The advantages are that the translation is done faster helping understanding the content of original text. The disadvantages are unknown sharing of the information and accuracy of translated material (text) depending on word ordering of original text.
I basically agree with the view expressed by Alberto Calderone, but would add that MT effective translations seem to vary depending upon the languages involved.. MT between European lgs are more likely to express the senses of the text, but MT between languages with very different grammars or structures are more like to go astray. E.G. English and Chinese or English and Thai.
They are good for translators who don't have certain dictionaries. It also helps in some ways but in general long texts need to be translated by a person. Nuances, cultural differences, and vocabulary that is very local need to be translated by a person. Yet,MT can help. .
I have used SDL Trados during localization of Windows and MS office into Albanian. Trados translates based on memories, which are pairs of sentences of the source and target languages.
I would also add that sometimes MT focus on individual word meanings and get the overall translation completely wrong, yet the problem of n-gram modeling is the sparsity that you will get for some if not most translations. Depending on the machine translation algorithm used it may also be biased to some particular translation over other. This biases translations which based on the context might be less appropriate than other alternatives.
My experience with Spanish and English translations both ways is less than favorable. Nonetheless, it may be helpful in studying (and modeling) some problems that could also be faced by students when studying other languages.