The machines can now translate a contract in under a second. They cannot tell when a client means ‘we’ll circle back’ as a polite way of saying ‘never again.’
For high-stakes work, the industry still insists on a human in the loop. That human is usually a freelancer billing by the word, reading a machine’s output and quietly fixing the parts where the speaker meant the opposite of what was said.
So who is actually translating the emotion? The answer is a tired person at 11pm, trying to figure out whether ‘with respect’ in a diplomatic memo is a cordial greeting or a declaration of war.
A Brussels negotiator once sent a message reading ‘your proposal is interesting.’ The translation system rendered it as ‘your proposal is fascinating,’ and the other delegation spent three weeks preparing a celebration. The original meant the proposal was a fire hazard.
The AI is fast, cheap, and confident. It is also fluent in every language and understands none of them when someone is being rude politely.
The translation business has reached the point where the most valuable skill is knowing when the machine is confidently wrong. That is now a job title, and it pays better than translating.