a short post:
i just found this article by Steven Pinker (author of The Language Instinct). though it's a few years old now, it's a succinct and pretty introductory discussion of some of my favorite lingusitics issues.
it touches on the rules/words divide (which, in the guise of overregularization made up several interesting classes in my child language acquisition class this semester), discusses the rumelhart-mclelland computational model of acquisition, and sums up how these cognitive issues (presumably concerned with individual speakers) effect long-term patterns of language change (presumably concerned with groups of people).
he also cites joan bybee's studies of regularity and frequency. i'm just reading a paper of hers (her 2005 LSA presidential address, downloadable on her website), in which she talks extensively about a relatively new usage-based approach to linguistics. as far as i can tell, it's at least related to the school known as functionalism or cognitive linguistics (see also functionalgrammar.com). she begins her speech by questioning some of the structuralist assumptions of more mainstream (read: mostly chomskyan) linguistics and, while she doesn't use the term 'post-structuralist,' the approach is based on treating mechanisms of change (as opposed to a static grammar) as language's universals.
kurzweilAI.net is not just about linguistics, as the 'AI' suggests (although the 'about' page suggests that it stands for 'accelerating intelligence' more than 'artificial intelligence'), but pinker's article reminds us that language will be a key part of any future research into machine learning/behavior. the whole thing is worth a look!