What Interpreters need to know about Machine Interpreting

Argh! The robots are taking our jobs!
Maybe. I’m Jonathan Downie and this is inside interpreting. Today we’re
looking at what professional interpreters need to know about machine
interpreting. Now, first a little caveat. I can’t tell you too much of what I found
out because there is a sequel to that book coming out very soon. I’m currently
writing it. So I can’t give away too much of what’s in the book. I just want to
give you three very basic things that you need to know about machine interpreting. Thing number one and this is gonna be a quick video, trust me. Number
one: machine interpreting software is trained on words and written language. It
has no social common-sense nor can it have. How important that is depends on
your view of interpreting. Number two: Largely the corpora that machine
interpreting solutions are being trained on are corpora of written language.
Certainly, in the machine translation parts are mostly trained on written
language. Anyone who knows linguistics can tell you that written and spoken
language are two entirely different things.
Number three: It’s entirely wrong to write off machine interpreting as
useless. There are certain tasks that it does pretty well. You’ll find out exactly
what tasks those are in my upcoming book. But on the basis of those three things,
and some of the previous videos on this channel, you should already be able to
put together a picture of what machine interpreting is going to mean for
the future of professional interpreters. When my new book comes out, I’m gonna
be doing a live Q&A on the patreon channel, which I’m going to tell you
about in a few videos’ time but in the meantime, think about
those three things. Machine interpreting has no social sense. The machine
translation section is trained on written language and is actually very
good or reasonably good for certain tasks.
What does that information mean for your business and for your future? How much do
you know about the way machine interpreting algorithms are trained and
built? How much do you know about the way they are evaluated?
The decisions that you make about the future of your business and the way that
you view the future of interpreting will largely be determined by how much you
know about those things. Certainly to make those decisions intelligently, you
need to know something about those things. Have a think about it, drop me your
comments. I look forward to hearing from you. This is Jonathan Downie and this is
Inside Interpreting.

Tags: , , , , , , , , ,


  • Riccardo Bravi says:

    Is it still sensible to study interpreting if your job is going to be replaced in a blink of an eye? It is such a creative task I don't think it can be phased out by machine interpreting over the next few decades. At least interpreters have a broad knowledge which can help them be onboarded in other working positions such as trade

  • Mateus Takayoshi Kida Souza says:

    Hi, have you seen real-time translation devices that are being currently advertised? I've used them and they work sometimes….. people are somehow satisfied with short message interpretations. they seem to work with long messages as well.

Leave a Reply

Your email address will not be published. Required fields are marked *