J. C. R. Licklider’s (3 initials? Really?) article on the man-computer, while somewhat outdated in terms of the modern technology that has now been created, posits some interesting questions about how man and machine might work together. As a linguist, the most interesting of these hypotheses, to me, was the idea that computers could recognize and produce recognizable human speech. In modern times, there are in fact, multiple speech recognition software programs, and computerized speech can be found in GPS navigation systems and various other computerized items. However, the question for me, as a linguist, is how a computer can be truly taught to understand language when we as humans do not fully understand it. English has one of the most expansive vocabularies of any language in the world, and furthermore, it is very adaptable to new words through the addition of prefixes and suffixes. Furthermore, there are a large number of synonyms, homophones, and words with multiple meanings. Intonation, speed, and non-verbal cues also provide a context for definition. Lastly, if one accepts the Universal Grammar hypothesis, language is not a learned behavior but an intuitive one, much like birds learn to fly, in which case, computers cannot be taught intuition. And while one tends to think of the human mind as finite, much like RAM in a computer, language has the ability to be infinite. (I spotted a cat on a mat under the stairs with my binoculars…) Can the computer process infinity? Can it understand odd sentence structure or questions based on intonation rather than wh-movement? I know that they have created computers that can reasonably mimic human language; however, to have the truly symbiotic relationship that Licklider imagines, I cannot envision a computer that is capable of processing language in the same way that humans do.
2 Comments
Well, Licklider went by “Lick” with his friends; his wife called him “Robin.” If you get interested, there’s a fascinating biography of him called “The Dream Machine” that I recommend highly. He was a gifted researcher, leader, and teacher.
Your analysis here is most insightful. Licklider was an AI guy who really did believe that computers could eventually parse and “understand” natural language, under certain conditions (not the sentence you cite, typically). Like many AI guys, he also wished natural language could be reformed, made less ambiguous, more precise. His book “Libraries of the Future” is very interesting along these lines.
Issues of the “semantic web” lurk here as well, though Licklider didn’t know or use that term of course.
What a coincidence! Today in sociology class, we watched a video on Feral children (which as a linguist, you prob know about) http://www.feralchildren.com/en/index.php
Much of the video was centered on whether these children could learn how to speak the human language after being isolated from humans for so long. From my perspective, it seemed extremely difficult. One girl learned much vocab but could not understand how to use grammar and put sentences together. Another eventually overcome and learned but could only speak short sentences.
Yes, language is very intuitive, and I don’t think a computer would have the ability to use language as humans do.