ChatGPT与意象感知问题:再论塞尔的人工智能语言批判  

ChatGPT and the Image Problem:A Re-examination of Searle's Critique of Al Language

作　　者：张杨[1] Zhang Yang(School of Philosophy,Heilongjiang University,Harbin 150001,China)

机构地区：[1]黑龙江大学哲学院,哈尔滨150001

出　　处：《外语学刊》2023年第6期19-24,共6页Foreign Language Research

基　　金：黑龙江省社科规划项目“AI技术时代的公共领域重构问题研究”(19ZXD185);黑龙江大学研究生创新科研项目“约翰·罗斯金美学思想研究”(YJSCX2022-015HLJU)的阶段性成果。

摘　　要：约翰·塞尔的中文屋论证是人工智能语言批判的经典思想实验。塞尔的理路接近后期维特根斯坦对理性主义还原论的批判,他也继承了奥斯丁的言语行为理论。由此,中文屋论证的构思要点触及语言的意象特质,其基础理论表现出亲近整体论语言观、抵对还原论语言观的倾向。塞尔依据中文屋论证揭示出人工智能、尤其是自然语言处理的发展瓶颈,即不论数字计算机与人的语言交流多么顺畅,它的语言表达始终是基于运算,而不是基于对意象的理解。ChatGPT的成功看似对塞尔的观点构成挑战,但它实际上并没有表现出感知语言意象的能力,也就没有跳脱出塞尔的批判范畴。Chat-GPT与人类高效互动的意义在于倒逼人类丰富自身的语言意象,从而揭示出人工智能发展的另一种可能性。John Searle's Chinese Room Argument is a classic thought experiment for criticizing the Al language technique.Searle's theory is close to Wittgenstein's criticism of rational reductionism in his later period,and he also inherited Austin's Speech Act Theory.The main idea of this experiment is the image feature of language,and the basic theory of it shows the inclination of being close to holism view of language and opposing the reductionism view of language.Searle reveals the dilemma of artificial intelligence,especially natural language processing through this experiment.That is to say,no matter how smooth the language communication between digital computers and humans is,its language expression is always based on operations,not on the understanding of images.ChatGPT's success seems to pose a challenge to Searle's position,but in fact,it still does not break through Searle's critic.It still does not show the ability of image perception.ChatGPT's efficient interaction with humans,however,forces human beings to enrich their own language imagery,which indicates another possibility of the development of artificial intelligence.

关 键 词：ChatGPT 语言 意象 人工智能 中文屋论证 言语行为理论 

分 类 号：B089[哲学宗教—哲学理论]