Integrated Photonic Computing beyond the von Neumann Architecture

Abstract

In the context of a doomed end of the Moore's law, various new types of computing architectures have been emerging, aiming to meet the demands of intractable computation and artificial intelligence. Photonic computing is a competitive candidate, in light of the inherent properties of photons, including high propagation speed, strong robustness, and multiple degrees of freedom to encode information. Also, the progress of integrated photonics continues to provide novel possibilities, apart from boosting the scalability and stability of photonic computing architectures. Moreover, an introduction of quantum technology might open a new chapter for photonic computing, from the view of single photons. In this Perspective, we highlight the unique features and advances of integrated photonic platforms, whose roles in constructing a non-von Neumann computing architecture are also outlined. We show their potential in solving problems beyond the reach of traditional computers and in machine learning and further discuss the conceivable challenges and opportunities.

Tuneable modulating metasurface at visible and near-IR wavelengths

“…7 However, one of the most promising applications lies in a potential solution to the limitation of Moore's Law. 8 Matrix multiplications form the basis of many mainstream computing systems, such as numerical simulations, artificial intelligence (AI) or neural networks, and can be classified as a parallel multiply and accumulate function (MAC). 9 Moreover, as MACs are typically a linear operation with minimal requirement on precision, optical processors excel in computing such calculations.…”

Grand challenges in neuromorphic photonics and photonic computing

“…Supercomputers are rapidly approaching the exascale era, in which requirements regarding latency, bandwidth, and energy consumption are challenging for digital electronics. In addition to this, the explosive growth of artificial intelligence (AI) and machine learning, and their penetration into everyday life, is forcing us to reconsider the traditional way computers work Xu and Jin (2023), and post-Moore paradigms and computer architectures need to be seriously considered Shalf (2020). Centralized architectures are indeed inefficient in implementing models used for artificial neural networks (ANN), which are inherently distributed and require massive parallel interconnections between a multitude of elementary computing units Jain et al (1996).…”