> - He talks about Linguistics, not NLP, so I don't know what your friends have to do with that part.
He specifically talks about computational linguistics, which is practically synonymous with NLP (there is a minor difference in connotation--the former is scientific rather than engineering, but it's really very minor).
That's right. After paying more attention to the paragraph in question I'd love the grandparent asking his friends about the new models that are being tried and how are they based on Wittgensteinian pragmatics. I haven't seen any, but I've only been exposed to the applied mainstream of NLP, not to the bleeding edge of the research.
I have the impression that Wittgenstein's musings were of an informal, philosophical nature, and that they are difficult to translate to a practical real world application. That doesn't have to mean that he's necessarily wrong, just that for practical work such as in NLP you are forced to make some rather strong assumptions which are actually just wrong, but which allow us to approximate language to an extent where current technology can do something useful with it.
He specifically talks about computational linguistics, which is practically synonymous with NLP (there is a minor difference in connotation--the former is scientific rather than engineering, but it's really very minor).