Having watched the reboot of Star Trek recently (and spending this morning musing of what to post for the blog), I remembered one of my favourite characters from the Star Trek of my generation, a.k.a. Star Trek – The Next Generation (TNG): Lt. Cmd. Data. [I was born in 1986, too young to appreciate the original Star Trek. Sadly, my memory of TNG is not exactly very lucid either, but here goes an attempt.]
Lt. Cmd. Data
Well, Data is definitely the pinnacle, one of the end points that NLP can reach. An android that is able to communicate with humans in human language, rather than in programming languages, clicks of a mouse, or touch on the screen. Given the fact that he’s able 60 trillions operations a second (and over 80 petabytes of memory), that might not be too surprising. Well, no, actually, now that I think about it, it is pretty surprising. Assuming 3GHz processor, current computers can perform those 60 trillions operations in under 10 minutes. Given the fact that natural language processing and computational linguistics sometime utilize tens of machines (heck, Google utilizes clusters of probably hundreds of machines and megahuge corpus), those 60 trillions operations suddenly seems very small. Yet, we are still so far behind what is portrayed by Data. Or do we really?
Truth be told, yes, I do believe we’re behind Data. But if you had been faithfully following Star Trek, Data isn’t that good with NLP (actually I only realized this now). He has not mastered contractions (shortening of “I am” to “I’m”, “is not” to “isn’t”, etc.). His speech is also more monotonous and rigid (although in one of the episode, where we were shown a future Data who is a professor at Cambridge, Data is portrayed as having more nuanced speech). Data is also less capable to use inflections. Minus these points, on the other hand, Data represents a possibility for NLP that we have not yet attained. It is very exciting indeed. (:
Lastly, of course, I can’t write a Star Trek post without alluding to universal language translator engine that is supposed to instantaneously translate languages into understandable language (well, in this case, English; or whatever language Star Trek is being dubbed into). Can we ever make such devices? What do you think about it? (: [Talking about universal language translator, I think Star Wars' protocol droids (such as C-3PO), which can translate between many languages, are more likely to be created than this one.]
Acknowledgment: Lt. Cmd. Data screenshot is taken from Wikipedia. As this is a screenshot of an actual films, the copyright likely falls to the studio. However, it is believed that usage as identification of a character in a character review qualifies under fair use under US Copyright Law.