C’mere and I’ll get those stats for you.”

Less than five minutes later, a happy French buyer was casting his eyes over a spreadsheet with the exact numbers he wanted. Within about two hours, happy handshakes matched jolly smiles, the British company was in receipt of permission to request a test order and, so long as the products were as good as they were said to be, a very nice multi-year contract was on its way to being signed. It would be a nice bonus when the chief executive of the British company would end the day assuring me that, even though he spoke French, they could not have done the deal without me.

Unpacking the story

Depending on the kind of interpreting you do, that story might seem fairly ­typical. In many cases, the success of an event, business deal or medical appointment hinges on the interpreter’s expertise and decision-making. In those environments, tiny decisions make a big difference. Do you explain a term or keep it as is? What do you do with that joke? What do you do if the parties misunderstand each other?

Good interpreting is the ability to take such decisions intelligently. The truth is that even if we want to keep using the traditional definition of interpreters as being impartial, accurate and terminologically exact, there will always be decisions we take as to how to solve specific problems. Those decisions are rarely clear-cut.

Stories like the one I just told show that there is a lot more going on than simply fetching the French equivalent of a word in English or the correct Hindi expression for a German phrase. Language comes with baggage, and it is always dangerous to assume that an accurate rendition of what was said will leave no room for misunderstanding and ambiguity. On the contrary, the very fact that language is tied inseparably to culture means that no term or phrase can ever be isolated from the culture in which it originated, no matter how much dictionaries and term bases might try to tell us otherwise.

Interpreters jump headlong into this potential confusion and have to try to find a way to swim to the other side of it. While lexicographers, terminologists and etymologists might bask in words for the sake of words, interpreters have to be aware that words are always doing something: expressing emotion or attitude, creating new realities, explaining, informing, goading, playing, deceiving, even flirting. For interpreting to work, words or signs must be found that do the same things in an appropriate way in an entirely different language.

Even the account I’ve just provided of interpreting is not enough. Interpreters don’t just play with words and reflect on the possibility of mutual understanding. No client ever hires interpreters because they want some theoretical account of terminological equivalence between Dutch and Swahili. For clients, interpreting isn’t important because of what it is but because of what it does. It might sound mercenary, but clients always want interpreting for their own reasons. They have treatment to administer, deals to strike, staff to hire or sack, treaties to sign, possible suppliers or customers to wow, journalists to tame or blame, details to explain or some other thing to do or achieve.

To interpret is to help someone achieve something that they could not otherwise achieve. Interpreting is purposeful, and its purpose may be only tangentially related to language.

The story I just told is a simple example of that premise. The purpose of the meeting was not to have accurate interpreting but to sell products. The purpose of the interpreting was to enable products to be sold. The need for interpreting only arose because the client happened to be French. The act and accuracy of interpreting were only important because of what they allowed the client to do. If that seems to go against our rarefied view of our job, then it may be because many of us love interpreting so much that it hurts to realise that, for our clients, it is merely a means to an end. And a rather expensive means at that.

Before we do any work towards defending human interpreting against the big bad boss of machine interpreting, we need to swallow that healthy dose of reality. Fighting against machine interpreting on the grounds of beauty or ethics or even human rights is a sure-fire way to be on the losing side. Building a case for human interpreting on the basis of its unique contribution to business growth or fair legal proceedings or more effective medical care or any of the hundreds of other tangible reasons for our work will be much more effective. Put simply, it’s not enough to prove to clients that they need interpreting; they need to be convinced that they actually want it.

It’s not enough to prove to clients that they need interpreting; they need to be convinced that they actually want it.

The first stage in building that kind of case is to have a realistic model of interpreting in terms of our relationship with the people for whom we interpret. To get to that point, we first need to understand two basic models of interpreting. Both are simplified representations of what interpreters do, but both have a lot to tell us about the nature of our work.

Enter the conduit model

Until the mid-1990s, there was only one model of interpreting in town, and it was a view that was held nearly universally, by interpreters and clients alike. In this model, the interpreter was nothing more or less than a communication channel through which people spoke when they needed their words to be understood by someone who didn’t speak the same language.

In a manner similar to landline telephones, interpreters were basically seen as taking a signal in one language, turning it into an exact representation in another language and passing it on. The best interpreting was therefore interpreting where there was no loss or gain in this process. The premise was simple: just say exactly what was said in the other language.

The name “conduit model” was borrowed from researchers looking at how people thought about communication processes in general (notably Reddy, 1979) but would prove to be a useful term. After all, in this view of interpreting, the interpreter had little more intrinsic value than a pipe or an electric cable.

To fully understand this model of interpreting, it helps to have a diagram. ...

Table of contents

• Cover
• Half-Title
• Title
• Copyright
• Contents
• Foreword
• Acknowledgements
• Introduction
• LEVEL 1 The fundamentals
• LEVEL 2 How machines gained the upper hand
• LEVEL 3 Choose your interpreting future
• LEVEL 4 Interpreting that beats the bots
• LEVEL 5 One last thought
• Bibliography
• Index