In this podcast, Ted Gibson explorers an empirically grounded theory of language as communication.
In the 90s and 00s, the American post-structuralist humanities, fueled by the Whorf hypothesis, reconceptualized language as coextensive with thought. Many scholars made strong claims that language encompassed reality, language concretized representations of the world, and without language, thought could not proceed.
In the 90s, Gibson began work on "difficult sentences" and discovered their propensity in many languages to minimize dependency distances between clauses.
In the 00s, Gibson used this insight to build towards a larger grammar rooted in the notion of dependency minimization with great power to explain nuances in word order across large sets of language families.
In the podcast, Gibson makes a compelling case against Chomsky's notion of a universal grammar, arguing that human language is a cultural artifact that nonetheless as generalizable features such as "head" and "dependent" structures.
Gibson's discussion with Bashir has interesting implications for our ongoing conversations about AI.
AI translates language to complex probabilistic equations. Each language family follows generalizable patterns. Even more abstractly, these equations suggest the dominance of a "head" and "dependent" structure."
And yet, we should not rush to the conclusion that our models are confirming the existence of Chomsky's universal grammar.
As mathematical descriptions of a communicative process, these equations only characterize what is happening, not why it is happening.
In order to get at the issue of nature vs. nurture, we researcher need recourse to more empirical studies as Gibson insists upon...
Thanks, Daniel, for this amazing interview!!!
In particular, I enjoyed thinking about music as a non-linguistic thought process!!!
I didn't know you played the violin. I grew up playing the cello!!!
In this podcast, Ted Gibson explorers an empirically grounded theory of language as communication.
In the 90s and 00s, the American post-structuralist humanities, fueled by the Whorf hypothesis, reconceptualized language as coextensive with thought. Many scholars made strong claims that language encompassed reality, language concretized representations of the world, and without language, thought could not proceed.
In the 90s, Gibson began work on "difficult sentences" and discovered their propensity in many languages to minimize dependency distances between clauses.
In the 00s, Gibson used this insight to build towards a larger grammar rooted in the notion of dependency minimization with great power to explain nuances in word order across large sets of language families.
In the podcast, Gibson makes a compelling case against Chomsky's notion of a universal grammar, arguing that human language is a cultural artifact that nonetheless as generalizable features such as "head" and "dependent" structures.
Gibson's discussion with Bashir has interesting implications for our ongoing conversations about AI.
AI translates language to complex probabilistic equations. Each language family follows generalizable patterns. Even more abstractly, these equations suggest the dominance of a "head" and "dependent" structure."
And yet, we should not rush to the conclusion that our models are confirming the existence of Chomsky's universal grammar.
As mathematical descriptions of a communicative process, these equations only characterize what is happening, not why it is happening.
In order to get at the issue of nature vs. nurture, we researcher need recourse to more empirical studies as Gibson insists upon...
Thanks, Daniel, for this amazing interview!!!
In particular, I enjoyed thinking about music as a non-linguistic thought process!!!
I didn't know you played the violin. I grew up playing the cello!!!
Nick
Thanks Daniel and Ted! Welcome to my project - General Intelligence System - https://www.linkedin.com/newsletters/natural-language-understanding-7020382304715911168/ - https://sigmoid.social/@MichaelMolin - https://vk.com/thematrixcom