When I studied NLP one of the first thing that was drilled into my head by that book (http://www-nlp.stanford.edu/fsnlp/promo/) was that these are all statistical techniques and context based understanding of a corpus is still a while away. The problem really is to condense all of that knowledge about the real world into a form that is computable. As that really famous example goes, how do you tell a computer that water is 'wet'?
However, this is still an extremely important development and it shows that sooner or later by leveraging the things computers are good at we will be able to solve such problems. I really don't think that it is something impossible to do. It just requires a new approach that none of us have thought about. After all not too long ago nuclear fission was said to be impossible, until a chain reaction was conceptualized.
Perhaps AI will go this way too, the difficult almost intractable problems will turn out to be easy under some new paradigm. Perhaps not. We'll never really know until we try.
So kudos Regina Barzilay and her team for pushing the limits.
The process is based on making significant assumptions - mainly that the language in question has some kind of mapping to a known language. Note that this is pretty much how things like Linear B were deciphered so it is quite an achievement.
Not really a "universal translator" as it presumably won't work with a language that has no shared structures with other languages.
From Wikipedia: "Darmok" is an episode of the television science fiction series Star Trek: The Next Generation, first broadcast in the United States on September 30, 1991; it was written by Joe Menosky. The episode contains a brief, but notable recounting of the Gilgamesh epic, which is communicated by Picard to the Tamarian captain. The episode has an average rating of 4.5/5 on the official Star Trek website (as of 18 August 2009 (2009 -08-18)
The most relevant part of that episode to this discussion being that the universal translator still left the language on the planet unintelligible because they spoke in metaphor & with allusions to their planet's version of the Gilgamesh epic.
However, this is still an extremely important development and it shows that sooner or later by leveraging the things computers are good at we will be able to solve such problems. I really don't think that it is something impossible to do. It just requires a new approach that none of us have thought about. After all not too long ago nuclear fission was said to be impossible, until a chain reaction was conceptualized.
Perhaps AI will go this way too, the difficult almost intractable problems will turn out to be easy under some new paradigm. Perhaps not. We'll never really know until we try.
So kudos Regina Barzilay and her team for pushing the limits.