反思智能：存在自主动机的最简机器模型设计

Reflective Intelligence: The Simplest Machine Model Design with Autonomous Motivation

摘要: 本文从“智能”定义反思的角度，反思人对“机器情绪”的认识。如果机器本身不存在一套“自我情绪体验”的机制，那么，机器对人或动物的情绪识别，终究是人对其他人或动物的情绪识别的延伸，机器不可能产生源自自身对人或动物情绪感知的“共情”。本文用内省、思辨、常识推理的方法，提出一种有别于图灵测试的智能测试方法，并针对此测试方法，构建出一个“存在喜恶之情”的最简抽象模型。该模型的“性质体数量”和“性质体的状态数量”都极少，但足够说明构成智能心理的每个抽象性质体的具体作用。从工程实现的角度上说，在现有计算技术的支持下，按模型位置给出的性质，设计出采集、加工数据的方法，然后一步步丰富数据，便可一步步使该模型模拟的动物越来越像“有自主动机”、乃至表面看起来更像“有自我意识”的动物。

Abstract: From the perspective of “intelligence” definition, this paper rethinks people’s understanding of “machine emotion”. If the machine itself does not have a set of “emotional experience” mechanism, then, emotion recognition machine to human or animal, it is the extension of emotion recognition of other human or animal, the machine could not produce from itself to the human or animal emotions “empathy”. This paper puts forward an intelligent test method which is different from Turing test by introspective, speculative and commonsense reasoning, and constructs a simple and abstract model of “love and hate” based on this test method. The number of properties and number of states of the body is very few, but it is enough to explain the concrete function of each abstract property of intelligent mind. From the view of engineering that the ex-isting computation technology, according to the nature of the design model of the position, method of collecting and processing data, then data can be enriched step by step, step by step, make the model animal more like “self motivation”, but to the surface looks more like “self con-scious” animal.