Neuromorphic photonics: 2D or not 2D?

The computing industry is rapidly moving from a programming to learning area, with the reign of von Neumann architecture starting fade, after many years dominance. new paradigms non-von architectures have started leading development emerging artificial neural network (ANN)-based analog electronic intelligence (AI) chipsets remarkable energy efficiency. However, size and advantages processing elements are naturally counteracted by speed power limits interconnects inside circuits due resistor-capacitor (RC) parasitic effects. Neuromorphic photonics has come forward as research field, which aims transfer well-known high-bandwidth low-energy interconnect credentials photonic circuitry in area neuromorphic platforms. high potential their well-established promise for fJ/Multiply-ACcumulate efficiencies at orders magnitudes higher neuron densities require number breakthroughs along entire technology stack, being confronted major advancement selection best-in-class material platforms weighting activation functions transformation into co-integrated computational engines. With this paper, we analyze current status available integrated technologies propose novel three-dimensional unit which, its compactness, ultrahigh efficiency, lossless interconnectivity, foreseen allow scalable computation AI that outperform electronics efficiency shape future computing.