@jalcine If you browse the early bits of a compilers textbook you'll often find a brief overview of the "Chomsky Hierarchy". He came up with this in an attempt to describe human languages, which it turns out it doesn't very well, but if you apply it to computer languages it actually captures some key insights
His contributions in Manufacturing Consent and the Chomsky Hierarchy are directly useful and insightful in the relevant fields (sociology/media studies and computer science, respectively)
In linguistics, his contributions have often suppressed useful field work & physiological research because his disciples have often treated his (kinda armchair-y) linguistics writing as dogma rather than engage with the subject as a empirical science
(To be fair, he's difficult to read in all three disciplines, but linguistics is the worst.)
The very first Markov chain bots were made to mock his convoluted writing style (and poke fun at his favorite arguments about the productivity of a Markov machine)
I literally read something last year that he wrote about LLMs and he went back and critiqued them because "they don't have syntax", which misses most of the forest (they don't have meaning) for the trees (hah)
Add comment