Quantum Computing and Artificial Intelligence in Learning and Device Engineering

Quantum computing and artificial intelligence interact through two different engineering problems. Quantum processors can implement subroutines for learning and optimisation, while classical learning can improve the operation of quantum devices. This critical narrative review assesses both directions using a common set of comparison fields covering task, input access, execution setting, comparator, resource boundary and uncertainty. Quantum kernel and variational models show task-specific results, but performance on conventional data depends strongly on preprocessing, tuned classical baselines, finite-shot estimation and circuit cost. Learning from quantum experiments provides a distinct route to advantage when coherent access changes the number of physical experiments needed. In device engineering, learned surface-code decoders can reduce logical error on matched processor data, while sustained throughput, feedback latency and robustness to drift determine whether an accuracy gain can be used online. Engineering diagrams identify the relevant system boundaries. Reproducible analytical plots quantify direct kernel acquisition, sampling precision, dense simulation storage and parameter-shift workloads. A separate plot reproduces reported decoder estimates from a single experiment. Across both directions, evidence for utility is strongest when the improvement survives matched inputs, a competitive baseline and end-to-end cost accounting at the scale of the intended application.