For most of the twentieth century, translation was a cottage industry of solitary polymaths. A translator of Russian literature needed not merely fluency but cultural intuition, historical awareness, and the literary sensibility to render Chekhov's pauses or Dostoevsky's fevers into English without flattening them. The work was slow, poorly paid relative to the expertise required, and invisible when done well. Then the machines arrived, and everything changed — except, perhaps, the parts that matter most.
Neural machine translation, the technology underpinning tools from Google Translate to DeepL, has reached a level of competence that would have seemed fantastical two decades ago. For straightforward commercial content — user manuals, corporate communications, product descriptions — the output is often serviceable on first pass. Translation agencies that once employed dozens of linguists now route work through algorithms, with humans relegated to what the industry calls "post-editing": cleaning up machine output rather than creating from scratch.
The economics of displacement
The shift has been brutal for translators working in high-volume, low-complexity segments. Rates for post-editing work typically run thirty to fifty percent below traditional translation fees, yet the cognitive load is arguably higher. Correcting a machine's plausible-sounding errors requires constant vigilance; the brain must simultaneously read for meaning and hunt for subtle mistranslations that would slip past a casual reviewer. Many experienced translators describe the work as deadening — all the tedium of quality control with none of the creative satisfaction.
Yet the profession has not collapsed. Demand for human translators in literary, legal, and medical contexts remains robust. A mistranslated pharmaceutical label can kill; a mistranslated contract can cost millions. Courts in most jurisdictions still require certified human translators for official documents. And literary translation, while never lucrative, continues to attract practitioners who view it as an art form rather than a commodity service.
The emerging hybrid role
What is genuinely new is the emergence of a translator who functions less as a writer and more as an editor, curator, and quality arbiter. Younger professionals entering the field increasingly describe themselves as "language engineers" or "localization specialists" — titles that would have puzzled their predecessors but accurately capture a role that involves managing translation memory databases, training custom machine models on client terminology, and making judgment calls about when automation suffices and when human craft is non-negotiable.
This hybrid role demands skills that traditional translator training rarely provided: project management, basic computational literacy, and the commercial acumen to price services that resist easy categorization. The translator who thrives is no longer the reclusive scholar but the adaptable generalist comfortable operating at the intersection of language, technology, and business.
Our take
The optimistic reading is that AI has liberated translators from drudgery, freeing them to focus on work that genuinely requires human judgment. The pessimistic reading is that the profession has been proletarianized, its practitioners reduced to algorithmic supervisors paid less to do work that is arguably harder. The truth, as usual, lies somewhere in the uncomfortable middle. Translation is not dying, but it is becoming something else — a discipline in which the machine does the heavy lifting and the human provides the quality control, the cultural nuance, and the accountability when things go wrong. Whether that constitutes progress depends entirely on what you thought translation was for in the first place.
