从语言数据到语言智能——数智时代对语言研究者的挑战

From Language Data to Linguistic Intelligence:Challenges for Language Researchers in the Age of Data-Based Intelligence

数基人工智能不仅破解了人类的“操作系统”,也正重新定义人类的知识,全方位挑战着人类的认知。数智时代的语言学家和认知科学家需要回答两个基本问题:为什么用我们能理解的方式,机器做不好?为什么机器能这样做,我们却理解不了?要想解开数据为什么会涌现智能之谜,我们可能需要回到真实文本,回到语言的概率性本质。本文认为在“数据→模式→知识→网络→智能”链中,语言研究的重点可放在模式和网络两个环节上,从而为构建可解释的人工智能贡献学科智慧。

Data-based artificial intelligence not only has cracked the human“operating system”,but also is redefining human knowledge,challenging human cognition in all aspects.Linguists and cognitive scientists in the age of data-based intelligence need to go back to two basic questions:Why can’t machines do it well in ways we can understand?Why can machines do it in a way that we can’t understand?To solve the mystery of why data emerges with intelligence,we may need to return to real texts,to the probabilistic nature of language.The article argues that in the“Data→Patterns→Knowledge→Network→Intelligence”chain,the focus of language research can be placed on patterns and networks,thereby contributing disciplinary wisdomto the construction of explainable artificial intelligence.