Translator to ChatGPT Evangelists: It's a Tool, Not a Magic Button

A freelance translator in Ottawa recounts a locker-room exchange that captures a widening gap in how people perceive AI. After explaining she had to skip a gym class to handle three overnight translation deadlines, a fellow gym-goer — a senior federal HR executive — suggested she simply upload the documents to ChatGPT to be done in minutes. The author’s response: machine output is grammatically plausible but unreliable, missing the adaptation, localization, terminology research, and consistency work that defines professional translation.

Far from being an AI refusenik, the author uses these tools daily where they actually help: checking her work against a client’s 500-page style guides, extracting specialized terminology to build glossaries, and getting second opinions on drafts. She frames this as continuity, not disruption — translators were pasting awkward sentences into Google Translate fifteen years ago, then DeepL. The tools changed the workflow; they did not eliminate the need for judgment. In her experience, AI invents acronyms and organization names, drops entire sentences, and ignores supplied terminology, requiring constant supervision ‘like a toddler.’

The piece lands two sharper points beneath the anecdote. First, an economic one: professionals shouldn’t see rates cut simply because AI exists, any more than a roofer earns less for using a hammer. Second, an irony: the executive pushing ChatGPT admitted she can’t use AI in her own HR work because ‘it’s really not reliable enough’ — a neat illustration of how people overestimate automation’s fitness for everyone’s job but their own.