类脑智能研究的回顾与展望

Retrospect and Outlook of Brain-Inspired Intelligence Research

人工智能学科诞生以来,实现人类水平的智能系统便是本学科探索的长期目标.然而经历了近60年的发展,目前还没有任何一个通用智能系统能够接近人类水平:具有协同多种不同的认知能力;对复杂环境具备极强的自适应能力;对新事物、新环境具备自主学习的能力等.随着脑与神经科学、认知科学的发展,在不同尺度观测各种认知任务下脑神经网络的部分活动并获取相关数据已成为可能.因此,受脑工作机制启发,发展类脑智能成为近年来人工智能与计算科学领域研究的热点.类脑智能是以计算建模为手段,受脑神经机制和认知行为机制启发并通过软硬件协同实现的机器智能.类脑智能系统在信息处理机制上类脑,认知行为和智能水平上类人,目标是使机器实现各种人类具有的多种认知能力及其协同机制,最终达到或超越人类智能水平.文中将从脑科学、认知科学、人工智能研究交叉的视角回顾与类脑智能研究有关的历史、现状与研究焦点,并展望该研究领域的发展方向、可能的应用领域及其潜在的深远影响.

Creating human-level intelligent system is the long-standing mission for the field of Artificial Intelligence （AI） since its establishment nearly 60 years ago. Until now, however, there is still no general purpose intelligent system which can reach the human intelligence level in terms of coordinating various cognitive behaviors, adaptability of complex environments, and autonomous learning under new environments. With the advancement of Brain Science, Neuroscience, and Cognitive Science, it is now possible for partially observing and obtaining data on the activities of brain neural networks at multiple scales while they are conducting various cognitive tasks. Hence, Brain-inspired Intelligence （Brain-inspired AI） is becoming a focus and promising trend of AI. Brain-inspired Intelligence is a field for creating machine intelligence through computational modeling and software-hardware coordination built with inspirations from the brain and human cognition. Brain-inspired intelligent system is similar to brain in information processing mechanisms and to human in cognitive behavior and intelligence level. Its goal is to realize various human cognitive functions as well as their coordination mechanisms by machine through brain-inspired principles, and eventually reach and go beyond human-level intelligence. This paper introduces the past and current status of brain-inspired AI, the main research focuses, the future directions and applications, and its potential profound influence to the society.