摘要
德雷福斯曾经提出:人工智能的问题既有科技属性,更有哲学属性。因为全部人工智能工作都依赖一个哲学假设:即人类世界是由可清晰表征的始基构成的,而智能就在某种独立的清晰元素之中。但从海德格尔和梅洛-庞蒂的现象学哲学来看,上述理性主义假设是不能成立的。德雷福斯就以现象学为基础来构建他的人工智能批判。近年来,深度学习算法大放异彩,而人工智能似乎已经完全克服了德雷福斯曾经提出的哲学批判。然而,在深入探查近些年来的人工智能工作之后,我们不难发现:尽管人工智能的神经网络进路的确取得了很大的进步,但它依旧没有放弃人工智能表征主义进路的哲学假设,因而仍然无法摆脱德雷福斯的哲学批判。
Hubert L. Dreyfus suggested that problems of artificial intelligence have both technological and philosophical attributes,because all works of AI rely on a philosophical hypothesis that human world is made up of clear and representable primitives,and that human intelligence lies in independent and definitive elements. However,according to phenomenological philosophy of Martin Heidegger and Maurice Merleau-Ponty,the above rationalist hypothesis can’t be correct. Dreyfus based his critique of AI on phenomenology. In recent years,algorithms of deep learning have made great progresses,and AI seems to have completely overcome the philosophical critique which Dreyfus once proposed. However,after deep exploring of AI works in recent years,we can easily find that neural network approach has indeed made great progresses,but it still does not give up the philosophical assumption of representationalism approach. Therefore it still can’t avoid Dreyfus’ philosophical critique.
作者
徐献军
XU Xian-jun(Department of Psychology,College of Humanities,Tongji University,Shanghai 200092,China)
出处
《自然辩证法研究》
CSSCI
北大核心
2019年第1期15-20,共6页
Studies in Dialectics of Nature
基金
同济大学"欧洲研究"一流学科建设项目子课题"欧洲思想文化与中欧文明交流互鉴"
关键词
德雷福斯
人工智能
海德格尔
梅洛-庞蒂
神经网络
Hubert L. Dreyfus
artificial intelligence
Martin Heidegger
Maurice Merleau-Ponty
neural network
作者简介
徐献军(1979-),浙江金华人,哲学博士,同济大学人文学院心理学系教授,主要研究方向:哲学心理学、现象学精神病理学。
