Unless you have been living in a cave for the past week, you could not avoid all the press and hype regarding Google’s official announcement and paper in Nature magazine that describes their successful completion of their quantum supremacy experiment. And you probably also saw IBM’s rebuttal that the speedup for Google’s random benchmark experiment was not as large as Google claimed. For those of you who want to read more about this, here are some links of relevant blogs and papers covering this.
Google has been working on this experiment for quite some time. We originally heard about it in December 2017 and it was originally envisioned as using a 49 qubit chip. The design was subsequently modified to be a 72 qubit Bristlecone chip and then modified again to become the 54 qubit Sycamore chip. The underlying purpose of the experiment was to provide a mechanism to drive the engineering team to produce the best possible quantum superconducting system that they could achieve in a reasonable timeframe.
Our first observation about their chip is that is a very good chip. Not only are the number of qubits (53 working out of 54) among the leaders, but each qubit has connectivity to its four nearest neighbors. This is better than what we have seen in other superconducting designs and will allow more gate operations to occur in the same time slot. Although cross-talk can be an issue the Google design incorporated tunable couplers that helps to isolate neighboring qubits from each other when necessary. The reported single qubit, two qubit, and readout fidelities also are very competitive with a reported average single qubit error rate of 0.16%, two-qubit rates of 0.93% or less, and readout errors of 3.8%. Another impressive aspect of the design are the gate delays which are in the 10’s of nanoseconds. In order to optimize the qubit quality, Google utilized Quantum Benchmark’s True-Q™software for suppressing residual calibration errors and providing information on the performance of quantum gate operations.
So an obvious questions arises. If Google has a 53 qubit design and IBM has a 53 qubit design, which is better? At this point, we don’t know because we are still waiting for IBM to release additional information on their recent 53 qubit computer. But we do think that the Google hardware design sets a high bar that will make for an interesting comparison.
Some people may think that the Sycamore chip was designed specifically for this quantum supremacy experiment and would not perform as well for potential applications programs such as quantum chemistry, quantum machine learning, or optimizations. That is not the case. All the improvements that Google has incorporated into the design will also allow it to perform very well on other programs that are intended for use with NISQ quantum computers.
We have previously remarked that progress in quantum computing will spark additional innovation in classical computing and we saw this again with IBM’s response. In Google’s original paper, they estimated it would take a classical computer 10,000 years to solve the same problem, but then IBM came back and said it could be done in 2 ½ days using a different type of classical simulator that leveraged hard disks as second storage for qubit states. Although we are no sure if this specific example is highly important, because the difference between classical and quantum performance will increase dramatically on this particular benchmark as the number of qubits is increased. However, it is representative of a phenomena we have seen time and again which is: Classical computing won’t give up without a fight! Classical computing scientists are always working to improve their algorithms so beating a classical computer represents a moving target. And we expect that performance battles will continue going on forever in quantum computers, just like they have for the past 70 years with classical computers.
As impressive as Google’s achievement is, there was another achievement we reported on a month ago that may be just as important. Whereas the Google benchmark was run on a specific artificial benchmark that may have very limited commercial use, Oki Data has reported that they have used a D-Wave quantum annealer to solve a problem of commercial significance. They developed a program that optimized the manufacturing flow in their LED Management Factory that reduced the flow line distances by 28%. Oki indicated that achieving this would not be possible if they only were able to use classical computers. While some folks have compared the Google experiment to the first Wright brothers airplane flight in December 1903, perhaps we can compare the Oki Data achievement to the establishment of the first commercial airplane passenger service in January 1914. This achievement did not receive as much press as the Google announcement; perhaps because Google has a much larger PR team than Oki Data.
But to summarize our view, completion of this quantum supremacy experiment is a significant achievement and the Google team needs to be congratulated for their work. But we should ignore much of the hype floating around that states it will start creating enormous change immediately. One should view quantum computing as a marathon race and achieving this benchmark is like passing the one mile marker in a very long race. Many decades of hard work still lie ahead. Not only will the engineers and scientists need to continue to optimize the chip and hardware systems designs, but just as much work, if not more, will be needed to develop software, libraries, application program, user communities and a quantum educated workforce that can take advantage of the technology.
October 26, 2019