factual

LLMs as evidence on Universal Grammar

Large language models, by acquiring fluent language from statistical exposure without an innate Chomskyan Universal Grammar, provide a data point suggesting that language can be learned without a dedicated built-in grammar module, but because transformers differ architecturally from the brain, they cannot definitively settle the linguistics debate about how humans acquire language.

factualpending

Speaker

Sean Carroll

Evidence Quote

“now LLMs do a convincing job at acting like they know language”

Source

AMA | October 2025— Sean Carroll's Mindscape
Created: 6/13/2026, 12:25:24 AM

Your notes

Loading notes...