Radical AI's Self-Driving Lab Uses Robots to Run Experiments

Automation Comes to the Research Bench

A New York City startup is testing whether artificial intelligence can fundamentally reshape how scientific research gets done. Radical AI operates what it calls a self-driving laboratory, where robotic systems physically conduct experiments while AI algorithms determine which tests to run next.

The company has secured more than $65 million in funding and counts the U.S. Air Force among its customers. Its focus areas include batteries, semiconductors, and heat-resistant alloys—materials critical to defense and energy applications where incremental improvements can have outsized impact.

How the System Works

In Radical AI's facility, the traditional division of labor in a research lab has been inverted. Robots handle the physical work of synthesizing compounds and running tests. The AI system analyzes results in real time and decides what experiment should come next, creating a feedback loop that operates without human intervention between cycles.

This approach represents a departure from conventional materials science, where researchers design experiments based on hypotheses, manually prepare samples, collect data, and then interpret results before planning the next round of tests. The automated system can potentially run far more experimental iterations in a given timeframe.

Why It Matters

The rise of autonomous laboratories forces a fundamental question about the nature of scientific discovery: Can machines generate genuine insights, or do they simply accelerate trial-and-error testing? For industries where materials development is both critical and time-intensive, the answer has significant competitive implications. If AI-driven labs can compress years of research into months, companies and governments that adopt the technology early gain substantial advantages in fields from energy storage to aerospace engineering.

Questions About Scientific Understanding

The scientific method has evolved over centuries, but human scientists have remained central to the process—forming hypotheses, designing experiments, and interpreting results within broader theoretical frameworks. Radical AI's approach challenges this assumption by removing humans from the experimental loop.

Whether AI systems can truly "do science" or merely optimize experimental parameters remains an open question. Critics might argue that pattern recognition and optimization, however sophisticated, differ from the creative leaps and contextual understanding that characterize breakthrough scientific work.

The Bulletin of the Atomic Scientists visited Radical AI's facility and produced a video report examining how the laboratory operates and exploring the implications of AI-directed research. The report was originally published by the Bulletin.