2 days ago
Quantum system beats classical AI in real test, powers greener supercomputing future
In a major leap toward the future of computing, researchers have shown that even small-scale quantum processors can outperform classical algorithms in machine learning tasks.
The finding offers a glimpse into a faster, greener era within the relatively new research field of Quantum Machine Learning, a space gaining momentum across both academia and new study combines quantum computing and machine learning, two of the most disruptive technologies of our time.
Recent advances in both fields are reshaping the technological frontier. While AI is already embedded in everything from personal assistants to scientific research, quantum computing promises a fundamentally new way of processing information. Their intersection has given rise to a rapidly growing field: quantum machine learning.
This emerging discipline explores whether quantum systems can improve the speed, accuracy, or efficiency of machine learning algorithms. However, proving such an advantage on today's limited quantum hardware remains a major challenge—one that researchers are just beginning to by an international team led by the University of Vienna, the experiment used a photonic quantum processor to classify data points, an essential task in modern AI systems.
The researchers found that the quantum system outperformed its classical counterpart, making fewer errors—a rare, real-world glimpse of quantum advantage with current hardware.
This breakthrough was achieved using a quantum photonic circuit developed at Italy's Politecnico di Milano and a machine learning algorithm proposed by UK-based Quantinuum. The experiment marks one of the first demonstrations of quantum enhancement in practical AI tasks, rather than simulations.
By isolating the quantum contribution in the classification process, the team was able to pinpoint specific scenarios where quantum systems results not only validate the potential of photonic quantum processors but also lay the groundwork for identifying machine learning tasks where quantum computing can make a real-world impact, even with today's limited-scale hardware.'We found that for specific tasks, our algorithm commits fewer errors than its classical counterpart,' said Philip Walther, project lead from the University of Vienna.
"This implies that existing quantum computers can show good performances without necessarily going beyond the state-of-the-art technology," adds Zhenghao Yin, first author of the accuracy, the experiment also reveals another important advantage in energy quantum systems process information using light and therefore consume significantly less power than traditional hardware, which is becoming increasingly important as AI's energy demands continue to rise.'This could prove crucial in the future, given that machine learning algorithms are becoming infeasible due to high energy demands,' said co-author Iris showing that today's quantum devices can already offer tangible improvements, the findings could steer both quantum computing and classical machine learning into a more symbiotic future, where quantum-inspired algorithms push conventional boundaries and photonic platforms help make AI more sustainable.
The study has been published in the journal Nature Photonics.
