In October 2019, the Google Quantum AI team published a well-known paper that used Random Circuit Sampling (RCS) to show a problem that could be solved by their Sycamore quantum computer in 200 seconds and would take 10,000 years to solve on the world’s fastest supercomputer. This experiment, called the Quantum Supremacy Experiment (now called the “Beyond Classical” experiment), used 53 qubits with a gate depth of 20 levels. (See our in-depth report about this here.)
Almost immediately, several challenges came to the experiment from both the classical and quantum sides. First, IBM published a blog describing how a different classical approach can implement this calculation in 2.5 days instead of 10,000 years. About two years later a Chinese group replicated the experiment using 60 qubits to a gate depth of 24 levels. And another Chinese group showed how the calculation could be completed in 5 days using a cluster of 60 NVIDIA GPUs.
One important lesson from this is that classical computing technology is not standing still either. There continues to be significant innovation in using classical techniques to solve problems previously thought insolvable. So the quest for quantum advantage is one where the bar is continually being raised year after year.
However, an important characteristic of the RCS problem, like many other quantum oriented algorithms, is that the resources needed to solve it doubles every time you add a qubit. So Google has rerun this experiment using a second generation Sycamore processor that used 70 qubits to a gate depth of 24 levels and they are again showing that their Sycamore processor can outperform the fastest classical system even with the classical system using the latest software techniques.
To read more about this new experiment, you can view a new arXiv preprint from the Google Quantum AI team here.
April 29, 2023