跳到正文
arXiv cs.CL· Peter Wallis·· 9 小时前AI 评分22

观点:当前大语言模型并不“理解”语言,而是“做”语言

Communication, Vocalisation, and Language

AI 导读

一篇 arXiv 论文提出,当前这批大语言模型并不“理解”语言,而是“做”语言;它们并非形成符号的“分布式表示”,而是捕获“文化智能”,并将符号扎根于社会学家所称的“实践”之中。该文为作者与 Roger 的合著论证,共 6 页并含一张 Venn 图。

正文 · 原文

View PDF HTML (experimental)

Abstract:The contention is that the current batch of Large Language Models don't *understand* language, but they *do* it. Rather than LLMs forming some kind of ``distributed representation'' of symbols, the claim is that LLMs are capturing ``Cultural Intelligence'' and grounding symbols in what the sociologists call practices.
Comments: A written version of my argument with with Roger, with input from Bruce. 6 pages and one (Venn) diagram
Subjects: Computation and Language (cs.CL)
Cite as: arXiv:2610.16490 [cs.CL]
  (or arXiv:2610.16490v2 [cs.CL] for this version)
  https://doi.org/10.48550/arXiv.2610.16490

arXiv-issued DOI via DataCite

Submission history

From: Peter Wallis PhD [view email]
[v1] Mon, 12 Oct 2026 09:43:18 UTC (46 KB)
[v2] Tue, 17 Nov 2026 09:19:18 UTC (46 KB)

来源:arXiv cs.CL · arxiv.org