大语言模型在哪里挑战了语言学?  

Where Do Large Language Models Challenge Linguistics?

作　　者：石锋[1] Shi Feng

机构地区：[1]南开大学文学院,天津300071

出　　处：《语言战略研究》2025年第1期87-96,共10页Chinese Journal of Language Policy and Planning

摘　　要：“人工智能教父”辛顿对乔姆斯基的批评,值得语言学者思考:大语言模型究竟在哪里挑战了语言学?本文讨论以下问题:(1)搞大语言学还是小语言学?这一问题涉及语言学研究对象拓展和研究范式转换。当前特别需要把小语言学观念转变为大语言学观念,建立基于数据和概率统计的多学科、跨领域的科学观。大语言学向外融合文、理、医、工等多学科,向内跨越语音、语法、语义、语用等多领域,海阔天空,大有作为。(2)语言和思维可分还是不可分?人类的思维可以离开语言,语言不可离开思维。思维是为了交流,没有新信息,思维会失去活力而枯竭。不能人为地把思维和交际分离开。(3)语言习得是先天的,还是经验的?人工智能弃用乔姆斯基的语言先天论,转而基于语言经验论,取得了里程碑式的成功。(4)人工智能会不会有思维,甚至有生命?人工智能不会具有生命。人工智能的语言是离开思维的语言。流利的语言并不等于自主的思维。离开人类智能的主宰操控,人工智能将一事无成。我们要学会驾驭人工智能,适应这个有了人工智能的世界,去创造更加美好的未来。The critique of Chomsky by Geoff rey Hinton,often referred to as the“Godfather of Artifi cial Intelligence”,raises profound questions about the intersection of linguistics and artifi cial intelligence.This paper explores four critical dimensions where large language models are fundamentally challenging traditional linguistic paradigms:(1)Should we pursue macro or micro linguistics?This debate fundamentally concerns the expansion of linguistic research objects and methodological transformations.The current academic landscape urgently requires a conceptual shift from a narrow,discipline-confined linguistic perspective to a comprehensive,interdisciplinary approach.The proposed“macro linguistics”advocates that macrolinguistics can integrate with disciplines such as humanities,science,medicine,and engineering,and span various domains like phonetics,grammar,semantics,and pragmatics,off ering vast potential.(2)Are language and thought inseparable or distinct?The question challenges traditional dichotomies,and the current research argues that Human thought cannot be entirely separated from language.Communication is the primary purpose of thought and thought loses vitality without novel information.Artifi cial attempts to disconnect thought from communication are counterproductive.(3)Is language acquisition innate or experiential?Artifi cial intelligence has signifi cantly challenged Chomsky’s innateness hypothesis by abandoning the theory of language as a pre-programmed,innate capacity and embracing language as fundamentally experiential,achieving milestone successes through experience-based learning models.(4)Can AI possess thought or even life?The paper presents a nuanced perspective on AI’s cognitive capabilities.We argue that artifi cial intelligence cannot possess life in the biological sense.AI’s language is fundamentally disconnected from genuine thought,linguistic fl uency does not equate to autonomous thinking and AI remains dependent on human intelligence and control.Without human intelligence to

关 键 词：人工智能 大语言学 自主思维 概率匹配 复杂适应系统 

分 类 号：H002[语言文字—语言学]