Quantum computers tested on real crime data — and they're no better than laptops

Researchers compared quantum computers, classical machine learning, and hybrid systems on 16 years of Bangladesh crime statistics. The quantum approach matched classical accuracy (84.6%) while using fewer parameters, but delivered no practical advantage for the actual job of predicting crime patterns.

This is a straightforward empirical test of a widespread claim: that quantum computing will solve hard real-world problems faster than classical approaches. The test finds that on an actual high-dimensional, imbalanced dataset from law enforcement, quantum-inspired methods work fine but offer no measurable edge. The authors propose the real selling point is memory efficiency on edge devices — a much narrower claim than the hype around quantum machine learning. This matters because it's the kind of honest result that rarely gets published in the quantum computing literature, which tends toward capability demonstrations rather than head-to-head failure documentation.