On Thursday, Anthropic unveiled the Model Hardware Standard (MHS), a new interface aimed at simplifying how AI agents interact with machinery. This standard is compatible with any device featuring a programmable interface, similar to how USB-C standardizes data transmission.
Elizabeth Kelly, Anthropic's head of beneficial deployments, emphasized that while the MHS was developed with scientific applications in mind, it also offers substantial advantages for enterprise and industrial use. This announcement indicates Anthropic's deeper foray into hardware development, an area where competitors such as OpenAI and Amazon have heavily invested.
The company is also expanding its silicon team to create custom chips for its AI models, having recently recruited Caitlin Kalinowski, a hardware expert with experience at OpenAI, Meta, and Apple. The MHS aims to streamline the integration and setup process for companies, and it is designed to be model agnostic, allowing users to utilize it without being tied to Anthropic's Claude models.
Initially, the MHS will be available to a select group of organizations in fields like science, robotics, and manufacturing as part of a research preview, with plans to eventually open-source the standard for broader adoption across various industries.
This follows Anthropic's earlier open-sourcing of the Model Context Protocol in 2024, which facilitated connections between AI agents and data sources