This article first introduces how ChatGPT discards the working paradigm of natural language processing(NLP)and challenges linguistic beliefs.It then discusses the critical and groundbreaking conclusion which was made by comparing the mechanisms of modern large language models with the research path of generative linguistics in Piantadosi(2023):modern machine learning has subverted and bypassed the entire theoretical framework of Chomsky's approach.To facilitate reading and understanding,this article focuses on the following three questions to introduce and evaluate Piantadosi's(2023)work:1)Can modern language models be considered genuine theories of language?2)Can modern language models and their principles be attested under linguistic scrutiny?3)Can modern language models and their performance refute Chomsky's linguistic claims?Finally,the article suggests that in the face of the outstanding performance of modern large language models in language application,we should abandon the methods of tenacity,authority,and a priori,and instead adopt a method of science that involves confronting problems,embracing skepticism,and continuously experimenting and validating.
ChatGPTmodern/large language modelsChomskyan/Generative Linguistic Claimsthe method of science
NETL
NSTL
维普
万方数据
