TL;DR

ARIA, the UK’s moonshot R&D agency, has funded 12 projects building “AI scientists” that can design and run lab experiments autonomously. The agency doubled its planned funding after receiving 245 proposals, demonstrating rapid advancement in this emerging field.

Automating the Scientific Workflow

ARIA defines an AI scientist as a system that can run an entire scientific workflow: generating hypotheses, designing experiments, executing tests, and analysing results. These systems can then feed results back into themselves and iterate continuously.

“There are better uses for a PhD student than waiting around in a lab until 3 a.m. to make sure an experiment is run to the end,” says Ant Rowstron, ARIA’s Chief Technology Officer.

The agency picked 12 projects from 245 proposals, doubling its intended funding due to the volume and quality of submissions. Half the winning teams are from the UK, with the remainder from the US and Europe. Each receives approximately £500,000 for nine months’ work.

Diverse Applications

Winning projects include Lila Sciences, a US company building an AI nano-scientist to discover optimal methods for composing quantum dots used in medical imaging and solar panels. A University of Liverpool team is developing a robot chemist that runs multiple experiments simultaneously and uses vision language models to troubleshoot errors.

A London startup in stealth mode is developing ThetaWorld, an AI scientist using LLMs to design battery experiments that will run in automated labs.

Testing the Frontier

The relatively modest funding represents an experiment for ARIA itself. “By funding a range of projects for a short amount of time, the agency is taking the temperature at the cutting edge,” explains Rowstron.

Current systems work by using large language models for ideation while calling on other models for optimisation and experiment execution. Rowstron envisions a future where AI scientists create entirely new tools on demand.

Looking Forward

For UK research organisations and science-intensive businesses, ARIA’s investment signals government commitment to AI-driven scientific discovery. While current systems still face reliability challenges, the rapid advancement suggests autonomous research capabilities will become increasingly important for competitive R&D operations.