UK Funds 12 AI Scientists That Run Their Own Experiments

Imagine a lab that never sleeps, where AI designs experiments, runs them, analyzes results, and starts again without human hands touching a beaker. That future just got a major funding boost in the UK.

Britain's Advanced Research and Invention Agency received 245 proposals from teams already building AI scientists, automated systems that handle entire research workflows from start to finish. The agency planned to fund a few projects but was so impressed they doubled their budget, awarding roughly $675,000 each to 12 teams across the UK, US, and Europe.

These aren't just fancy calculators. An AI scientist can form hypotheses, design experiments to test them, run those experiments using robotic equipment, and analyze what happened. Then it loops back and does it all over again, learning as it goes.

One winning team at Lila Sciences is building an AI NanoScientist to discover better ways to make quantum dots, the tiny particles that power medical imaging and solar panels. Another team at the University of Liverpool is creating a robot chemist that runs multiple experiments simultaneously and even troubleshoots its own errors using vision technology.

London startup Humanis AI is developing ThetaWorld, an AI scientist using language models to design battery experiments that will run in automated labs at Sandia National Laboratories in the US. The goal is understanding the physical interactions that make batteries perform better.

Human scientists become supervisors rather than lab technicians. "There are better uses for a PhD student than waiting around in a lab until 3am to make sure an experiment is run to the end," says Ant Rowstron, ARIA's chief technology officer.

The technology still has growing pains. A recent study found that AI agents fail to complete scientific workflows 3 out of 4 times, sometimes from what researchers called "overexcitement that declares success despite obvious failures."

But Rowstron sees rapid progress ahead. Right now, these AI scientists call up existing tools like AlphaFold when needed. Within a decade, he believes they'll create entirely new scientific tools on the fly when they hit a problem that existing solutions can't solve.

The Ripple Effect

The teams have nine months to prove their AI scientists can generate novel findings. If they succeed, ARIA will use what it learns to fund larger projects that could transform how science gets done.

Faster discovery means faster solutions to problems affecting millions of people, from better batteries for clean energy to improved medical treatments. And freeing scientists from repetitive tasks lets them focus on the creative, big-picture thinking humans do best.

The race to automate discovery is heating up, with most major AI companies now fielding science teams. This funding wave shows the technology has moved from concept to reality, ready to prove what it can do in real labs solving real problems.