Google scientists announce
that they have achieved quantum supremacy. This is a long-awaited milestone for quantum computing. This announcement was published in Nature on 23 October. It follows a leak of a paper from five weeks ago that Google didn’t comment on.
A world-first, John Martinis, an experimental scientist at the University of California Santa Barbara and Google Mountain View, California led the team that created the quantum computer. It performed a particular calculation that was beyond the capabilities of regular, “classical” machines. Google estimates that the same calculation would take 10,000 years for even the most powerful classical supercomputer.
Martinis says that quantum supremacy is a landmark because it shows that quantum computers outperform traditional computers. The advantage was only demonstrated in one case. However, it is a demonstration that quantum mechanics can be used to solve complex problems.
“It seems like Google has given us first experimental evidence that quantum speeds-up can be achieved in a real-world system,” Michelle Simmons, a quantum scientist at the University of New South Wales, Sydney, Australia.
Martinis compare the experiment with a “Hello World” programme that instructs a system to display a phrase. It’s not particularly useful but it informs Google that both the quantum hardware and the software are functioning correctly.
After a draft of the paper had been leaked to NASA’s website, the Financial Times and other outlets first reported the feat in September. It was quickly removed from the site. NASA collaborates with Google on quantum computing. The company didn’t confirm the publication of the paper nor comment on it.
While the Google calculation — which is checking outputs from a quantum random number generator — may not have practical applications, Scott Aaronson, a theoretical computer scientist at The University of Texas at Austin, says that the scientific achievement “is huge, assuming and I’m guessing that it will”.
Google researchers are working to improve the classical algorithms that were used to solve the problem. This will help to reduce the company’s 10,000-year estimate. IBM, which is a rival to Google when it comes to building the best quantum computers in the world, published a preprint on 21 October claiming that the problem can be solved in 2.5 days using a new classical technique. This paper was not peer-reviewed. If IBM is right, it would lower Google’s feat to demonstrating a quantum ‘advantage. It allows Google to perform calculations much faster than a traditional computer but still within its reach. Simmons says that this would still be a landmark. “It’s the first time this’s been done, so it’s definitely a significant result.”
Quantum computers operate in a fundamentally different manner from classical machines. A classical bit can only be one or zero, but a quantum bit (or qubit) can exist in multiple states simultaneously. Qubits are interconnected in a way that allows physicists to exploit the interference between quantum states to do calculations that would otherwise take thousands of years.
Scientists believe quantum computers could one day be able to run new algorithms, such as searching large databases or factoring large numbers, including those used in encryption. These applications are still decades away. It is more difficult to keep the fragile state of qubits in a device’s operating state if there are many of them. Google’s algorithm is run on a quantum chip that consists of 54 qubits each made up of superconducting loops. This is only a fraction of the 1 million qubits needed to build a general-purpose computer.
Christopher Monroe, a University of Maryland physicist in College Park, says that the task Google gave its quantum computer was “a little bit of a strange one”. The problem was first created by Google physicists in 2016. It is extremely difficult to solve for an ordinary computer. Sycamore, the team’s computer, was challenged to predict different outcomes using a quantum-version random-number generator. This is done by running a circuit that passes 53 qubits through a series of random operations. This creates a 53-digit string consisting of 1s, 0s, and 2 53 combinations. One of Sycamore’s 54 qubits was lost so only 53 qubits were utilized. It is difficult to predict the outcome from the first principles because it is complex. However, due to the interplay between qubits, some strings of numbers are more probable to occur than others. This is similar in that it produces random numbers even though certain outcomes are more probable than others.
Sycamore determined the probability distribution by sampling a circuit, running it one million times, and then measuring the output strings. This method is similar in that it reveals the bias of the die by rolling it. Monroe says that the machine is performing a scientific task: it is using an experiment to solve a quantum problem, which is difficult to calculate classically. He says that Google’s computer can be used to create quantum circuits with any settings.
The verification of the solution was another challenge. The team then compared the results to simulations of smaller, simpler circuits. This was done using classical computers, including the Summit supercomputer at Oak Ridge National Laboratory, Tennessee. The Google team extrapolated these examples and estimated that simulating the entire circuit would take 10,000 years even with a computer with one million processing units (the equivalent of around 100,000 desktop computers). Sycamore took only 3 minutes and 20 seconds.
Google believes their proof of quantum supremacy is solid. Hartmut Neven, the head of Google’s quantum computing team, said that even if external researchers reduced the time required to perform the classical simulation, quantum hardware continues to improve. This means that conventional computers will never be able to solve this problem.
Applications are limited
Monroe believes that Google’s success might help quantum computing by encouraging more computer scientists to join the field. He also warned that the media could make quantum computers appear more mainstream than they actually are. He says, “They’ve finally beaten a regular computer. So here we are, two years and one will be in our home’.”
Monroe says that scientists have yet to prove that a programmable quantum computer can solve any useful task. Monroe cites an example: calculating the electronic structure for a specific molecule. This is a difficult problem that requires multiple quantum interactions to be modeled. Aaronson also believes that proving quantum supremacy using an algorithm known as error correction is important. This process corrects for noise-induced errors which could otherwise ruin a calculation. This is a crucial step in enabling quantum computers to work at scale, according to physicists.
Martinis says that Google is working towards these milestones and will release the results of its experiments over the next few months.
Aaronson believes that the experiment Google created to prove quantum supremacy may have practical applications. He has developed a protocol to use such calculations to show users that the bits generated from a quantum random number generator are truly random. This could prove useful in cryptography or some cryptocurrencies whose security depends on random keys.
Martinis says that Google engineers had to make several hardware improvements to run the algorithm. This included building new electronics to control quantum circuits and devising a new way for qubits to be connected. This is the foundation of how we will scale up in the future. He believes that this is the best way forward.