Highly interesting article about how Google Translate differs from other machine translation programs, and what the consequences are:
This is of course a tremendously clever thing for those Google computers and their algorithm buddies to achieve. A couple of weeks ago there was actually a comparison of the Google Translate version of the opening paragraphs of The Little Prince with other translation program results. Google's was near perfect. The others weren't nearly as good. Now we know why. Google just went out and tracked down the standard English translation, out of a number of available ones, and produced that, almost unchanged. With its millions of digitized books, it can usually find translations of what you want translated; if not it has o be cleverer, and go looking for sentences that are like the ones in the text you want translated, then find published translations of them in some book or other, and make adjustments. And since the great majority of the sentences we use are used over and over again, the likelihood is good of a match, or many matches.
I'm stunned and marvelling. Also a bit afraid.