Built to assist, not replace: inside Intercall’s real-time AI for professional interpreters

Among the interpreters who rely on it, the verdict is consistent: at last, something built for the way they actually work. The premise is simple. Interpreting works best as human and machine together, not machine in place of the human.

Interpreting is one of the hardest things a person can do in real time. The interpreter has to listen, understand, reframe and speak almost at once, often through an unfamiliar accent or a run of numbers that arrives faster than anyone could write it down. Researchers describe a “tightrope hypothesis”: for most of an assignment, interpreters work at the very edge of their mental capacity, and accuracy slips as soon as that load rises. The strain is physical, too. Time pressure raises stress and heart rate, and quality declines in turns longer than half an hour, which is why interpreters are rotated roughly every 30 minutes.

That pressure has only...

Copyright of this story solely belongs to thenextweb.com. To see the full text click HERE