When Julie Tay first worked as a freelance interpreter in New York in the 1990s, language services didn’t formally exist as an industry, even in our city of immigrants. It was tied to the dominance of an English-speaking, “white-bread” America, she says, one where immigrants were expected to eventually assimilate to survive. The idea was that the language problem would be a temporary one.
So the job came to her, not the other way around, she says. Tay was teaching and freelancing when word got around she had interpretation skills in Cantonese and Mandarin. Agencies sent her on interpreting assignments in the courts, and she fell into translation work through those cases.
Back then, decades before machine-assisted interpretation and translation became fixtures, a career in language services opened up various opportunities, including working as tour guides or cultural liaisons for delegations from foreign companies.
But today, widespread use of new tools that are supposed to make translation and interpretation easier have language workers concerned. Some worry machine translation will make their jobs obsolete. And students entering the field are second-guessing their career choice.
Now Tay, who was named director of Hunter College’s Master of Arts in Translation and Interpreting (MATI) program, is part of a group of industry leaders fighting to help language workers tech-proof their career and assert their role beyond supporting the dominance of the English language. Some experts say the future of their field lies in making technology work for them, not fearing it, depending on it, or pretending the problem will go away. Co-existing with AI means changing the way language workers think about their job, Tay says.
Being a language worker in the age of AI
AI is a concern despite an increased demand for translation and interpreter services in New York, as well as the rise in opportunities for remote work for independent translators and interpreters since the Covid-19 pandemic.
At a “fun” industry mixer hosted by the Circle of Translators in March, the most common questions for the newly formed state Office of Language Access weren’t such fun ones.
Recently laid-off or underemployed local language workers asked: How can we get interpreter or translation jobs? Could we get the list of state-contracted language service vendors to ask them for jobs? With the increase in virtual interpreters already, would NYC Health and Hospitals be hiring more face-to-face interpreters who know the culture and community?
In May, at a conference hosted by MATI at Hunter College, worries about AI’s role in the industry took center stage.
The shift towards AI and other tech solutions in the language services industry is part of a broader trend that began with the digital revolution, Tay says. The emphasis on efficiency often sidelines the human aspects of interpreting. This, in turn, further marginalizes language workers, limiting their ability to exercise agency and build community connections in their roles.
“Linguistically interpreting is the lowest level of interpreting; it’s brainless mechanical interpreting,” Tay said. “If students want to be over-obsessed with it, and just narrow themselves down to this at the expense of learning life skills, … you discount yourself in the age of AI.”
Interpreting culture when the stakes are high
Unlike others Epicenter NYC spoke with at the industry events, Luis Montes de Oca isn’t so worried about his job security. Before his 15-year career as an interpreter, he developed the “life skills” that Tay cited, through jobs in sales, entertainment and media. Now he applies his diverse cultural knowledge in healthcare and legal settings and at hotels, interpreting for and to asylum seekers. The work goes beyond mechanized interpreting, he says; it means knowing when to ask doctors for clarifications and when to ask the same of patients.
“In our culture, sometimes we say ‘me quiero morir,’ when [we’re] stressed out, or there’s so much going on,” he says. The literal interpretation — “I want to die” — through remote interpreters unfamiliar with the idiom or unaware of the speaker’s nonverbal cues would sometimes land patients in the psychiatric emergency department. It’s just one example where Montes de Oca sees the value of interpreting culture.
While he says medical providers prefer in-person interpreters for these reasons, they’ll sometimes use remote interpreters for convenience — they can just plug in the interpreter machine — and cost savings.
But he doesn’t see AI getting close to meeting the need anytime soon in settings where interpretation errors can have critical consequences.
“The human presence will always be a key factor, especially in the field of health, because any misunderstanding, they’ll give you the wrong treatment, the wrong medicine, the wrong diagnosis,” Montes de Oca said.
In over a decade of interpreting at a psychiatric unit in a local hospital, he says he encountered many cases like this. One incident still haunts him: a Black Honduran man who had recently immigrated and had spent time in jail shared that, due to a miscommunication, hospital staff were forcing him to take psychiatric medications. The patient refused, saying in Spanish that the medications made him really sick, and that he wasn’t mentally ill.
Montes de Oca was in the room when the doctor got frustrated and threatened to send the Honduran man back to jail, he says.
“I’ll never forget that case, because I felt so bad, powerless,” he said. “God knows how many people are like that [due to language barriers], with no family, sent to jail or a psychiatric hospital, nobody’s saving them and they’re in hell.”
Working with and outside of AI
Even experts on the tech side of the AI quandary have qualms about over-relying on machine-assisted language services.
Large language models like ChatGTP and machine translation tools like Google Translate have a built-in bias because of the data they’re trained on. Sören Eberhardt, a program manager at Microsoft who used to work on localization, sees this in translations of gendered job descriptions in the German language, for instance. “Nurse” becomes female and “doctor” becomes male when translated.
In addition, AI has limitations.
The machines do a good job with clear-cut content like templates and technical instructions, he says. They also help speed up translations and ensure terminology stays consistent, as with the “translation memory,” a database that stores previously translated text. Together with machine translation, it’s what Microsoft translators use.
But Eberhardt also knows many translators who won’t work with machine translation, not even with computer-assisted translation tools (CAT) at all. Some tools can be costly. Or it might not be worthwhile if you don’t have text that repeats. Or if you get really bad translations, and you have to change them, that might take you more time than translating from scratch, he says.
And Eberhardt, who has a master’s degree in comparative literature, German linguistics, and philosophy, wouldn’t trust AI to translate language as complex as poetry. No language has exactly the same words for the same feelings, he says. Emotions are layered, and finding the right word is crucial, which he sees as the beauty of literary translation.
He hears, from translation agencies, about language workers’ concerns over AI as a threat to their job stability.
And he gets it: some of the more repetitive translation tasks — like instructions and help articles — might be replaced by machines, he says. But that’s not necessarily a bad thing: earlier in his career, Eberhardt used to translate error messages for Windows Server.
“It wasn’t very enjoyable,” he said. “If a machine can handle that, so be it. Many translators likely prefer working on more creative texts, which are more challenging for machines to produce. The more cultural context required for a translation, the safer human jobs will be.”
After all, companies looking to save a buck or speed up the translation process might assume technology can solve everything. Eberhardt sees this in upper management’s push for automation across industries.
“My hope is that technology actually takes away some of the menial tasks and really enables people to focus on the deeply human side of what needs to be translated,” he said. “Ultimately, you’re enabling communication between different human beings. In a way, the language industry is so privileged to have a problem they can solve. But the problem itself [of differences in languages] is beautiful.”
MATI’s program director, Julie Tay, says what gives her hope is language workers pushing their field past simply interpreting or translating — and not just to ensure more job stability in the age of AI. For her, going beyond tasks performed by AI should also be part of a larger resistance to what she sees as the industry’s primary role: serving the state and speakers of “privileged” languages (like English, French, and German), as opposed to helping speakers of diverse languages navigate the U.S. without having to assimilate.
And she sees hope in the language needs of the city’s migrant situation, in “the unrelenting influx of immigration, regardless of what the government says,” Tay said.
“Therein lies the power of community, but it’s not really the direct work of interpreters and translators,” she said. “I worry that, translation, interpreting, the way it’s going, in the sheer monetization, … that this would only reinforce linguistic hegemony and would diminish linguistic diversity even more.”
