复杂性视阈下的当代人工智能发展——以深度学习为例

The Development of Contemporary Artificial Intelligence from the Perspective of Complexity-Take Deep Learning as An Example

人工智能在过去的六十年时间里呈现出波浪式发展特征,深度学习近年来革命性的突破更是引人瞩目。本文立足复杂性视角,以复杂性认识论和方法论在该领域的运用为线索,对深度学习革命性突破的成因给出崭新诠释——深度学习呈现出的革命性能力是作为复杂系统的深度学习模型涌现性的结果;基于同样的复杂性视角,本文还对深度学习有待解决的可解释性问题、泛化问题、算理不明问题之深层原因给出独特解释。最后,基于深度学习发展所呈现出的得失,还对复杂性方法论运用过程中如何处理整体与局部、简单性与复杂性方法进行了探究。

Artificial intelligence has shown the characteristics of wavy development in the past 60 years,and the revolutionary breakthrough of deep learning in recent years is even more remarkable.From the perspective of complexity and taking the application of complexity epistemology and methodology in this field as clues,this paper gives a new interpretation of the causes of the revolutionary breakthrough of deep learning-the revolutionary ability of deep learning is the result of the emergence of deep learning model as a complex system;Based on the same complexity perspective,this paper also gives a unique explanation for the underlying causes of interpretable problems,generalization problems,and computational ambiguity problems that need to be solved in deep learning.Finally,based on the gains and losses of the development of deep learning,it also explores how to deal with the whole and part,simplicity and complexity methods in the application of complexity methodology.