Google’s 67-Qubit Sycamore May Outperform Top Supercomputers

According to a new experiment, quantum computers can outperform our fastest conventional computers in extremely particular domains.

Google’s Sycamore chip performs better than supercomputers, demonstrating previously unheard-of capabilities.

Google’s 67-qubit Sycamore processor can now exceed the fastest conventional supercomputers, according to recent developments in quantum computing. This discovery, which was described in a report that was published on October 9, 2024, in Nature, points to a new stage in quantum processing called the “weak noise phase.”

Researchers at Google Quantum AI have identified a “stable computationally complex phase” that may be attained with current quantum processors, sometimes referred to as quantum processing units (QPUs).

Study under the Weak Noise Phase led by Google Quantum AI’s Alexis Morvan, a quantum computing researcher at Google, shows how quantum computers may reach this steady stage of computational complexity. The Sycamore chip can do computations during this stage that surpass the capability of conventional supercomputers. Representatives from Google claim that this discovery is a major step toward practical uses of quantum technology that are impossible for traditional computers to duplicate.

The Role of Qubits in Quantum Computing, which uses the laws of quantum physics to execute computations in parallel, are the foundation of quantum computers. This stands in stark contrast to the sequential processing of information by bits in conventional computing. Problems that would take conventional computers hundreds of years to solve can be solved in seconds by quantum machines thanks to the exponential power of qubits. Nevertheless, qubits have a greater failure rate due to their extreme sensitivity to interference; for example, 1 in 100 qubits may fail, whereas classical systems have an exceptionally low failure rate of 1 in a billion billion bits.

Despite its promise and overcoming challenges: the noise and error correction from quantum computing has several obstacles to overcome, chief among them being the noise that degrades qubit performance. According to a LiveScience study, efficient error correction techniques are required to attain “quantum supremacy,” particularly as the number of qubits rises. The biggest quantum computers available now contain about 1,000 qubits, and scaling them comes with challenging technological challenges.

Experiments from Random Circuit Sampling done by Google researchers recently tested the performance of a two-dimensional grid of superconducting qubits using a method known as random circuit sampling (RCS). RCS is considered to be one of the most difficult benchmarks in quantum computing and is used to evaluate the capabilities of quantum computers to those of traditional supercomputers.

The results showed that the researchers could move qubits into the “weak noise phase” by adjusting quantum correlations and noise levels. The calculations grew sufficiently intricate in this condition to show that the Sycamore chip could perform better than classical systems.

“We are focused on developing practical applications for quantum computers that cannot be done on a classical computer,” officials of Google Quantum AI emailed Live Science. “This study represents a major advancement in that regard. “The next challenge is to show a ‘beyond classical’ application that has practical implications.”

But since quantum computers still generate noisy data, they still have to perform a lot of quantum “error correction” as the number of qubits increases to keep the qubits in the “weak noise phase,” they said.

In contrast to traditional computer bits, which can only process data in sequence, qubits, which are integrated into QPUs, use the laws of quantum physics to do computations in parallel. A machine’s power increases exponentially with the number of qubits on its QPU. A quantum computer might do tasks that would take a classical computer thousands of years to complete in a matter of seconds because of their parallel processing capabilities.

However, qubits are “noisy,” which means that interference can cause them to malfunction; around 1 in 100 qubits fail, compared to 1 in 1 billion, billion bits. Environmental disruptions like temperature variations, magnetic forces, or even space radiation are a few examples.

A quantum computer with millions of qubits or highly advanced error-correction technologies, which are not currently available, would be required to achieve “quantum supremacy,” due to this high error rate. Since there are now only around 1,000 qubits in a single machine, scaling quantum computers is difficult.

However, a recent experiment conducted by Google scientists indicates that quantum computers can beat traditional computers in some operations and endure the present levels of noise. However, when machines grow in size, error correction could still be necessary.

The researchers also tested the fidelity of a 2D grid of superconducting qubits—one of the most prevalent kinds of qubits—using a technique called random circuit sampling (RCS). Superconducting qubits are composed of a superconducting metal held at temperatures that are very near to absolute zero. According to the experts, RCS is the most difficult benchmark to complete on a quantum computer and compares the performance of a quantum computer to that of a traditional supercomputer.

Working qubits may switch between a first phase and a second phase, known as a “weak noise phase,” by setting off specific circumstances, according to the research. The researchers either delayed the spread of quantum correlations or intentionally enhanced the noise in the tests. They came to the conclusion that a quantum computer may do better than a classical computer in this second, “weak noise phase,” since the calculation was so complicated. They used Google’s 67-qubit Sycamore chip to illustrate this.

Some representatives of Google Quantum AI stated, “This is a waypoint on the journey to get to real-world applications, or beyond classical commercial applications,” “On a traditional computer, the applications shouldn’t be reproducible. Our research’s findings represent a major advancement in that regard. You can’t win at anything else if you can’t win at the RCS standard.