通用人工智能时代关于语言研究的思考  

Reflections on Linguistic Study in the Age of Artificial General Intelligence

作　　者：毕玉德 闫丹辉 BI Yude;YAN Danhui

机构地区：[1]复旦大学,上海200433 [2]国防科技大学,南京210039

出　　处：《语言文字应用》2024年第3期119-131,共13页Applied Linguistics

基　　金：国家社科基金项目“朝鲜语语言知识结构化构建及深度语义理解研究”(22BYY163);国防科技大学外国语学院2023年度青年项目“基于修辞结构理论的军事领域自然文本与大模型伪造文本对比研究”(23A09)的资助。

摘　　要：本文综述了不同学科研究人员对以Chat GPT为代表的大语言模型的认识,认为虽然大语言模型展现出了令人印象深刻的语言能力,但是本质上仍然属于语言模型的范畴,只能在大规模数据基础上对人类的语言行为进行模仿,并没有真正地理解自然语言。本文探讨如何认识大语言模型的语言能力、人工智能时代语言学研究的任务和意义等问题,认为通用人工智能时代的语言学研究应采取不同研究范式互相促进的思路,进一步推动对人类语言本质的认识。This paper reviews the understanding of researchers from different disciplines on large language models represented by ChatGPT.It suggests that although large language models have demonstrated impressive language capabilities,they fundamentally still belong to the category of language models.They can only imitate human language behaviors based on large-scale data,without truly understanding natural language.This study explores issues such as understanding the language capabilities of large language models,the tasks and significance of linguistic research in the era of artificial intelligence.This paper argues that linguistic research in the era of general artificial intelligence should adopt the approach of mutually reinforcing different research paradigms to further advance our understanding of the nature of human language.

关 键 词：大语言模型 人工智能 语言研究 研究意义 研究范式 

分 类 号：H0-05[语言文字—语言学]