当期目录首页 > 在线期刊 > 当期目录
如何在通用人工智能系统中实现“知识锁定标杆浮动效应”——一种基于“时间压力”的简易心智模型发布时间:2019-08-30  点击数:
作 者:徐英瑾
关键词:通用人工智能;时间压力;知识锁定标杆浮动效应;比对主义;语境主义;不变主义;固知需求
摘 要:

 

所谓“知识锁定标杆浮动效应”,是指面对同样的信念内容,认知主体会在某些环境下将其判定为“知识”,而在另一些环境下将其判定为“非知识”。该效应的存在,使得人类能够根据环境信息的变化,灵活地改变自身信念,更好地适应环境。人工智能体对于该效应机制的模拟,也能够更好地适应环境。不过,这种模拟必须建立在对于该效应的正确理解上,而西方主流知识论学界对于该效应的解释,如语境主义、比对主义与不变主义提出的解释,要么缺乏足够的普遍性,要么本身建立在一些有待解释的概念上。与之相比较,基于“时间压力”的模型,则将智能体的知识指派倾向的强度视为与其感受的时间压力彼此负相关的一项因素,而“时间压力”本身被视为“主体所预估的问题解决所需要的时间”与其“所愿意付出且能够付出的时间”之间的差值。这样的模型不仅能够对所谓的“银行案例”与“斑马案例”作出简洁的解释,而且在原则上可以被算法化。

 

How to Simulate Epistemic Shifts in a General Artificial Intelligence System?

A Cognitive Model Based on the Notion of “Time Pressure”

 

Xu Yingjin (Fudan University)

 

Abstract The term “epistemic shifts” refers to the phenomenon that knowledge ascribers would ascribe different epistemic statuses to same beliefs under different internal/external conditions. Hence, one belief would be judged as knowledge in one circumstance, and as non-knowledge in another. The existence of this phenomenon makes human subjects be able to respond to environmental changes in a flexible manner, and hence, an Artificial General Intelligence (AGI) system is also expected to simulate this function. But this simulation is impossible if there is no convincing high-level explanation of “epistemic shifts”, whereas the mainstream epistemological explanations of it (which are provided by contextualism, contrastivism, invariantism, etc.) are either too ad hoc or assuming notions which are more troublesome than the explanans in question. My competing theory appeals to the notion of “time pressure”, and the intensity of time pressure is supposed to be inversely proportional to the intensity of the disposition of attributing knowledge to the target belief. Time pressure is construed in terms of the numerical difference between the estimated time needed by the completion of the task and the time that the subject can and wants to spend to complete the same task. And such account can be algorithmically treated to fit the theoretical requirement of AGI.

Key words Artificial General Intelligence (AGI); time pressure; epistemic shifts; contrastivism; contextualism; invariantism; need-for-closure

 

作者简介 徐英瑾,哲学博士,复旦大学哲学学院教授、博士生导师;上海 200433。


[PDF](下载数: