Google's quantum chip beats fastest supercomputer

Experiments on Google's 67-qubit Sycamore processor show operations entering the "weak noise phase," where calculations are complex enough to outpace supercomputers.

Quantum computers can beat the fastest classical computers in a specific area, according to a groundbreaking experiment. The Google Quantum AI team found that it is possible to achieve a 'complex and stable phase' with off-the-shelf quantum processing units (QPUs). This means that when quantum computers enter this low-noise phase, they can perform complex calculations faster than the fastest supercomputers. Alexis Morvan, a quantum computing researcher at Google, and colleagues published their findings on October 9 in the journal Nature.

Google's Sycamore quantum chip. (Photo: Cnet).

'We're focused on developing practical applications for quantum computing that are not possible on classical computers ,' said a Google Quantum AI representative. 'This research is an important step in that direction. Our next challenge is to demonstrate an application that has real-world impact.'

Quantum bits (qubits) in a QPU rely on the principles of quantum mechanics to run calculations in parallel, while classical computer bits can only process data sequentially. The more qubits a QPU has, the more powerful the machine becomes. Due to parallel processing capabilities, calculations that would take thousands of years to perform on a classical computer can be completed in seconds on a quantum computer.

But qubits are noisy, meaning they are extremely jumpy and prone to malfunctioning when subjected to disturbances—roughly one in 100 qubits compared to one in a billion billion bits. Examples include environmental disturbances such as temperature changes, magnetic fields, or even radiation from space. The high error rate means that to achieve quantum supremacy, researchers need either highly sophisticated error-correction technology or quantum computers with millions of qubits. Scaling up quantum computers is no easy task; the largest number of qubits in a machine today is only about 1,000.

New experiments by Google scientists show that quantum computers can withstand current noise levels and outperform classical computers on certain calculations. However, error correction will still be necessary as the machines scale up.

The team used a random sampling method called random circuit sampling (RCS) to test the reliability of a 2D superconducting qubit lattice , one of the most common types of qubits made of superconducting metals suspended at temperatures close to absolute zero. RCS is a measure of how well a quantum computer performs compared to a classical supercomputer.

The results of the experiment revealed that qubits can switch between the first phase and a second phase called the 'low noise phase' by activating certain conditions. In the experiment on Google's 67-qubit Sycamore chip, scientists increased the noise, or slowed down the spread of quantum correlations. In the low noise phase, the computation was complex enough for them to conclude that quantum computers could perform better than classical computers.