SBIR Phase I: Superconducting Neural Units with Adjustable Weights

SBIR 第一阶段：权重可调的超导神经单元

• 批准号: 2213032
• 负责人: Yuchao Zhou
• 金额: $ 25.56万
• 依托单位: WOLFZYC LLC
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-09-15 至 2023-11-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2213032&HistoricalAwards=false
• 关键词: SBIR; Phase; Superconducting; Neural; Units

The broader impact of this Small Business Innovation Research (SBIR) Phase I project aims to develop a new neuromorphic computing platform, the effects of which may be seen in many different fields, ranging from the automobile industry to robotics and biomedical applications. The reduction in energy use afforded by the technology could save up to 2% of the total electricity consumption nationwide, making it one of the most environmentally friendly computing platforms. Furthermore, the superconductor-based hardware has the potential to provide one of the fastest commercialized high-performance computing platforms for both industry and research, offering the potential to solve current computational problems in many fields. The development of high-frequency neuromorphic computing, with specific application to artificial intelligence applications, promises to strengthen the US economy. This Small Business Innovation Research (SBIR) Phase I project aims to bring to the technological forefront a high-speed superconductor electronics-based neuromorphic computer centered around neural units with adjustable weights serving as a disordered array. Currently, high-speed computing necessary for large artificial intelligence models is bottlenecked at semiconductor device architecture as Moore’s Law reaches its limits. Much of the necessary computing power is only accessible via paid run time on vast server banks owned by a select few companies that can afford to build them. By unlocking computing power based on SQUID (superconducting quantum interference device) arrays and rapid single flux quantum (RSFQ), the project aims to develop an affordable, reliable, and carbon friendly high-speed computing platform, thereby helping to solve the power/cost crisis facing currently accessible computing systems.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

这个小企业创新研究（SBIR）第一阶段项目的更广泛影响旨在开发一种新的神经拟态计算平台，其影响可以在许多不同领域看到，从汽车工业到机器人和生物医学应用。该技术提供的能源使用可以节省全国总电力消耗的 2%，使其成为最环保的计算平台之一。此外，基于超导体的硬件有潜力提供最快的商业化高性能之一。两个行业的计算平台高频神经拟态计算的发展以及人工智能应用的具体应用有望增强美国经济。旨在将基于高速超导电子学的神经形态计算机推向技术前沿，该计算机以具有可调节权重的神经单元为中心，充当无序阵列。目前，大型人工智能模型所需的高速计算在半导体器件架构中遇到了瓶颈。法律达到其大部分必要的计算能力只能通过在少数有能力建造它们的公司拥有的大型服务器银行上付费运行来获得，通过释放基于 SQUID（超导量子干涉装置）阵列和快速单通量的计算能力。量子（RSFQ），该项目旨在开发一种经济实惠、可靠且碳友好的高速计算平台，从而帮助解决当前可用计算系统面临的功耗/成本危机。该奖项是 NSF 的法定使命，并被认为是值得的通过评估提供支持基金会的智力价值和更广泛的影响审查标准。