← Home
Christopher Manning from Stanford discusses the intersection of linguistics and large language models — whether LLMs truly understand language or process it as statistical patterns, and what this distinction means for the field.
This summary was generated from show notes and public descriptions, not from a full transcript review. Details may contain inaccuracies.
Canon
•
Manning traces the history: rule-based NLP researchers resisted statistical methods in the 1990s. Statistical NLP researchers resisted deep learning in the 2010s. Deep learning researchers are now debating whether scaling alone can produce understanding. Each generation resists the paradigm shift that makes their expertise less relevant.
Highlights
•
LLMs have learned more about language structure than any previous AI system — the debate is whether statistical learning constitutes understanding
Manning argues that LLMs have learned syntactic structures, semantic relationships, and pragmatic conventions that previous NLP systems could not capture. Whether this constitutes understanding depends on how you define the term — and linguists and computer scientists define it differently.Was this useful?