One of quantum computing's most significant milestones has been achieved, with a processor that can be programmed across all of its quantum gates beating the most powerful classical computers at a task. “Beating”, however rather undersells the achievement, as the quantum processor took 36 microseconds to perform a task that would take existing supercomputers 9,000 years.
The phenomenon of quantum superposition allows the ones and zeros of classical computers to be simultaneously both. In theory, taking advantage of this should enable quantum computers to perform simultaneous calculations that existing machines must do in sequence, often over impossibly long periods of time. Quantum computing has proven harder to implement in practice than in theory, but steady progress has been made.
We're still a long way from fully operable quantum computers, however, so an interim goal was created, known as quantum advantage. Processors will display quantum advantage when they can beat the best classical equivalent at a well-defined task. A paper in Nature claims to have done just that on a task known as Gaussian boson sampling.
To test Gaussian boson sampling, photons of light are sent through a network of beam splitters to be counted at a detector. The computer seeks to reconstruct the photons' probability distribution based on the number counted and certain attributes. The more photons there are, the longer it would take classical computers to perform the calculations required to establish the distribution.
The photonic processor described in the new paper, known as Borealis, has been used to detect an average of 125 photons in repeated trials, peaking at 219 and easily shattering the previous record of 113.
Its makers, Canadian company Xanadu, describe Borealis as “the world's first photonic quantum computer offering full programmability over all of its gates and capable of quantum advantage.” Borealis can also be accessed over the cloud.
Although quantum processors have been achieved using many different designs, the calculations they have performed have mostly been more like party tricks than serious problem-solving. Most have also been riddled with errors, a problem that clearly needs to be addressed.
Gaussian boson sampling is not terribly important in itself but it represents a way to measure quantum computers' progress. It is one of the first areas in which quantum computers should be able to beat classical computers, analogous to chess as a way of testing machine intelligence against humans. Computers beating chess grandmasters didn't mean machines had achieved complete intellectual superiority over humans, but it marked a milestone on that path. Similarly, it has been considered an important marker for quantum computers to outshine classical equivalents at something, and Gaussian boson sampling has been considered a prime possibility, now accomplished.
Borealis achieved quantum advantage by passing squeezed light through loops of optical fiber that act as delay lines, and by classifying photons by time of arrival rather than direction.
Previous, more limited, examples of quantum advantage have been demonstrated, but only with static gate sequences and only small advantages relative to classical computers. Borealis beats these by a factor of 50 million. Its makers also claim it is far more resistant to classical spoofing attacks than previous quantum computers.
As Dr Daniel Brod of the Federal Fluminense University, Niterói, Rio de Janeiro explains in an accompanying editorial, this does not mean practical quantum computing is just around the corner. To achieve the touted goals in cryptography cracking or pharmaceutical research “a quantum computer capable of such tasks would require millions of controllable, robust quantum bits (qubits), whereas current quantum processors have fewer than 100 qubits,” Brod notes.
1