IBM QUANTUM COMPUTER RUNS TASK IN 19 SECONDS, RESEARCHERS SAY

IBM's Nighthawk r2 quantum processor generated a million computational samples in 19 seconds, a task researchers estimate would take a conventional supercomputer around 110 years to complete, according to a preprint study. The research was led by Tigran Sedrakyan, a theoretical condensed matter physicist at BlueQubit in the United States, and has been published on the arXiv repository ahead of formal peer review. The team says the demonstration used a technique called random-circuit sampling, run on hardware that is commercially and broadly accessible rather than confined to a specialist laboratory. The researchers describe it as, to their knowledge, the first demonstration of this kind of quantum advantage on a processor that non-expert users can access and replicate themselves.

Random-circuit sampling has been a central benchmark in efforts to prove quantum computers can outperform classical machines since researchers at the University of California, Berkeley, set out theoretical evidence for the approach in 2018. The method works by applying a sequence of largely randomised operations to qubits, the quantum equivalent of the bits used in ordinary computers, building up an increasingly complex quantum state. That state is then measured repeatedly, producing strings of ones and zeros known as samples, which grow exponentially harder for classical computers to simulate as complexity increases. IBM's dilution refrigerators keep the superconducting processors at extremely low temperatures to maintain the qubits' quantum properties.

The contest to demonstrate a clear, practical quantum advantage over classical supercomputing has continued since 2018, with successive experiments from various research groups adding to the evidence. Because the findings are currently presented in preprint form, they have not yet undergone formal peer review or independent verification by outside scientists. The researchers say the accessibility of the processor used is significant, since it allows other non-expert users to attempt to reproduce the result themselves via commercial cloud access. No further experiments or verification timeline were specified in the material reviewed.