【作 者】：
刘 凯， 贾 敏， 孙常新， 马玉慧， 王伟军
【关 键 词 】：
通用人工智能； 机器教育； 教学原则； 学习效果； AGI
【栏 目】：
理论探讨
【中文摘要】：
人工智能分为专用与通用两个分支，二者皆可与教育融合，但前者关注如何用人工智能手段解决教育问题，如智慧教育；后者却反向聚焦怎样用教育手段解决人工智能问题，如机器教育。既有研究已实证后者的必要性与可能性，本研究则通过实验探索人类教育经验对通用人工智能系统学习效果的影响，尝试验证“机器教育”的有效性。实验自变量来自人类教学过程的四类重要影响因素，分别是教学目标、教学内容、教学节奏和教学空间。结果发现，通用人工智能系统对教学目标的激活频率、教学内容的正确率、教学节奏的时间间隔以及教学空间的大小等指标具有与人类学习者高度相似的敏感性，都可由教学参数的调整而获得更好的学习效果。因此，人类教育经验同样可推广至通用人工智能系统。对“人机兼容”客观教育规律的确证，不仅用科学证据有力地回击了对教育理论科学性的质疑，亦有望实现教育学对人工智能研究的逆向反哺。
【英文摘要】：
Artificial Intelligence(AI) has been divided into two branches of Special purpose AI(SAI) and Artificial General Intelligence(AGI). Both can be integrated with education, but the former focuses on how to use AI to solve educational problems, such as intelligent education. The latter, on the other hand, focuses on how to solve AI problems by educational means, such as machine education. Existing studies have demonstrated the necessity and possibility of the latter, and this study explores the influence of human educational experience on the learning effect of artificial general intelligence systems through experiments, trying to verify the effectiveness of "machine education". The independent variables of the experiment come from four kinds of important factors in human teaching process, which are teaching goal, teaching content, teaching rhythm and teaching space. The results show that the artificial general intelligence system has highly similar sensitivity to the activation frequency of teaching objectives, the accuracy of teaching content, the time interval of teaching rhythm and the size of teaching space as human learners, which can be adjusted by the teaching parameters to obtain better learning results. Therefore, the human education experience can be generalized to artificial general intelligence systems as well. The confirmation of the objective education law of "human-machine compatibility" not only effectively counteracts the questioning of the scientific nature of educational theories with scientific evidence, but also is expected to realize the reverse feedback of pedagogy to AI research.