Anthropic's New AI Tool: Conducting Scientific Experiments
Anthropic has unveiled a groundbreaking system that empowers artificial intelligence (AI) assistants, such as Claude, to perform scientific experiments by controlling various equipment. This marks the company's initial foray into creating tools designed for real-world applications.
The Model Hardware Standard (MHS) enables AI agents to operate devices like microscopes, liquid handlers, lasers, and robotic arms, programming them through computer code. Initially, the research preview will be accessible to select scientific research labs and advanced manufacturers.
Key Features:
- Autonomy: AI agents can autonomously conduct experiments, operating equipment around the clock.
- Real-time Adjustments: They can adjust parameters and recover from hardware errors without human intervention.
- Exploratory Abilities: Similar to human scientists, Claude can interact with hardware in an exploratory manner.
Anthropic highlights successful trials where Claude successfully identified specific parts of brain tissue during a neuroscience experiment.
The company plans to open-source MHS but will initially focus on implementing safeguards to minimize errors during AI-hardware interactions. This launch follows the introduction of Claude Science, Anthropic's first product tailored for scientists, which includes applications like rendering protein structures and aiding drug discovery. MHS extends this capability by allowing AI agents to move from analyzing data to interacting with physical equipment.