This Is How Anthropic Thinks AI Agents Should Navigate the Physical World
The potential for AI to automate scientific research and manufacturing must be balanced with new risks, Anthropic says.
Anthropic, an artificial intelligence company, has unveiled a new framework called Model Hardware Standard to guide AI agents in interacting with physical systems such as microscopes, quantum computing hardware, and robot arms. This development comes as a growing belief that AI could revolutionize scientific research and industries, provided it can do so safely. The company plans to collaborate with trusted partners to ensure maximum safety before making the framework generally available.
Despite potential misuse concerns, such as the development of biological weapons, Anthropic believes that built-in guardrails in AI models will prevent malicious actors from exploiting the new standard for harmful purposes. Alek Kemeny, a quantum physicist who co-led the development, emphasizes that the goal is to accelerate scientific progress by bridging the gap between literature review and data analysis, bringing this power to the experimental world.
AI agents, which are designed to take actions on computers and perform tasks like answering emails, are considered the next step beyond chatbots. They have the potential to use various hardware, and Anthropic aims to establish rules for their safe use. Several startups, including Periodic Labs, LILA Sciences, Edison Scientific, and Discovery Loop, are exploring AI-driven scientific discovery, envisioning a recursive loop where AI develops and tests hypotheses.
Configuring scientific equipment to interact with other hardware typically demands significant expertise, but AI could automate much of the complex engineering by configuring machines and enabling them to communicate. This framework is being developed in collaboration with manufacturers. Kemeny notes that the new standard allows Claude to view robots on a factory line and optimize their behavior.
Concerns about the potential risks of AI agents using physical systems, such as damaging equipment or endangering people, are valid. Experiments have demonstrated how AI models can manipulate robots to misbehave. To address these risks, Anthropic's new standard enables scientists and engineers to specify how AI models should avoid using different hardware, thereby preventing mishaps.
Previously, Anthropic introduced the Model Context Protocol, which outlines rules for AI models interacting with different software programs.
Written by urgent.news from Wired Business's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
