通用人工智能需要在私人语言的层面上进行知识表征吗——来自大森庄藏的启发

Is It Necessary to Construct Private Language in the Knowledge Representation of an Artificial General Intelligence System?Some Remarks Inspired byŌmori Shōzō

通用人工智能语境中的私人语言,指的是这样一个意思:表征A在系统甲那里的知识表征方式与同一个表征在系统乙那里的表征方式必然会有所差异。因此,在预设推论主义语义学自身有效性的前提下,A在甲中的意义集,总会有一个子集(无论这一子集有多小)仅仅为甲自身所拥有,而无法被任何一个别的系统所拥有。因此,任何一个与甲不同的别的系统,都无法彻底地理解甲对于A的意义把握方式。很显然,这样一种将机器表征与哲学史上的“第一人称哲学”传统相互结合的思路,是无法见容于后期维特根斯坦对于私人语言的著名反驳的。而为了与后期维特根斯坦论战,日本哲学家大森庄藏的思想资源便具有了很高的引用价值,因为他本身的哲学就可以被视为“维特根斯坦的话语方式与胡塞尔的思想内核”的日本式混合体。在对大森的哲学进行面向机器表征问题的重建的过程中,对于局域性原则与历史性原则的引入也是题中应有之义,以便为通用人工智能语境中建设私人语言的必要性提供辩护。而非公理性推理系统(纳思系统)所提供的技术手段,则会为这种想法的技术落地提供可能。

In the context of Artificial General Intelligence(AGI),the existence of private language implies that how a certain symbol is presented in systems’inner data-bases is different from one specific AGI system to another.Accordingly,when the validity of the“inferentialist semantics”is assumed,for any symbol S and two different systems A and B,the set of meaning-encoding units concerning S in A has to embrace a subset which is only a part of A’s data-base but not of B’s.Hence,for any system other than A,it cannot fully understand how S is construed in A’s case.The revival of the idea of private language in AGI is definitely conflicting with later Wittgenstein’s criticism of the possibility of private language.For refuting Wittgenstein,Ōmori Shōzō’s insights may be fairly helpful here,given that his own philosophy can be viewed as a hybrid of Wittgensteinian terminology and a Husserlian core.In the process of reconstructingŌmori’s arguments,the“principle of locality”and the“principle of historicity”will also be introduced to strengthen the necessity of bringing the idea of private language into AGI.As to how to technically realize the very idea in AGI,Non-Axiomatic Reasoning System(NARS)may provide a promising approach.