Below is an example of a type of catastrophically erroneous translation produced by machine translation systems such as Google Translate, as a warning of the risks of using such systems. It serves as an example of why, while machine translation is an interesting field of research and has genuine applications when used with caution, it is not usually a substitute for a professional translation produced by a human translator.
You may be used to seeing translation disasters take the form of incomprehensible garbage. In such cases, although the system fails to produce an accurate or even comprehensible translation, there is arguably little harm done in the sense that the failure can at least be detected easily by a speaker of the target language. Or in other words, the reader hopefully knows to simply take no notice of a translation that is obviously nonsense.
However, a more subtle danger of machine translation is the possibility of producing an apparently fluent sentence but which unfortunately means something different from the original source text. In some cases, the sentence produced can even mean the opposite of the original intended meaning.
In this example, virtually the opposite of the intended meaning arises because Google Translate is unable to capture the meaning of French ne ... que, meaning "only". Take the following sentences:
Google Translate offers the following translations:
Google Translate mistranslates these sentences to more or less the opposite of their meaning by inserting the presence of the word not. The actual interpretation of these sentences is that the apartment is to be rented only at the stipulated price and "in the following circumstances".
Google Translate offers the option to view some of the alternative translations of a particular section of the sentence that it was considering. Doing so in this case reveals an underlying problem: the system is considering as "alternative" translations phrases that essentially mean the opposite of one another.
So when deciding whether to rely on a machine translation, bear in mind that the system could produce serious but subtly undetectable errors. It isn't just the possibility of producing disfluent garbage that you need to watch out for if you rely on machine translation!