For an interpreter, just as for a translator it is almost humiliating to suggest that a machine will ever be able to do your job, just as well as you do it. I guess many skilled worker have felt the same during the whole era of industrialization and digitalization. Computer Aided Interpreting is a different thing, it is already here. Bringing your computer and you internet key to the mission in order to use dictionaries or search the internet is a habit for many of my colleagues, both community and conference interpreters. But suggesting that you could actually put a computer in my place and that would be just as good is hard to swallow. Luckily, although many argue the difference, it will take some time before we get there. As Nataly Kelly says in an article in TCWorld: “Unfettered bidirectional speech-to-speech communication is still the Holy Grail in the automation space.”
A recent study made by Milam Aiken et. al. is reported in Translation Journal, an on line journal on translation. They claim that in their test between Korean and English there were a lot of errors but the result was still acceptable. The problem with their report is that they don’t qualify “acceptable”. Acceptable to whom, and does acceptable mean understandable?
The reason I believe that it will take some time to develop this is first of all the problems that machine translation tools has. I don’t deny that they are a very good tool, especially the ones you pay for (haven’t we all had a lot of fun with Google Translate?), but hardly as a human translator, especially when it comes to smaller or rarer languages.
Secondly, the problems with speech recognition software, OK, they are getting a lot better, especially for English, but just think about how difficult it is speaking to a speech recognition software over the phone: “Did you say “too”?”, “No, two”, “I don’t understand, did you mean “not”, and so forth. My own experience of these services is that in the end the computer decides for you and you get information that you do not want.
And finally, a gentleman points out on Proz that: “almost nobody, including most interpreters, really know how human interpretation functions. Thus, until now nobody has thought of modeling into a computer engine and replicate the mental processes applied and the information used by human interpreters (which, by the way, are only half-explored). This a huge task, but POSSIBLE.” I mean that this translator has not read the interpreting studies literature very well, there are not many interpreting researchers around but MANY of them have tried to model the mental process. Actually, the modeling of the mental process is the “shibboleth” of the interpreting studies research. See for instance, Moser-Mercer, Seleskovitch, Gile or Setton on that issue. The problem is that it is, just as the gentleman said, a huge task, but also a very difficult task.
So, I’m sure that people will continue to improve computer software and make wonderful tools for interpreters and translators, but I’m still not convinced that they will completely replace the interpreters.