论深度学习的语用特征

On the Pragmatic Characteristics of Deep Learning

深度学习的根本困境来源于真实世界中广泛存在的语境问题。一个有效的深度学习系统,只能执行某类特定功能,无法跨越类别去处理多个领域的任务,具有明显的语用特征。为完成某个特定的语用任务而设计的深度学习系统难以适应纷繁复杂的真实世界,因此也无法具有真正的智能,更不可能实现强人工智能。

The fundamental dilemma of deep learning originates from the contextual problems in the real world. An effective deep learning system can only perform a certain kind of function,and cannot deal with tasks in other fields across categories,showing obvious pragmatic characteristics. The deep learning system designed to accomplish a specific pragmatic task is difficult to adapt to the complicated real world and therefore cannot have real intelligence,let alone realize strong artificial intelligence.