To start, each constructs bilingual dictionaries without the aid of a human teacher telling them when their guesses are right. That’s possible because languages have strong similarities in the ways words cluster around one another. The words for table and chair, for example, are frequently used together in all languages.
So how does the computer know that table and chair are often used together? What about cultures that do not have chairs, but do have tables? How is the computer mapping co-occurances that have a significant variation by culture, or simply do not exist?
This article uses Chinese and Arabic as the example languages for mapping. The underlying cultural principles are rather different, for community, behaviors, constructs, and as these are mapping IN language, does one become more like the other? Does the machine create an Arabic with Chinese sentiments? [The papers use French and English, which are much more similar, culturally and linguistically.]
Treating language like math is not going to turn out well. Though I haven’t yet read the papers that support these assumptions.
Back to my regularly repeated statement: machine translation and language that does not address the significant cultural components of language, as communication, as culture, as transfer mode of ideology, will, in the end, create a different or new culture, and now would be a very good time to be paying more attention to this.