Photonic technologies for implementing neuromorphic computing: results, research directions, and AI hardware

Making use of photonic technologies significantly transforms an entire area of high-performance computing by offering new generation of ultra-fast computing systems, overcoming limitations pertinent to traditional electronic hardware in terms of energy efficiency, parallelism, and scalability. Such unique properties of light as an ensured high throughput, low time delay, and capability of parallelizing large data arrays and streams, photonic integrated circuits (PICs) offer a promising alternative for performing complex computations having a direct impact on artificial intelligence (AI) development. This review summarizes recent advances in photonic neuromorphic architectures (analog optical computing, photonic tank computing), and provides some insight into future research.