Researchers Challenged Google’s ‘Quantum Supremacy’ With Ordinary Supercomputer

The goal of quantum supremacy is to complete the most difficult task faster than the most powerful supercomputer

Back in 2019, Google boldly said that they had attained the proof that the esoteric approach could surpass conventional ones—something quantum computing specialists had been looking for for years. However, scientists are disputing this instance of "quantum supremacy," asserting that they outperformed Google on a pretty ordinary supercomputer.

No one is accusing Google of lying or misrepresenting its work, and the meticulous and innovative research that resulted in the quantum supremacy declaration in 2019 is still of utmost significance. The rivalry between classical and quantum computing, however, is still anyone's game if this latest report is accurate.

Although it may seem like a cop-out, the goal of quantum supremacy is to demonstrate the method's viability by identifying even one extremely specialized and peculiar work that it can complete faster than the most powerful supercomputer. Since that opens the door for the quantum to add more tasks to its task library. Perhaps, in the long run, all jobs will be performed more quickly in quantum, but for Google's needs in 2019, only one was, and they thoroughly explained how and why.

Currently, a group from the Chinese Academy of Sciences directed by Pan Zhang has released an article outlining a novel method for simulating a quantum computer that seems to take a minuscule fraction of the time predicted for classical computation in 2019.

To depict the 20 cycles that the Sycamore gates went through throughout the simulation, they represented the issue as a sizable 3D network of tensors, with the 53 quantum bits in Sycamore being defined by a grid of nodes that had been extruded out of 20 times. A network of 512 GPUs was then used to calculate the mathematical relation between these tensors.

Although it should be noted that this was Google's estimate for 54 quantum bits performing 25 cycles; 53 qubits undertaking 20 cycles is significantly less complex but would still require on the order of a couple of years by their estimation method. The original Google paper estimated that running this scale of modeling on the most robust supercomputer at the time would take approximately 10,000 years.

According to Zhang's team, it only took them 15 hours. And if they had access to a powerful supercomputer like Summit, they could complete the task more quickly than Sycamore, in a matter of seconds.

There is no reason to believe that these results are a mistake or a scam, but they have not yet been thoroughly examined and verified by persons with relevant information. As it is extremely difficult to construct and program quantum computers, while conventional ones and their software are continually being upgraded, Google even acknowledged that the baton may be transferred back and forward a few times before supremacy is permanently established.

Google hasn't been sitting still either, so it very well could respond with fresh assertions of its own. However, the fact that it is even competitive is a positive thing for everybody involved because this is a fascinating field of computing and research like Zhang's and Google's keeps everyone's standards higher.