像教育人一样教育机器——人类教学原则能用于通用人工智能系统吗?  

Educate Machine as Human:Can Human Teaching Principles be Available for Artificial General Intelligence System?

作　　者：刘凯 贾敏 黄英辉 胡祥恩[3] 王培[4] LIU Kai;JIA Min;HUANG Yinghui;HU Xiangen;PEI Wang(Department of Educational Science, Bohai University, Jingzhou 121013, China;Institute of Artificial General Intelligence of Bohai University, Jingzhou 121013, China;Department of Psychology, Central China Normal University, Wuhan 430079, China;Department of Computer and Information Sciences, College of Science & Technology, Temple University, PA Philadelphia 19122, USA)

机构地区：[1]渤海大学教育科学学院,辽宁锦州121013 [2]渤海大学通用人工智能研究所,辽宁锦州121013 [3]华中师范大学心理学院,湖北武汉430079 [4]天普大学计算机与信息科学系,美国宾夕法尼亚州费城19122

出　　处：《开放教育研究》2022年第2期11-21,共11页Open Education Research

摘　　要：通用人工智能旨在探索智能的一般规律,使其软件系统具有感知、记忆、情感、推理与决策等多种认知功能,乃至拥有与人类学习者相似的学习能力。本文将通用人工智能系统作为机器学习者,考察人类教学原则对机器学习者的教育干预效果,探究机器教育的可能性。受控实验结果显示,教学时长、教学内容及教学方法的变化都能引发通用人工智能系统的记忆网络结构显著改变。这不仅证实了人类教学原则迁移至通用人工智能系统的可能性,也折射出存在贯通“人—机”的普适的、可证伪的教育客观规律。在人工智能等新兴技术赋能教育的时代背景下,教育学的原生创新仍相对乏力。通用人工智能系统的出现不仅丰富了教育学研究对象,也拓宽了教育学研究边界。在通用人工智能框架下,认知机器不但能够而且也应该受到教育,用人类教育手段反哺人工智能将成为教育科学今后的重要研究课题之一,教育科学也有望成为促进科学发展的新动力。Aiming to explore the general laws of intelligence,artificial general intelligence(AGI)has many human-like cognitive functions such as perception,memory,emotion,reasoning,decision-making,etc.This article takes the AGI System as a machine learner to investigate whether the pedagogical effect of human teaching principles could be found in the machine learners.The results of controlled experiments show that changes,including teaching time,teaching content,and teaching methods,lead to significant changes in the network structure of memory of the AGI System.It not only confirms the possibility of transferring human teaching principles to the AGI System,but also reflects the existence of universal and falsifiable educational rules beyond human and machine.In the high-tech era,Artificial Intelligence(AI)has become a key external source of innovation,but the internal source in education is often in a weak position.However,AGI expands the research objects of pedagogy,and broadens the boundaries of pedagogical research.Taking AGI into consideration,machine learners not only can be but also should be educated,which offers educational science an opportunity to manifest the power of human education to enrich AI reversely.Anyway,education science is well on the way to standing in the center of our science.

关 键 词：教育 通用人工智能 机器教育 教学原则 记忆网络结构 

分 类 号：G434[文化科学—教育学]