Veteran Technical Translators Struggle to Adapt or Be Ousted by AI

I met Warren Smith through the “honyaku” email listserv, where translators of Japanese gather to palaver and share translations of tricky terminology. Smith has 39 years’ experience in translation, a portfolio of more than 40 million words, and a doctorate from Harvard. He formerly worked as a semiconductor engineer for Signetics and Mitsubishi and regularly attends court cases as an expert witness. Smith is as qualified and capable as any technical translator could hope to be; yet this past Friday, he reached out to the listserv for help, not with finding the right translation, but in search of something else: a job.

“All of my major clients have suddenly gone entirely to the AI/post-editing model,” he wrote. “...My income has disappeared, so I am scrambling.”

Starting with the public release of Chat GPT in November 2022, the surge of new services, technologies, and business models built around Large Language Models (LLM) has shown no sign of abating. Translators, not to mention subtitlers, copyeditors, copywriters, technical writers, proofreaders—basically anyone who spins text into gold—have begun gathering to swap prophecy and debate: is AI an existential threat?

Author and researcher Tom Gally, himself a 20-year veteran of the translation industry, has taken particular interest in answering this question. In March, he gathered some four-dozen freelance translators for an impromptu colloquium, which he later summarized on Youtube. In the video, Gally establishes a difference between “reader-driven” translation and “writer-driven” translation. In other words, translation in which the client seeks access to information or seeks to deliver their own message to others.

In the case of the first, Gally predicts that AI will inevitably outcompete human professionals, who will instead be relegated to tidying the cheap words of garrulous robots (a process referred to as “post-editing”). In the case of the latter, Gally is more optimistic. He believes that translators will find job security by embracing their human traits. That is, by offering something more personable than the faceless interface of a mute, panglottic scribe—as Gally puts it, “by interacting with clients, by solving problems, by being more complex [and] less deterministic in the work that [you] do.”

For years now, much of a translator’s success has been decided by their ability to specialize. The US Bureau of Labor Statistics Occupational Outlook Handbook notes that “Students who study technical subjects... may be able to provide a higher level of interpreting and translation,” and the American Translators’ Association’s Guide to Starting Out as a Translator?repeatedly references the importance of specialization. Pharmaceutical, legal, technical—the more esoteric the subject matter, the more pennies per word. That rationale still holds water today, but this industry truism may be on its way out.

Patent translation requires wide-ranging technical expertise, and Warren’s bona fides can be seen in his contributions to the listserv. In February 2023, he chimed in to assist with a nebulous Japanese term: “I used to be an engineer in this field, where I ran some of the etching processing for semiconductors. My take on it is that this is likely an anisotropic etching process.” The discussion continued, ranging across two languages and several topics that would make a non-engineer’s eyes cross.

Just a few short months later, Smith elected to sit the bar exam and seek work as a patent agent. His career as a full-time translator was done. He approached several firms, reportedly encountering “zero interest in the Japanese skills I have been developing for the last four decades.” For the members of the listserv, his email thread is a chilling elegy for what increasingly seems to be a moribund profession. As fellow long-time translator Matthew Schlecht commented, “the harsh market churn is affecting an increasing percentage of our colleagues.”

That same day, Schlecht was one of several speakers at a virtual conference held by the American Translator’s Association. The event specifically addressed the looming presence of AI in the translation industry, as reflected by its title, “Translating & Interpreting the Future: Empowering Individuals to Innovate and Thrive.” Its keynote speaker, Jay Marciano, is the director of machine translation outreach at Lengoo, a producer of AI-powered translation technology. Regardless of how its footsoldiers will fare, translation as an industry is mobilizing for revolution.

In 2021, the US Bureau of Labor Statistics predicted 20-percent job market growth for translators over the coming decade—four times the average for a given occupation. The report notes that “computers have made the work of translators and localization specialists more efficient,” but adds that “many of these jobs cannot be entirely automated, because computers cannot yet produce work comparable to the work that human translators do in most cases.” Less than three months after this statement, Chat GPT was released, to be followed by GPT-4 in March of 2023. Looking back now, the report’s assuredness seems painfully na?ve.

Whether by embracing the nascent technologies that promise to reshape their profession or by leaning into their human appeal, it is clear that translators—like so many others—are scrambling to adapt to an industry changing beneath their feet. For some of the old guard, the transition may prove to be insurmountable. Smith is not unfamiliar with AI-powered tools, having recently completed a distance learning course. But as he commented in his email, “Despite being unusually good with AI (at least when compared to most translators, I think), I don't see any way to tie that into a bright future.”