I use Facebook's translation of comments a fair bit and it enables me to usually get at least the gist of what's being said - and sometimes a clear view of it. If that level of accuracy could be applied to in-person audio it could be a lot more useful than not having it.
The point is that an imperfect translation engine can be adequate for making sense of written text in an asynchronous way (if one sentence is confusing your eyes can jump back to the previous one to try to guess the connection, etc.), but be near-unusable for on the fly translation of a real world conversation. Context matters a lot.
Then your argument is actually that they won't be able to do real time audio translation to a similar level of accuracy to Google Translate. The comment I replied to did not make that argument.
The point I am making that you are referring to is that while you may be able to understand a piece of text in a foreign language translated visually by Google Translate (because your eyes can jump forward/backwards in the text, reread a confusing sentence, you can pause for a second and try to make sense of it, etc.), if the exact same text were translated in audio form, as part of a live, 1-1 conversation, you wouldn't be able to make sense of a lot of it because you can't do all the things I just listed which help deal with imperfect translations.
It's the same thing as if I read you a very complicated scientific paper you're not familiar with out loud, versus you reading it yourself. You will understand more in the latter situation.
(imagine if we had this back and forth in 2 different languages using Google earbuds, instead of over a textual medium :-)
I don't agree. No-one knowing you're using such translation is going to talk to you in a normal conversational manner. They'll talk to you in a simplified, slowed-down fashion. And having such translation ability in such a situation would still be better than the alternatives.
