AI translators are getting more fluent – but communication is never just a matter of words

AI speech translation has become impressively fluent. But fluency in words doesn't guarantee understanding, according to The Conversation—and in high-stakes settings, the difference can matter. Consider a doctor asking a patient about missed treatment sessions. A straightforward "no" might mask hesitation, embarrassment, or ambiguity that a human interpreter would catch but an AI translator might miss. In England and Wales, over a million residents need language support in daily life, yet qualified interpreters for some languages are scarce. A recent survey found a third of UK health, social care, legal, and emergency service workers use machine translation. The temptation is obvious: when human interpreters are hard to find, automated translation feels like a solution. But AI excels in controlled conditions—clear speech, people taking turns, straightforward exchanges. Real conversation is messier. In asylum interviews or mental health assessments, a language-related miscommunication can have serious consequences for someone's health, legal status, or liberty. Researchers argue the answer isn't to ban AI translation, but to use it strategically: reserve it for genuinely low-risk interactions, keep human interpreters in the loop for complex cases. Under-represented languages face a particular risk—sparse AI training data and scarce qualified interpreters compound the problem. A responsible approach means investing in professional expertise and oversight, letting technology support human judgment rather than replace it.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton