Anthropic Previews Standard for AI Control of Physical Devices

Anthropic has unveiled a research preview of a shared specification, known as the Model Hardware Standard (MHS), which aims to guide artificial intelligence agents in managing physical devices like lab equipment and manufacturing tools. This standard allows agents to perform tasks ranging from routine drug discovery to advanced operations such as quantum computer laser calibration. They can work continuously and recover from hardware errors without human assistance.

Anthropic is initially sharing MHS with partners in sectors like science, robotics, electronics, and manufacturing. After working with these partners to create safety evaluations and best practices for AI systems that control physical equipment, the company plans to make the standard open source. According to Anthropic, MHS has helped reduce the time needed to integrate devices, enable faster iteration in various experimental settings, and assist in real-time fault detection for machines.

Physical AI, which involves integrating AI with physical devices, is gaining traction as sensing, perception, and large AI models enhance machines' capabilities beyond traditional automation. Companies specializing in this field are attracting investor attention, with venture funding pouring into startups focused on physical AI and vertical AI applications.

Anthropic is also supporting scientists by opening 10,000 free and discounted Claude subscriptions for scientists worldwide and expanding its AI for Science program to cover more scientific fields. Anthropic aims to accelerate scientific discovery by making AI more accessible to researchers.