Abstract: Currently, brain-inspired models, represented by artificial neural networks, are primarily implemented at the software level. Although these models have achieved significant breakthroughs in terms of intelligence, their development is still constrained by the physical limitations of electronic chips. With the advancement of optical computing technology, photonic chips based on analog optical computing can directly construct physical neuromorphic computing units at the hardware level, enabling efficient intelligent processing and adaptive learning. This paper first provides an overview of recent research progress on photonic chips in the fields of digital optical computing and optical quantum computing, highlighting that developing brain-inspired photonic chips based on analog optical computing can circumvent the issue associated with logic gate design and simulate the computational advantages of the brain. By reviewing the evolution of artificial neural networks and the corresponding optical implementation technologies, this paper further presents the following viewpoints regarding the bottlenecks in existing brain-inspired photonic chips, specifically in nonlinear components and scalability. In terms of hardware implementation, it is necessary to further explore nonlinear optical components and construct all-optical nonlinear operation layers to simulate the nonlinear characteristics of the brain. In terms of algorithm implementation, it is crucial to study the memory-based learning and cognitive principles of the brain to design brain-inspired intelligent algorithms that match photonic characteristics, thereby overcoming scalability limitations.

Key words: Photonic Chip, Optical Neural Network, Photonic Synapse, Brain-Inspired Photonic Chip, Brain-Inspired Algorithm