Researchers at Google Quantum AI have applied a machine-learning technique called reinforcement learning to automatically correct errors in their Willow quantum processor (a 105-qubit chip unveiled in December 2024) while calculations continue uninterrupted, according to a study published this week in the journal Nature. The system works by feeding measurements from auxiliary qubits into an AI agent that continuously adjusts microwave pulses sent to the processor's data qubits, learning to compensate for environmental disturbances more quickly and precisely than human experts can. The advance is considered a meaningful step toward fault-tolerant quantum computing, which both Google and IBM are racing to achieve by the end of the decade — a timeline the US government has also endorsed in a recent executive order signed by President Trump.