Anthropic 推出了一项名为模型硬体标准的全新人工智慧工具，使软体代理能够自主控制显微镜和机械手臂等实体的科学与工程设备。这项技术的突破在于让人工智慧从数位世界的资料分析与假设生成，跨足到实验室的真实物理环境中执行实验。

该公司已经与生物技术、机器人以及量子计算领域的实验室和制造商合作测试这项新技术。科学家观察到，人工智慧能够像人类研究员一样追踪、调整并重复实验，例如成功操作复杂的雷射与反射镜，在神经科学实验中准确聚焦于特定的大脑区域。

这项发展符合 Anthropic 在首次公开募股前扩展企业业务的策略，尤其针对制造业与制药业。随著人工智慧在大型制药公司科学研究中的应用日益重要，该公司计划在确保系统安全、避免潜在生物风险的前提下，最终将这项新技术开源。

Anthropic has launched a new artificial intelligence tool called the Model Hardware Standard, enabling software agents to autonomously control physical scientific and engineering devices such as microscopes and robotic arms. This technological breakthrough allows AI to transition from digital data analysis and hypothesis generation to executing experiments in a real-world physical laboratory environment.

The company has already been testing this new technology in collaboration with laboratories and manufacturers in the fields of biotechnology, robotics, and quantum computing. Scientists have observed AI tracking, adjusting, and repeating experiments just like human researchers, such as successfully manipulating complex lasers and mirrors to accurately focus on a specific brain region during a neuroscience experiment.

This development aligns with Anthropic's strategy to expand its enterprise business, particularly in manufacturing and pharmaceuticals, ahead of its planned initial public offering. As artificial intelligence becomes increasingly critical in scientific research at big pharmaceutical companies, the company plans to eventually open-source this new technology, provided that the system is properly safeguarded against potential biological risks.