How long will humans be needed for translation? AI is taking over, but can it handle the nuances of legal and medical language without making critical errors?

A well-trained system might know to translate 'cats and dogs' as 'cães e gatos' in Portuguese, but what about idioms in Indonesian? More and more translation is being done by machines, but on topics with higher stakes than the weather.

Direct sales of machine translation were worth $1.4bn last year, and for many use cases, raw AI translation is now the right call. But for certain fields, like legal and medical, the question is no longer whether AI translation is good, but whether it’s good enough to remove the human.

Take legal documents, for example. AI translated 'Le demandeur' as 'the requester' and 'Madame X, épouse Y' as 'Ms X marries Y'. The consequences of mistranslation in the legal sphere can be serious, as seen in the US where errors in machine-translated asylum applications have led to delays in the release of detainees.

Accuracy is crucial, especially in the medical field. AI translated the Greek letter μ as milli instead of micro, which could lead to a thousand times the dose. As AI translators have improved, catching errors has become more difficult. If it's something that will bear your signature, there's a risk for your reputation.