Universal photonic artificial intelligence acceleration

• 成果类型: Article
• 署名作者: Ahmed, Sufi R.; Baghdadi, Reza; Bernadskiy, Mikhail; Bowman, Nate; Braid, Ryan; Carr, Jim; Chen, Chen; Ciccarella, Pietro; Cole, Matthew; Cooke, John; Desai, Kishor; Dorta, Carlos; Elmhurst, Jonathan; Gardiner, Bryce; Greenwald, Elliot; Gupta, Shashank; Husbands, Parry; Jones, Brian; Kopa, Anthony; Lee, Ho John; Madhavan, Arulselvan; Mendrela, Adam; Moore, Nicholas; Nair, Lakshmi; Om, Aditya; Patel, Subie; Patro, Rutayan; Pellowski, Rob; Radhakrishnani, Esha; Sane, Sandeep; Sarkis, Nicholas; Stadolnik, Joe; Tymchenko, Mykhailo; Wang, Gongyu; Winikka, Kurt; Wleklinski, Alexandra; Zelman, Josh; Ho, Richard; Jain, Ritesh; Basumallik, Ayon; Bunandar, Darius; Harris, Nicholas C.
• 署名单位: OpenAI
• 刊物名称: Nature
• ISSN/ISSBN: 0028-1694
• DOI: 10.1038/s41586-025-08854-x
• 发表日期: 2025-04-10
• 关键词: deep; LEVEL

摘要

Over the past decade, photonics research has explored accelerated tensor operations, foundational to artificial intelligence (AI) and deep learning1, 2, 3-4, as a path towards enhanced energy efficiency and performance5, 6, 7, 8, 9, 10, 11, 12, 13-14. The field is centrally motivated by finding alternative technologies to extend computational progress in a post-Moore's law and Dennard scaling era15, 16, 17, 18-19. Despite these advances, no photonic chip has achieved the precision necessary for practical AI applications, and demonstrations have been limited to simplified benchmark tasks. Here we introduce a photonic AI processor that executes advanced AI models, including ResNet3 and BERT20,21, along with the Atari deep reinforcement learning algorithm originally demonstrated by DeepMind22. This processor achieves near-electronic precision for many workloads, marking a notable entry for photonic computing into competition with established electronic AI accelerators23 and an essential step towards developing post-transistor computing technologies.