Proven quantum advantage: cut time for a learning task from 20M years to 15 mins (phys.org)

🤖 AI Summary
Researchers at the Technical University of Denmark and international partners report a provable quantum advantage in a photonic learning task, published in Science. Using entangled squeezed light on a scalable optical platform, the team learned the noise “fingerprint” of a bosonic random-displacement channel in about 15 minutes—what a comparable classical strategy would need roughly 20 million years to achieve. This is the first demonstrated quantum advantage for a photonic system and validates that entanglement can fundamentally reduce the number of measurements required for characterizing noisy quantum systems, with clear potential for sensing and machine‑learning applications. Technically, the experiment probes an optical channel where multiple pulses share the same noise pattern: two squeezed beams are entangled, one acts as the probe and the other as a reference, and a joint measurement compares them in a single shot. That comparison cancels much of the measurement uncertainty and extracts far more information per trial than measuring the probe alone. The setup runs at telecom wavelengths with off‑the‑shelf components and tolerates ordinary losses, showing the advantage comes from the measurement strategy rather than idealized hardware. Backed by prior theory (“Entanglement‑Enabled Advantage for Learning a Bosonic Random Displacement Channel”), the work proves a task no classical method can match and points to practical domains where entanglement-enabled learning will pay off.
Loading comments...
loading comments...