D-Wave CEO Responds to Criticisms About Quantum Supremacy Claim

Insider Brief

• D-Wave CEO Alan Baratz defends the company’s claim of demonstrating quantum computational advantage in complex materials simulations, responding to critiques from classical computing researchers.
• Studies from the Flatiron Institute and EPFL suggest that classical algorithms, including belief propagation and time-dependent variational Monte Carlo methods, can match or exceed D-Wave’s results in certain cases.
• Baratz argues that competing studies tested only a subset of problems addressed in D-Wave’s work and asserts that the company’s quantum simulations covered a broader range of lattice geometries and conditions.

D-Wave CEO Alan Baratz is pushing back against recent critiques of the company’s latest study, which asserts that its Advantage2 quantum annealer has achieved quantum supremacy in complex materials simulations. The study, published in Science, has drawn scrutiny from researchers who argue that classical algorithms can still rival or exceed quantum methods in some cases.

A key challenge to D-Wave’s claim comes from a team led by Joseph Tindall of the Flatiron Institute, who developed a competing classical approach using belief propagation, an artificial intelligence-based algorithm. Their method, based on a 40-year-old technique, reportedly produced more accurate results than the quantum annealer for select two- and three-dimensional systems, according to Science News.

The team’s findings, posted on arXiv but not yet peer-reviewed, question whether quantum annealing holds a definitive advantage in all scenarios.

Tindall and his colleagues assert in their paper that their classical approach “demonstrably outperforms other reported methods” for the specific spin glass problem tested. In two cases, they claim to have achieved lower errors than the D-Wave Advantage2 system.

However, their study examined only a subset of problems addressed in D-Wave’s work, according to Baratz, leaving open the question of whether classical techniques can consistently match or surpass quantum simulations in broader applications.

Another test of D-Wave’s claims emerged from a separate arXiv study by researchers at EPFL’s Institute of Physics and Center for Quantum Science and Engineering. Led by Linda Mauron and Giuseppe Carleo, the team employed time-dependent variational Monte Carlo (t-VMC) techniques to simulate quantum annealing in spin glass models. Their findings suggest that classical simulations can rival quantum annealers on a scale once considered infeasible, with correlation errors below 7% for systems up to 128 spins in a three-dimensional diamond lattice.

The study raises questions about the idea that infinite-dimensional systems inherently favor quantum annealers.

Baratz, however, rejects these critiques, stating that comparisons to classical approaches misrepresent the scope and depth of D-Wave’s results.

“Let me be clear, as there seems to be confusion and misleading claims around our computational supremacy work that Science published in a peer-reviewed paper today,” Baratz said in a statement. “We have shown quantum supremacy on complex materials simulation problems. This is a first for the industry.”

He further argued that competing studies did not attempt the most complex lattice geometries, nor did they reproduce the largest simulations or simulate the full range of conditions tested in D-Wave’s study.

“Other physicists’ papers published this week do not come close to achieving what we accomplished on the D-Wave Advantage2 quantum computer, and their claims are confusing the public,” he said in the statement.

Baratz acknowledged that classical researchers have made progress in pushing the limits of existing techniques but maintained that their work remains a “small subset” of what D-Wave achieved.

“We demonstrated deeper and larger simulations on a greater variety of lattices than what anyone else has shown,” he said, emphasizing that D-Wave’s results should be recognized as a significant milestone in quantum computing.