Is quantum advantage moving too fast for you to keep up?
An overview of the recent developments in quantum computing
Vienna, 12th of October, 2019:
Eliud Kipchoge became the first human to run a sub 2-hour marathon. Once believed to be unattainable, his time, 1:59:40, will not be officially recognized as a world record [1]. The reason for this is that the experiment to run 26.2 miles in under two hours was conducted under specific conditions which are not comparable to standard racing situations. Having an electric car and a team of 41 world-class runners in front of him, setting the pace, he broke “the last barrier of modern athletics”. Or to put it in his own words: “Vienna is about running and breaking history, like the first man on the moon.”
Although the final proof in terms of a sub-2 hours world record is still missing, he demonstrated that it is actually achievable, thereby inspiring a whole generation of upcoming athletes, thanks to a storm of #nohumanislimited hashtags and the vast amounts of money invested by the organizers and sponsors of this “historical” event.
Now you might wonder why is this quantum physics nerd telling you all this in a science page? Maybe -just maybe- because she simply enjoys running. But, no. There is indeed something similar happening in the latest progress of quantum computing!
Quantum computers work fundamentally different than any other so-called “classical” computer, may it be an ordinary laptop or a supercomputer. Many hopes have been placed in quantum computers; an unbreakable encryption, the discovery of new materials and drugs, solutions to our energy and climate problems, an economic boost. However, they are still at the development stage. One major step towards fulfilling these expectations is to prove their advantages over any other computer. And here lies the conjunction between Eliud Kipchoge‘s run of the century and the recent progress in quantum computing.
This article is an attempt to approach the topical question: Did a quantum computer outperform any computer on this planet and therefore broke the barrier of classical computing?
Google’s headquarters in Mountain View, California, 23rd of October, 2019:
In their nature paper “Quantum supremacy using a programmable superconducting processor” [2], Google claims that they have achieved the first experimental evidence of a computational task that can only be performed on quantum processors. They report that their quantum processor, called Sycamore, needs about 200 seconds to solve this specific task, whereby today’s best supercomputers would take approximately 10,000 years. That is 26 million times faster than any state-of-the-art supercomputer, or in other words, practically impossible.
IBM Q’s headquarters in New York, two days earlier, 21st of October, 2019:
Since a first draft of the paper was leaked on the website of NASA in September (yep, nerdyness does not make immune against leaking), the scientific community outside Google was already prepared and have started to carefully examine the findings. As a result, IBM, the strongest actual competitor in building quantum computers, argues in a non peer-reviewed preprint [3] that the classical algorithm can significantly improve. IBM claims that “an ideal simulation of the same task can be performed on a classical system in 2.5 days”. This is “only” a thousand times faster.
The speed up of 26 million and a thousand times faster seems far apart from each other, however, there is only a thin line between performing a task much faster than on a supercomputer, and performing a task so much faster that it is beyond reach for any classical machine. From a broader point of view, speed is only one alerting parameter. It is also about the task. So, let us get our hands dirty and enter the depths of quantum computing.
The algorithm runs on a quantum processor named “Sycamore” which consists of 53 qubits*, each forming a loop of on qubit coupled to the four nearest neighbors.
A Qubit is fundamentally different from a classical bit, which is either a 1 or a 0 and forms the basic unit of information processing in ordinary computers. Qubits, however, exploit the wave characteristics of quantum particles in order to create a new type of informational unit which is neither a 1 nor a 0 but best described by multiple states at once, i.e. a superposition of 1s and 0s. The unique computing potential of a quantum computer is not only based on these single quantum states, but rather on the wave-like interactions between them. In theory, this enables a network of qubits to perform computations exponentially faster than conventional computers of the same size in terms of informational units.
The problem, Sycamore was asked to solve, was carefully chosen, keeping in mind that it needs to be a tough issue for classical computers. Having said this, we try to formulate the task in just one sentence: “Sycamore, toss 53 loaded coins and tell us how likely the obtained 53-digit string of heads and tails is!” Doesn’t sound like a really hard problem at all? Oh, yes, there are ²⁵³ (1 quadrillion) combinations!
Nonetheless, Sycamore is capable of solving this by applying a circuit of random operations on the qubit network. The result leaves the 53 qubits in a random state. Once measured, the state collapses into a sequence of 0s and 1s or heads and tails, if you wish. Some outcomes, however, are more likely than others due to the quantum features of the pre-processed qubit state. In their paper Google describes this as follows: “Owing to quantum interference, the probability distribution of the bitstrings resembles a speckled intensity pattern produced by light interference in laser scatter, such that some bitstrings are much more likely to occur than others.” The speckled intensity pattern is just another metaphor as throwing a loaded coin. It essentially means that the quantum circuit outputs a random string of 0s and 1s with some results having a higher probability. Throwing the coin often enough can reveal its bias. Sycamore does so by running the circuit one million times and subsequently measuring the observed output strings. It takes the quantum processor approximately 3 minutes and 20 seconds to obtain the final probability distribution of the outcomes.
Emulating this computation on a state-of-the-art supercomputer is much more difficult. By extrapolating the results from simpler versions of the quantum random-number generator, the Google team estimates that today’s best classical computer would take 10,000 years to solve the problem. But strong criticism has being made by IBM: “We urge the community to treat claims that, for the first time, a quantum computer did something that a classical computer cannot with a large dose of skepticism due to the complicated nature of benchmarking an appropriate metric.” In their eyes, an improved simulation, based on performance-enhance data storage and an optimized usage of the classical hardware, can be performed on a classical system within 2.5 days. Another doubtful point, addressed by IBM, is the usefulness of the task since any application is still missing. However, back in the mid-seventies, almost no-one could have imagined all the applications of the newly emerging personal computer.
Assessing the findings of Google and IBM finally boils down to the question: What does it mean that a quantum computer is “supreme” over classical computers? ⁺
In search of an answer, quantum physics is possibly experiencing the most lively and intriguing debate since the correspondence between Niels Bohr and Albert Einstein about the fundamental nature of the quantum world [4]. What is certain as of now is that the Google machine is solving a computational problem in a fundamentally different way than classical computers.
Eliud Kipchoge was the first man to run 26.2 miles in less than two hours. Maybe Sycamore is the first quantum computer to outperform any supercomputer. In the former case, agreeing on a set of rules for defining the world record helped to judge his performance but did not diminish any enthusiasm or fascination.
Therefore, the recent debate and the ongoing race of technologies has the potential to change the public view on the feasibility of quantum technologies, and inspire a new generation of scientists to come up with improved and truly applicable usage of quantum computing. Maybe, we are only a few creative ideas away from breaking history, like the first man on the moon. We have already taken off.
* In fact, Sycamore was constructed out of 54 qubits but one is broken, that is why we refer to 53.
⁺ Or shouldn’t we better think about them as a fundamentally different concept which can also work in concert with supercomputers?
[1] https://www.nytimes.com/2019/10/12/sports/eliud-kipchoge-marathon-record.html
[2] https://www.nature.com/articles/s41586-019-1666-5
[3] https://arxiv.org/abs/1910.09534
[4] https://en.wikipedia.org/wiki/Bohr%E2%80%93Einstein_debates
[5] https://ai.googleblog.com/2019/10/quantum-supremacy-using-programmable.html
