Abstract

The development of artificial intelligence (AI) that could rival their biological counterparts is the long-cherished dream of human. Recent progress in AI (such as AlphaGo) based on software simulation has shown its capability of handling cognitive tasks in restricted specific fields. However, these systems still suffer from the lack of generalizability and the ability to transfer learned knowledge from one task to another. Learning from information processing mechanisms of the brain, development of scalable hardware elements intrinsically capturing neuromimetic dynamic processes is clearly a promising way forward in building more general machine intelligence.¹

© 2017 Optical Society of America