量子计算机的稳定运行对激光频率精度有极高要求。1QuEra近日部署了一套由AI代理Claude驱动的自动激光锁定恢复与调优系统，通过模型硬件标准（MHS）实现了这一关键过程的自动化控制。1该标准由Anthropic与HHMI Janelia研究园区合作制定。1
这套AI系统在实际测试中表现出色。在695次恢复尝试中，成功率达到99.3%，其中波长保持故障可在0.9至5.4秒内恢复，硬故障恢复时间为10至14秒，相比传统人工专家操作的5至10分钟效率提升显著。1在验证期间，激光发生了43次自发模态跳跃，均被该系统自动捕获并恢复。1
通过优化反馈参数，该AI系统还将残余频率噪声降低了5倍。1在一次为期19小时的无人值守测试中实现零次锁定丢失，相比之前每小时平均丢失1.6次的情况大幅改善。1此外，第二波长的调试时间也从首次试点的数周缩短至一个无人值守夜间运行即可完成。1这使得量子计算机能够在无需人工干预的情况下稳定运行。1
QuEra has automated the laser frequency stabilization process in quantum computing systems by deploying an AI agent based on Claude, leveraging the Model Hardware Standard (MHS) framework developed through collaboration between Anthropic and HHMI Janelia Research Campus 1. The system automatically restores laser lock when disruptions occur, achieving a 99.3% success rate across 695 out of 700 recovery attempts 1. Recovery times range from 0.9 to 5.4 seconds for wavelength-holding faults and 10 to 14 seconds for hard faults, dramatically faster than the 5 to 10 minutes required by human experts 1. During validation, the system captured and recovered from 43 instances of spontaneous laser mode jumping without manual intervention 1.
The AI-driven approach has significantly improved quantum computer reliability and operational efficiency. Through feedback parameter optimization, the system reduced residual frequency noise by a factor of five 1. In a 19-hour unattended testing period, the system maintained laser lock with zero losses, compared to 1.6 losses per hour under previous conditions 1. Debugging time for a second laser wavelength was reduced from several weeks during initial trials to a single unattended overnight run 1. The laser frequency precision requirement of ±1 kHz or better—approximately ±0.000000001 nanometers—is maintained consistently by the automated system 1, enabling quantum computers to operate reliably without continuous human oversight 1.
评论
还没有评论，欢迎留下第一条。