Deep learning has an unquenchable thirst for processing resources. Classical deep learning has silently developed a bottleneck due to the interference of electrical impulses, energy usage, and physical constraints, even though electronic parts established on silicon can still sustain it presently. Academic and corporate circles are attempting to find new approaches for resolving electrical flaws that are less computationally intensive. It has huge benefits in information transmission and OC due to its high speed of 300,000 km per second, which is 300 times faster than that of an electron, and its data-carrying capacity and variety, which is 2 × 104 times greater than that of electric channels, as well as high parallelism and strong anti-interference. Switching electricity with light has emerged as a viable and sustainable work style, following the current trend.