Swedish AI Scientists Design and Run Their Own Experiments

Scientists just taught artificial intelligence to do science itself, and the results could transform how we discover new medicines and solve biology's biggest puzzles.

Researchers at Chalmers University of Technology and the University of Gothenburg in Sweden created an AI scientist that generates its own hypotheses, designs experiments, runs them using lab robots, and interprets the results. The system completed nearly 2,000 experiments on yeast cells with minimal human guidance.

The AI analyzed a database of 60,000 biological relationships in baker's yeast to form testable predictions about how nutrients affect cell growth and stress resistance. Then it converted those ideas into step-by-step instructions that laboratory equipment could follow automatically.

After each experiment finished, the AI evaluated whether its predictions were correct. When hypotheses were wrong, the system automatically refined them for the next round of testing.

"It is too much information for a human to analyze, but our AI scientist could identify promising biological questions, recommend experiments to test them, evaluate experimental outcomes and iteratively refine its understanding based on new evidence," said researcher Ievgeniia Tiukova.

The system combines multiple large language models, extensive databases, and physical lab robotics into one closed loop. This approach minimizes human bias and reduces errors that can occur when people manually record experiments or work under varying conditions.

The Ripple Effect

This breakthrough arrives as automation transforms scientific research worldwide. The WHO recently published guidelines for responsible AI use in health research, recognizing both the opportunities and ethical considerations involved.

Senior author Ross King emphasized that AI scientists will work with human researchers rather than replace them. "AI Scientists will collaborate with human scientists to accelerate discoveries across biology, medicine and biotechnology," he said.

Human scientists will still set research priorities, interpret broader contexts, develop research programs and ensure ethical compliance. The AI handles the time-consuming work of sifting through massive datasets and running repetitive experiments that would take human teams months or years to complete.

The technology could reduce the time needed to explore complex scientific questions and optimize how laboratories use their resources. By speeding up research cycles, these AI systems might help scientists develop new treatments faster and tackle challenging medical problems that have stumped researchers for decades.

Future generations of autonomous laboratories will become increasingly capable partners in solving biology and medicine's toughest questions, working side by side with the humans who guide them.