论文部分内容阅读
近年来人工智能研究的许多重要进展反映了一个趋势:来自脑科学的启发,即使是局部的借鉴都能够有效地提升现有人工智能模型与系统的智能水平。然而,想要真正逼近乃至超越人类水平的人工智能,还需要对脑信息处理机制更为深入的研究和借鉴。类脑智能研究的目标就是通过借鉴脑神经结构及信息处理机制,实现机制类脑、行为类人的下一代人工智能系统。文章从受脑启发的新一代人工神经网络、基于记忆、注意和推理的认知功能模型、基于生物脉冲神经网络的多脑区协同认知计算模型等角度,并结合研究团队在类脑智能领域的研究进展,论述类脑智能的研究进展、发展方向和对未来发展的思考。
In recent years, many important advances in artificial intelligence research reflect a trend: inspired by brain science, even partial reference can effectively enhance the intelligence of existing AI models and systems. However, artificial intelligence that really wants to approach or even surpass humankind needs further research and reference on brain information processing mechanism. The goal of brain-like intelligence research is to realize the next-generation artificial intelligence system of mechanism-like brain and behavior-like person by drawing lessons from brain nerve structure and information processing mechanism. The article from the brain-inspired new generation of artificial neural networks, cognitive function model based on memory, attention and reasoning, multiple implicit neural network based cognitive cognitive model and other aspects of the brain, combined with the research team in brain-like intelligence Research progress, discusses the research progress of brain intelligence, development direction and thinking of the future development.