D-Wave proclaims quantum superiority – wrongly

Contents

The Canadian manufacturer D-Wave Quantum Inc. announced on Wednesday that it had solved a useful problem for the first time with a quantum computer that surpasses the capabilities of conventional supercomputers.

The D-Wave machine is said to have run a simulation in just a few minutes that would take a supercomputer almost a million years and more energy than the entire world population consumes in a year. The results were presented by the international research team led by D-Wave in a paper published am Wednesday in the journal Science. The research builds on prework from 2023.

Quantum computer vs. supercomputer

In of its press release, the company calls its experiment a demonstration of quantum superiority. This term usually refers to the point at which quantum computers solve problems that no classical supercomputer could solve in a realistic amount of time. Other researchers prefer the less politically charged term quantum advantage.

Manufacturers of quantum computers repeatedly claim to have beaten classic supercomputers. Just last December, Google claimed to have provided just such proof with its new quantum chip “Willow”. Google's first Behauptung of quantum superiority from 2019 was disproved after a short time. The core problem with these experiments so far has been that the problems solved were highly academic and had no practical use whatsoever.

Simulation of magnetic materials

The current case should be different. “Our demonstration of the superiority of quantum computing on a useful problem is an industry first. All other claims that quantum systems outperform classical computers have been disputed or involved the generation of random numbers with no practical value,” says D-Wave CEO Alan Baratz.

“This is highly exciting work that looks for quantum advantages in the right place,” explains Jens Eisert, quantum researcher at Freie Universität Berlin. In their experiments, the D-Wave researchers simulated the behavior of magnetic materials.

They investigated the behaviour of so-called spin glasses, materials with a complex magnetic order that are relevant for research, but also for solving optimization tasks. For example, scientists hope to be able to research new materials using such simulations. “These are not overly important model systems for solid-state issues, but they are plausible candidates for exploring a paradigmatic test case,” says Eisert.

Quantum computers and quantum annealers

Traditionally, such optimization issues are simulated using supercomputers. However, the qubits of a quantum annealer already have quantum mechanical properties that a classical computer must first simulate. This makes quantum computers particularly suitable for simulating quantum systems.

However, Eisert explains that the D-Wave machine is “not yet a full quantum computer”, but a “quantum simulator that can very accurately simulate model systems from solid-state physics.” A quantum annealer is a specialized machine designed to solve optimization problems, such as in materials research, the financial market and machine learning.

While other manufacturers are still creating prototypes of quantum chips, D-Wave is already selling its quantum annealers commercially. For example, a D-Wave device has been in Forschungszentrum Jülich since 2022. “In the past, D-Wave has often attracted attention due to exaggerated claims and otherwise there is a certain basic scepticism towards this company,” notes Markus Heinrich from the University of Cologne. Nevertheless, the latest publication has been well received by the specialist community.

Quantum annealer

In the case of only two freely selectable parameters, an optimization problem can be imagined as a hilly landscape. The task is to find the lowest point in the landscape. However, real problems usually depend on far more parameters: The fastest route from A to B depends not only on the road layout, but also on traffic lights, speed limits, traffic conditions and so on. While this is a very difficult task for classical computers, quantum annealers are particularly good at finding the optimum for such multidimensional problems.

A quantum annealer consists of quantum objects, for example atoms or tiny circuits. First, the optimization problem is translated into the structure of the quantum objects. Then energy is supplied to the particles. Put simply, they are then in all possible states at the same time. Then the quantum system is slowly cooled down and it slips into the most energetically favorable state on its own. This corresponds to the sought-after, optimal solution to the optimization problem.

The D-Wave researchers simulated their chosen problem both on their D-Wave Advantage1 and Advantage2 machines and on the Frontier supercomputer at the United States Department of Energy's Oak Ridge National Laboratory. On the supercomputer, the team used various algorithmic methods to solve the problem. It started with an arrangement that is known to be easy to simulate classically, gradually increased the complexity, and finally reached the point where the supercomputer could no longer solve the problem in a realistic amount of time.

But no quantum advantage

However, Eisert warns that the claim of quantum superiority is dangerous. “Because it's not just quantum simulators that are evolving, but also classical simulation methods.” A few days ago, contrary to D-Wave's expectations, a research team from New York and Trieste simulated the issue under consideration on a classical computer.

This result came about a year after D-Wave had published his results on the pre-print server arXiv in advance. A second team from Lausanne demonstrated a similar result only a little later. Both manuscripts follow different paragraphs and appeared on arXiv, so they have not yet been reviewed by independent experts.

Nevertheless, Heinrich considers the original comparisons between D-Wave and classical algorithms to be fair. “I believe that the estimates were made very conscientiously. However, such figures should always be considered a challenge to adapt and improve existing classic algorithms,” he says.

Eisert also says: “The race between quantum and classical computers is healthy, good, and important. It develops both fields further.” However, researchers should better refrain from making statements about a quantum advantage. “The field of quantum computing, however, can make good use of calm method development – without excitement and hype –.”

(spa)