When Google Quantum AI reported in July that it had taught a quantum computer to “learn from its errors,” the obvious story was better error correction. Quantum computers are extremely sensitive machines. Small changes in their environment can push them out of tune and introduce errors, which can force researchers to stop a calculation and recalibrate the hardware. Google’s AI system learned to make some of those adjustments while the machine was still running. In tests, it made the system 3.5…