辛顿与乔姆斯基在语言知识来源问题上的分歧  

Disagreement Between Hinton and Chomsky on the Source of Linguistic Knowledge

作　　者：丁晓梅[1] 冯志伟 DING Xiaomei;FENG Zhiwei

机构地区：[1]大连海事大学外国语学院,辽宁大连116026 [2]新疆大学中国语言文学学院,新疆乌鲁木齐830046

出　　处：《外语研究》2025年第2期25-33,F0003,共10页Foreign Languages Research

基　　金：国家社科基金项目“生成式人工智能赋能语言学多模态学科知识图谱建设与应用研究”(编号:24BYY079)的阶段性成果。

摘　　要：本文简述了大语言模型的发展历程,分析了大语言模型的机理和大语言模型的局限性,指出了辛顿和乔姆斯基在语言知识来源问题上的分歧,说明了所罗门诺夫归纳法是大语言模型的理论基础,建议把数字智能与人类智能结合起来,构建人机协作共同体。本文认为,语言学又将面临一场“人工智能革命”,这是继“乔姆斯基革命”之后的又一次语言学革命,这场人工智能革命将比“乔姆斯基革命”更加深刻,它将重塑语言学的面貌,把语言学的发展推向一个崭新的阶段。This article introduces the development history of large language models,analyzes their mechanisms and limitations,points out the disagreement between Hinton and Chomsky on the source of linguistic knowledge,explains that Solomonoff's inductive method is the theoretical foundation for large language models,and suggests combining digital intelligence with human intelligence to build a human-machine collaborative community.The article holds that linguistics will face another“artificial intelligence revolution”,which is another linguistic revolution after Chomsky's revolution.This artificial intelligence revolution will be more profound than Chomsky's revolution,reshaping the face of linguistics and pushing its development to a new stage.

关 键 词：大语言模型 数字智能 人类智能 语言机能 所罗门诺夫归纳法 人机协作共同体 

分 类 号：H030[语言文字—语言学]