Affiliation:
1. PRIMALIGHT, Faculty of Electrical Engineering; Applied Mathematics and Computational Science, King Abdullah University of Science and Technology , Thuwal 23955-6900, Saudi Arabia
Abstract
Photonic accelerators for Artificial Intelligence (AI) are rapidly advancing, promising to provide revolutionary computational speed for modern AI architectures. By leveraging photons with a bandwidth higher than 100 THz, photonic accelerators tackle the computational demands of AI tasks that GHz electronics alone cannot meet. Photonics accelerators integrate circuitry for matrix–vector operators and ultra-fast feature extractors, enabling energy-efficient and parallel computations that prove crucial for the training and inference of AI models in various applications, including classification, segmentation, and feature extraction. This Perspective discusses modern challenges and opportunities that optical computations open in AI for research and industry.
Subject
Computer Networks and Communications,Atomic and Molecular Physics, and Optics
Reference46 articles.
1. Communication-efficient federated learning via knowledge distillation
2. A federated graph neural network framework for privacy-preserving personalization
3. Federated learning for predicting clinical outcomes in patients with COVID-19
4. J.
Lu
, C.Clark, R.Zellers, R.Mottaghi, and A.Kembhavi, “Unified-IO: A unified model for vision, language, and multi-modal tasks,” arXiv:2206.08916 [cs.CV] (2022).
5. D.
Amodei
and D.Hernandez, https://openai.com/research/ai-and-compute (2018); accessed 27 August 2023.
Cited by
3 articles.
订阅此论文施引文献
订阅此论文施引文献，注册后可以免费订阅5篇论文的施引文献，订阅后可以查看论文全部施引文献