IBM and University Researchers Make Noisy Quantum Computing Useful
By Mark Smotherman, Deputy Editor
The eagerly anticipated NISQ era of quantum computing is drawing closer. With the advent of useful quantum computers, researchers will be able to tackle complex problems that classical computers simply cannot handle. These problems include simulating molecular interactions and developing new materials. Despite the challenges posed by noisy quantum computing, there are still quantum algorithms that can achieve results beyond the capabilities of classical computers. Now, researchers at IBM, in collaboration with the University of California at Berkeley and Purdue University, have found a way to use a noisy quantum processing unit (QPU) to obtain these results.
In their groundbreaking paper titled “Using Noise-Extrapolation to Achieve Correct Results from Noisy Quantum Computations,” the researchers describe how they utilized a 127-qubit “Eagle” QPU to run quantum algorithms that produce results unattainable by classical computers. They achieved this by employing a technique called zero noise extrapolation (ZNE) and establishing a feedback mechanism between the Eagle QPU and supercomputers at UC Berkeley and Purdue University.
Anthony Laing, manager of media relations at IBM Research, explains the significance of this work: “This work represents an important advance in quantum computing. It shows that the power of quantum effects can be harnessed for real-world problems in a noisy quantum environment, which is essential for quantum computing to achieve its potential.”
This breakthrough paves the way for quantum utility, where quantum computing becomes useful for real-world applications, to arrive sooner than anticipated in the NISQ era. Laing adds, “It was a great collaborative effort between the groups at UC Berkeley and Purdue University and the IBM QX team. It required the integration of the NISQ quantum computers with powerful supercomputers, which can perform trillions of operations per second.”
To handle the complexity of qubit simulation, the researchers at UC Berkeley employed compression algorithms known as tensor network states, which represent calculations in a three-dimensional space. This approach surpasses the limited capabilities of standard 2D solutions in representing complex information relationships and volumes.
The researchers successfully ran substantial and deep circuits on the Eagle QPU, obtaining accurate results. This demonstrates that NISQ quantum computers can produce results that surpass the capabilities of standard supercomputers or would require significant time and resources to achieve.
Laing explains that the collaboration will continue as the number of useful qubits increases. The goal is to explore circuits with depths greater than 60, while ensuring that standard supercomputers can keep up with the growing number and quality of qubits.
Furthermore, the researchers are exploring new classical methods to improve the accuracy of results obtained from tensor network simulations. Accurate results depend on the ability to predict and control noise in a quantum system.
Laing further clarifies, “Researchers had to understand how different factors, such as qubit activation and arrangement, affected the noise in the Eagle QPU. To compare quantum simulations run on the Eagle QPU with the results from supercomputers, a ground truth can be established.”
IBM collaborated with the National Energy Research Scientific Computing Center and the Anvil supercomputer at Purdue University to perform these comparisons.
By employing ZNE, the capabilities of NISQ quantum computers are already enhanced, allowing them to perform calculations that surpass current classical Turing machines. As hardware and algorithm optimizations continue to improve, the power of supercomputers will increase, enabling more accurate verification of quantum computing results.
Ultimately, as quantum computing progresses, it will no longer need to rely on classical techniques for verification. This will lead to further advancements in quantum error correction.
For more details, refer to the full paper on arXiv.
