The UK government is backing AI scientists that can run their own lab experiments

The UK government is backing AI that can run its own lab experiments

UK agency ARIA selected 12 AI scientist projects from 245 proposals, doubled its planned funding, and will give each team about £500,000 for nine months. ARIA defines an AI scientist as a system that hypothesizes, runs experiments, analyzes results, and iterates; the funded projects still rely on existing tools. The key signal is reproducible lab-loop execution, not press-release heat: one cited external study reports LLM agents failed to complete a scientific workflow 3 out of 4 times.

Why it matters: HKR-H/K/R all pass: 'AI runs its own lab experiments' is a strong hook, and the piece includes 12 teams, 245 proposals, ~£500k each, a 9-month term, and a cited 75% failure rate. Important for agentic science, but this is funding for early systems, not a proven breakthrough.