摘要
符号奠基问题的核心是要设法使符号具有关联于外部世界的意义。随着过去解决方案的接连失败,该问题的框架也在反思中得到了不断完善,其中就包括“零语义承诺条件”和“错误表征条件”的提出。然而在实际的研发过程中,零语义承诺条件却一直难以得到贯彻;不仅如此,尽管错误表征条件要求智能体能够自主识别并修正表征的错误,但来自外部的评估始终却都是一个不可避免的要素。因此,我们有理由怀疑在严格的零语义承诺条件与错误表征条件之间存在着不兼容性,有必要考虑适当放宽前者对符号奠基问题的约束。
The symbol grounding problem(SGP) can be roughly expressed as:how to enable an agent to give symbols semantic content by its own mechanism without relying on external observer.With the successive failures of past solutions,the framework of SGP has also been continuously improved in reflection,which includs the addition of “zero semantical commitment condition”(ZSCC) and “misrepresentation condition”(MC).However,in actual research,the ZSCC is found difficult to obey.Not only that,when the ability of misrepresentation requires that the agent can identify and correct from representational errors automatically,external evaluation standards are always indispensable,thus breach the ZSCC.So it can be concluded that the strict ZSCC is incompatible with the MC,and therefore it is necessary to appropriately relax the former's constraints on the SGP.
作者
张洋
ZHANG Yang(School of Marxism,South China University of Technology,Guangzhou 510640,China)
出处
《系统科学学报》
北大核心
2023年第3期29-34,共6页
Chinese Journal of Systems Science
关键词
符号奠基问题
零语义承诺条件
错误表征
指称的因果理论
symbol grounding problem
zero semantical commitment condition
misrepresentation
causal theory of reference