论文部分内容阅读
人工智能追求的长期目标是使机器能像人一样学习和思考。由于人类面临的许多问题具有不确定性、脆弱性和开放性,任何智能程度的机器都无法完全取代人类,这就需要将人的作用或人的认知模型引入到人工智能系统中,形成混合-增强智能的形态,这种形态是人工智能或机器智能的可行的、重要的成长模式。混合-增强智能可以分为两类基本形式:一类是人在回路的人机协同混合增强智能,另一类是将认知模型嵌入机器学习系统中,形成基于认知计算的混合智能。本文讨论人机协同的混合-增强智能的基本框架,以及基于认知计算的混合-增强智能的基本要素:直觉推理与因果模型、记忆和知识演化;特别论述了直觉推理在复杂问题求解中的作用和基本原理,以及基于记忆与推理的视觉场景理解的认知学习网络;阐述了竞争-对抗式认知学习方法,并讨论了其在自动驾驶方面的应用;最后给出混合-增强智能在相关领域的典型应用。
The long-term goal of artificial intelligence is to make the machine learn and think like a human being. Due to the uncertainty, vulnerability and openness of many problems facing human beings, no machine of any intelligent level can completely replace human beings. Therefore, it is necessary to introduce human functions or human cognitive models into an artificial intelligence system to form a hybrid - Enhance the form of intelligence that is a viable and important growth model for artificial intelligence or machine intelligence. Hybrid - enhanced intelligence can be divided into two basic types: one is man-machine collaboration in man-machine loop to enhance intelligence, and the other is to embed cognitive model into machine learning system to form cognitive intelligence-based hybrid intelligence. This article discusses the basic framework of hybrid-enhanced intelligence, and the basic elements of hybrid-enhanced intelligence based on cognitive computing: intuitionistic inference and causal model, memory and evolution of knowledge. In particular, we discuss the application of intuitionistic reasoning in solving complex problems Function and basic principle, as well as the cognition learning network based on the understanding of the visual scene of memory and inference; expounds the competitive-adversarial cognitive learning method and discusses its application in automatic driving; finally, Typical applications in related fields.