An earlier prototype of Jiuzhang 4.0, the quantum computer that achieved quantum advantage
Chao-Yang Lu/University of Science and Technology of China
A quantum computer may have reached “quantum advantage” by carrying out a task that is firmly out of the reach of the world’s best supercomputers. Experts have estimated replicating the calculation on a classical machine would take trillions of trillions of times the age of the universe. But what does this feat mean for the development of truly practical quantum computers?
The new record holder is a quantum computer called Jiuzhang 4.0 that makes calculations using particles of light, or photons. Chao-Yang Lu at the University of Science and Technology of China and his colleagues used it for Gaussian boson sampling (GBS), a task where a sample of photons is measured after the particles have travelled through the computer’s sprawling and complex arrangement of mirrors and beam splitters.
Previous records for this task involved fewer than 300 photons, but in this case Jiuzhang used 3090 particles. That is a tenfold improvement, which signals an increase in computational power. Lu and his colleagues estimated a state-of-the-art algorithm run on the world’s most powerful supercomputer would take 1042 years to simulate what Jiuzhang completed in 25.6 microseconds.
“The results are, without a doubt, an impressive technical achievement,” says Jonathan Lavoie at the Canadian quantum computing start-up Xanadu, which held a previous GBS record of 219 photons. Chris Langer at the quantum computing company Quantinuum, which has previously demonstrated quantum advantage with a different type of quantum computer, says this is a significant advance. “I think it is important that quantum systems can prove that they are not simulable,” he says.
But a Jiuzhang machine has been here before. Several times, researchers used earlier versions of the quantum computer to demonstrate GBS with high numbers of photons that seemed impossible for traditional computers to simulate. Each time, they were thwarted as classical computers replicated their results, sometimes in under an hour.
Bill Fefferman at the University of Chicago in Illinois, who worked on one of these victorious classical algorithms, says a crucial concern has impeded the photonic device: many photons get lost as they move through the quantum computer, and so the device is noisy. “Here, they reduced their rates of noise, and at the same time made the experiment larger, which – at least at the moment – seems to cause our algorithm to struggle,” Fefferman says.
Lu says overcoming photon loss was the biggest challenge his team had to meet in the new experiment. But Jiuzhang is still not fully free of noise, which leaves some room for new classical simulation strategies to challenge its champion status.
“In my opinion, they are not in the regime yet where we can be confident that no such strategy is possible,” says Jelmer Renema at the University of Twente in the Netherlands.
There is a “virtuous cycle” here, where the competition between classical algorithms and quantum devices keeps bringing us closer to understanding the elusive boundary between the classical and quantum worlds, says Fefferman. In terms of fundamental science, this is a win for everyone – but whether it moves quantum computing towards machines that are more powerful in a useful way is a separate issue.
Langer says GBS is an “entry-level benchmark” in the sense it establishes a quantum computer’s difference from conventional computers, but the achievement does not directly reflect on the computer’s usefulness. From the point of view of rigorous mathematical theory, it is difficult to assess when GBS is “smoking gun” evidence of quantum advantage and to identify a clear path for making a machine that excels at GBS into one that excels at some more applied task, says Nicolás Quesada at Polytechnique Montréal in Canada.
This is in part because Jiuzhang’s hardware is highly specialized, so the quantum computer cannot be programmed to carry out just any calculation. “While it may demonstrate a computational advantage for a narrow task, it lacks crucial elements for fault-tolerant and useful quantum computation,” says Lavoie. Here, fault tolerance refers to calculations where the quantum computer identifies and corrects its own errors, a long-sought ability that has not yet been achieved in practical quantum computers.
At the same time, Lu and his team have put forward several applications for Jiuzhang’s exceptional capability when it comes to GBS. The process can enhance computations relevant for image recognition, chemistry and certain mathematical problems related to machine learning. Fabio Sciarrino at the Sapienza University of Rome in Italy says this approach to quantum computing is still in its infancy – but if successful, it might give rise to a whole new paradigm.
Specifically, advances in hardware – like this latest Jiuzhang device – could enable researchers to build exceptional light-based quantum computers, says Sciarrino. They would be programmed in a completely new way and excel at tasks related to machine learning.
Topics:
