Quantum Utility! IBM ushers in the practical era of quantum computing

Today, IBM (NYSE: IBM) announced a new breakthrough, published on the cover of the scientific journal Nature.

-- The team has demonstrated for the first time that quantum computers can produce accurate results at a scale of more than 100 quantum bits; and beat supercomputers in at least one type of computation. More importantly, they managed to get around quantum noise.

This time, IBM is not claiming what has often been discussed before as "quantum supremacy" or "quantum advantage": leaving aside questions about the future of traditional algorithms or the exponential quantum advantage. IBM has chosen to demonstrate impressive practical capabilities in a real-world application problem, leaving aside the debate about possible future improvements to traditional algorithms or the definition of exponential quantum advantage.

One of the ultimate goals of quantum computing is to simulate material compositions that have never been effectively modeled by classical computers. Modeling these materials is a key step in the ability to tackle challenges such as designing more efficient fertilizers, making better batteries and creating new drugs. However, today's quantum systems are inherently noisy and they generate a lot of errors, hindering performance; this is due to the fragile nature of quantum bits and interference from their environment.

In their experiments, the IBM team demonstrated that quantum computers have the potential to outperform leading classical simulations by learning and mitigating errors in the system. The team used the IBM Eagle quantum processor, which consists of 127 superconducting quantum bits on a chip, to generate large entangled states, simulate the spin dynamics in a material model, and accurately predict its properties such as magnetization.

IBM has used their Eagle R3 processor to demonstrate the efficient execution of a 127Q circuit with a gate depth of 60, which it did in a class of problems (time evolution of a two-dimensional transverse field Ising model).

To verify the accuracy of this modeling, a team of scientists at UC Berkeley ran these simulations on both the National Energy Research Scientific Computing Center (NERSC) at Lawrence Berkeley National Laboratory and an advanced classical computer at Purdue University. As the models grew in size, the quantum computers produced accurate results with the help of advanced error mitigation techniques: even the classical computing methods, even when they eventually went awry, were no match for the IBM quantum system.

"This is the first time we've seen a quantum computer accurately model physical systems in nature beyond the leading classical methods." Darío Gil, IBM senior vice president and director of research, said, "This milestone is an important step for us, demonstrating that today's quantum computers are capable, scientific tools that can be used to model extremely difficult (and perhaps impossible) problems with classical systems, marking the fact that we are now entering a new practical era of quantum computing era."

Back in 2017, researchers on IBM's Quantum team announced a breakthrough: the successful simulation of the energy of small molecules, a Nature cover story Hardware-efficient Variational Quantum Eigensolver for Small Molecules and Quantum Magnets, shows how to implement a new quantum algorithm that can efficiently calculate the lowest energy states of small molecules (lithium hydride and beryllium hydride) using a quantum computer.

But because of the noise in the system at the time, they fell far short of the accuracy or scale of interest to chemists. These simulations were exciting - the scientists did something with quantum computers; at the same time, the real breakthrough was that scholars had an idea of what was wrong with these simulations.

Around the same time, the team published a theoretical paper Error Mitigation for Short-Depth Quantum Circuits and set an important signpost: if we can really understand what causes noise, then we can potentially eliminate its effects; then maybe we can take the noisy quantum computers and then maybe we can extract useful information from noisy quantum computers to solve certain types of problems.

The team realized that the same technique used to control quantum bits could be used to amplify the effects of noise through a technique called "pulse stretching. Essentially, if we increase the time to operate on each quantum bit individually, then we amplify the amount of noise by the same factor.

Pulse spreading allowed the team to dramatically improve the accuracy of LiH simulations with four quantum bits in 2019. But one question remains: how far can these methods be scaled?

In the team's 2023 paper Scalable error mitigation for noisy quantum circuits produces competitive expectation values, published in Nature, they extend these error-mitigating simulations to 26 quantum bits, a model that scales reasonably well with the size of the quantum computer - modeling the noise of a large processor is no longer a daunting task. This implies to the team that these methods can produce results that are much closer to the ideal answer than the approximations effectively obtained from classical computers.

This essentially lays out the blueprint for our current work. The team says, "If we can improve the scale and quality of the hardware and develop methods that can control the amplification noise to a greater extent, perhaps we can estimate the desired value to a degree of accuracy that will make it useful for practical applications."

Amplification of noise is the final piece of the puzzle. With a representative noise model, one can manipulate and amplify the noise more accurately; then, using the Zero Noise Extrapolation (ZNE) method, one can extrapolate the results of a noise-free calculation.

This diagram illustrates the basic principle of ZNE noise amplification, a method for mitigating errors in quantum systems. With the ZNE technique, the team amplifies the noise in the system to different levels and evaluates the noise at each level. The evaluated data can then be combined with a number of extrapolation methods, enabling the scientists to extrapolate back to the zero-noise limit.

Meanwhile, error mitigation requires high-performance hardware. With IBM's 127-bit IBM Eagle processor, the team finally had a system capable of running circuits large enough. According to the company, "Now, it's time to use error mitigation to test state-of-the-art processors."

There are several ways to run quantum circuits with classical computers. The first is the "brute force (brute force)" method - calculating expectations, which is similar to how a physics student would calculate expectations by hand. This involves first writing all the information about the wave function into a list and then creating a lattice of numbers (also known as a matrix) to perform the calculation.

The difficulty of these methods doubles with each additional quantum bit, so that ultimately the complexity of a large enough circuit cannot be captured. But for a small fraction of quantum circuits, these circuits and methods can be used first to benchmark classical and quantum methods.

This graph shows the performance of quantum computers compared to state-of-the-art classical approximation methods on a range of increasingly challenging computational problems.

The graph compares the performance of quantum computers with classical approximations in areas beyond the capabilities of exact classical "brute force" methods.

The experiment will be conducted as follows: The IBM team will use all 127 quantum bits of the IBM Eagle processor to simulate the changing behavior of a system (the quantum Ising model) that naturally maps to a quantum computer. The Ising model is a simplification of nature that represents interacting atoms as a lattice of quantum dichotomous systems in an energy field. These systems look a lot like the two-state quantum bits that make up a quantum computer, which makes them well suited for testing the experiments described above. the IBM team will use the ZNE to try and accurately calculate one property of the system, average magnetization, which is an expectation value: essentially weighted average of the possible results of the circuit.

Meanwhile, the UC Berkeley team will attempt to simulate the same system using a tensor network approach with the help of advanced supercomputers at Lawrence Berkeley National Laboratory's National Energy Research Scientific Computing Center (NERSC) and Purdue University. Specifically, the calculations will run partly on NERSC's Cori supercomputer, partly on the Lawrencium cluster inside Lawrence Berkeley National Laboratory, and partly on Purdue's NSF-funded Anvil supercomputer. The team will then compare the two methods with the exact method to see how well the two work.

Initially, the quantum method continued to be consistent with the exact method; however, as the difficulty increased, the classical approximation began to "falter.

Finally, the team asked the two computers to perform calculations that went beyond the exact calculation: Eagle returned the exact answer every time. Observing the performance of both computational models in simulating increasingly complex situations led both teams to agree that the quantum computer could return more accurate answers than the classical approximation.

127 Characterization of quantum bit circuits.

Zero-noise inference with probabilistic error amplification.

Classically verifiable expectation values for 127-bit Clifford and non-Clifford circuits.

Estimation beyond precisely verifiable expectations.

However, it is not clear whether quantum computing indisputably beats the classical techniques of the Ising model.

Many believe that commercial applications in the NISQ era are increasingly unlikely. Now, IBM has not proven that these will be realized, but they have moved the dial significantly in that direction.

IBM Quantum and the University of California at Berkeley have presented evidence that NISQ quantum computers will be able to deliver value sooner than expected, all thanks to advances in IBM quantum hardware and the development of new error mitigation methods.

These results validate IBM's short-term strategy, which aims to provide useful computing by mitigating, rather than correcting, errors. However, this work excites us! For a number of reasons:

- It is a realistic scenario for exploring meaningful computation and realistic applications using currently available IBM quantum processors, ahead of the era of fault tolerance.

- In addition to providing a proof of principle, the team has provided results that are accurate enough to be "useful".

It is important to note that this is not to say that today's quantum computers exceed the capabilities of classical computers - other classical methods and specialized computers may soon return computationally correct answers to tests. But that's not the point. The continued back and forth between quantum running complex circuits and classical computers verifying quantum results will improve both classical and quantum systems simultaneously, while giving users confidence in the capabilities of near-term quantum computers.

Right now, the field mostly agrees that realizing the full potential of quantum computers (e.g., running Shor's algorithm to decompose large numbers into prime numbers) will require error correction: fault tolerance is the ultimate goal, and error mitigation is the path to making quantum computing useful. However, there is still debate as to whether recent quantum hardware can provide computational advantages for useful problems before error correction is fully implemented. This paper gives us good reason to believe that noise-containing medium-scale quantum computers will be able to provide value before the era of error tolerance arrives - and that includes processors that are available today.

"The key to this work is that we can run a fairly substantial deep circuit right now using all 127 of Eagle's quantum bits, and the numbers turn out to be correct." said Kristan Temme, principal investigator and manager of IBM Quantum's Quantum Algorithm Theory Group.

IBM scientists see error mitigation as an interim solution that can now be used to solve increasingly complex problems outside of the Ising model. The paper is a milestone that shows we are entering the era of quantum dominance. Quantum dominance will be a continuous path that requires two things:

First, we must show that quantum computers can outperform classical computers.

Second, we must find problems useful for such acceleration and figure out how to map them to quantum bits.

This paper gets us across the first point. the IBM team says, "It immediately points to the need for new classical methods. And they're already working on those methods. Now, we're asking ourselves, can we take the same error mitigation concept and apply it to classical tensor network simulations and see if we can get better classical results?"

Meanwhile, for quantum researchers, it's a learning process: "How can we optimize our calibration strategy to run quantum circuits like this one? What can we expect, and what do we need to do to improve things in the future? These are all good things we discover along the way as we run our projects."

IBM's 100×100 Challenge: IBM believes that successful implementation of "quantum advantage" requires at least 100 quantum bits and 100 gates deep. This experiment shows that IBM has achieved 127×60.

"This work gives us the ability to perhaps use quantum computers as a validation tool for classical algorithms."

IBM researchers, including from left, Abhinav Kandala, Kristan Temme, Katie Pizzolato, Sarah Sheldon, Andrew Eddins and Youngseok Kim, with their quantum computer.

Following this groundbreaking work, IBM also announced that its IBM Quantum systems running in the cloud and at partner sites will be powered by at least 127 quantum bits and are expected to be completed next year. "With this transition, all of our users will have access to systems like the ones used in this study."

These processors offer computational power large enough to surpass classical methods for some applications and will provide improved coherence times and lower error rates than previous IBM quantum systems. This capability can be combined with advancing error mitigation techniques to bring IBM quantum systems to a new threshold for the industry - what IBM calls "utility-scale" - at which point quantum computers can be used as a scientific tool to explore new scale problems that classical systems may never be able to solve.

"As we achieve useful quantum computing, we have solid evidence of the building blocks needed to explore entirely new computational problems," said Jay Gambetta, IBM Fellow and vice president of IBM's Quantum Division. processors, we are inviting our customers, partners and collaborators to bring their hardest problems to explore the limits of today's quantum systems and begin extracting real value."

Now, all IBM quantum users will be able to run problems on utility-scale processors larger than 100 quantum bits; researchers and industry leaders worldwide are leveraging IBM quantum technology in pursuit of value.

Because of the immediate potential of quantum, as IBM expands its quantum technology stack, leaders in research institutions and the private sector are mobilizing industries and equipping themselves with more powerful quantum technologies. This includes advanced hardware and tools to explore how error mitigation can enable accuracy; a number of pioneering organizations and universities are working with IBM to advance the value of quantum computing. These working groups include

- Healthcare and Life Sciences: led by organizations such as the Cleveland Clinic and Moderna, which are exploring the application of quantum chemistry and quantum machine learning to challenges such as accelerated molecular discovery and patient risk prediction models;

- High-energy physics: consisting of groundbreaking research institutions such as CERN and DESY, which are working to identify the most appropriate quantum computing methods for areas such as particle collision event identification and reconstruction algorithms, and to investigate theoretical models of high-energy physics;

- Materials: led by teams from Boeing, Bosch, the University of Chicago, Oak Ridge National Laboratory, ExxonMobil and the RIKEN Institute in Japan, aiming to explore the best methods for establishing materials simulation workflows;

- Optimization: aims to build collaborations between global institutions such as E.ON, Wells Fargo and others to explore key issues and drive the identification of optimization problems best suited to quantum advantages in sustainability and finance.

As a leading global provider of hybrid cloud and artificial intelligence and consulting expertise, IBM helps clients in more than 175 countries leverage the insights of their data, streamline business processes, reduce costs and gain a competitive advantage in their industries.

IBM expects to unveil its most powerful processor, the 1121-quantum-bit Condor chip, later this year (2023). IBM also has "utility-scale processors" with up to 4,158 quantum bits in development, said Jay Gambetta, head of IBM's quantum technology efforts.

"To achieve the long-term goal of building a 100,000-bit quantum machine capable of full error correction by 2033, researchers will need to solve a large number of engineering problems," Gambetta added. Gambetta added.

Reference links:

[1] https://newsroom.ibm.com/2023-06-14-IBM-Quantum-Computer-Demonstrates-Next-Step-Towards-Moving-Beyond-Classical-Supercomputing

[2]https://www.nature.com/articles/nature23879

[3]https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.119.180509

[4]https://www.nature.com/articles/s41567-022-01914-3

[5]https://research.ibm.com/blog/utility-toward-useful-quantum

[6]https://www.nature.com/articles/d41586-023-01966-2

[7]https://phys.org/news/2023-06-technique-error-prone-quantum-classical.html

[8]https://www.nytimes.com/2023/06/14/science/ibm-quantum-computing.html

[9]https://www.newscientist.com/article/2378229-ibm-quantum-computer-beat-a-supercomputer-in-a-head-to-head-test/

[10]https://www.forbes.com/sites/karlfreund/2023/06/14/ibm-achieves-breakthrough-in-quantum-computing/?sh=68757d5b2832

[11]https://quantumcomputingreport.com/ibm-cuts-through-the-noise-whats-the-potential-impact/