Quantum computing could revolutionize information technology by harnessing the strange principles of quantum mechanics. While there is growing hype surrounding its potential, the reality is a mix of groundbreaking progress and persistent technical challenges.
In 2019, Google claimed to have reached quantum supremacy, a technological breakthrough where quantum computers overtake classical computers in capabilities. In reality, a classical computer managed to reproduce the results a few months later. More recently, Microsoft announced progress toward developing topological quantum bits, or qubits, which are designed to be robust against errors and well-suited for quantum computing architectures. While the announcement has generated interest, some researchers have expressed caution and are awaiting further validation. Both cases show an eagerness by some to claim early successes towards a practical, useful quantum computer, which in reality may still be decades away.
There is enormous potential in quantum computing, and companies and governments are investing billions in R&D programs in quantum computing technology. There have been many impressive results already, but this environment can also create an incentive towards hype. "The hype is sustaining the field a bit," says Vincenzo Savona, professor at the EPFL Laboratory for Theoretical Physics of Nanosystems. "It is detrimental because it might create a feeling of disillusionment if progress takes longer. But the fact that there is hype doesn't mean that quantum computing is not feasible."
An altogether different computer
Quantum computing takes advantage of the properties of subatomic particles - superposition, entanglement and interference - to carry out calculations. While classical bits can take two discrete values, 0 or 1, qubits can take any combination between these two values, opening the door to a new universe of computation possibilities.
In theory, quantum computers could tackle problems that classical computers might never solve or would need a very long time to do so. However, that does not mean that quantum computers can solve any problem, or even any problem that classical computers can solve; rather there are some problems that quantum computers are more suited to addressing efficiently. For this reason, quantum computers are not necessarily faster computers, and will never replace classical computers completely.
Building quantum computers requires creating new types of designs to take advantage of the unique rules of quantum physics. Among the most promising quantum computer architectures are those that use superconducting qubits and trapped ions. "Superconducting qubits are, at the moment, the most mature architecture, but we are still far from having a type of qubit comparable to the transistor that enabled microprocessors," says Edoardo Charbon, professor at the EPFL Advanced Quantum Architecture Lab.
In fact, the future of computing envisions a new paradigm where classical and quantum coexist and complement each other. "The integration of quantum and classical architectures is probably the most interesting and challenging of the problems that we need to solve today," Charbon says. Currently, there are large European projects integrating quantum computers in high-performance computing centers to explore this approach.
Applications on the horizon
One of the areas where quantum computing is expected to have a real advantage over classical computers is in the simulation of quantum mechanical systems. Proposed by Richard Feynman more than 40 years ago, the simulation of quantum systems can help explain the phenomena that underlie some physical processes. This knowledge can directly translate to the development of new drugs, better batteries and new materials for electronic circuits, for example. "We could accurately predict properties such as the ground-state energy, excitation spectrum, or thermal behavior of a new material, enabling the identification of phase transition points under different conditions," says Yihui Quek, who was recently appointed as assistant professor of computer science at EPFL.
Another area where quantum computers can represent a significant step forward is in cybersecurity and cryptography. For example, Shor's algorithm shows that future quantum machines might crack widely used encryption like RSA much more efficiently than today's computers. Although such powerful quantum computers are still years away, this possibility has prompted researchers to develop post-quantum cryptography - new encryption methods designed to resist quantum attacks. Meanwhile, quantum cryptography, such as quantum key distribution (QKD), uses the laws of quantum physics to create ultra-secure communication channels. Together, these approaches aim to protect digital systems in a world where quantum technology becomes mainstream.
Correcting errors, one qubit at a time
Quantum computing faces many significant challenges before it can meet practical applications. The main obstacle is the susceptibility to interactions with the external environment, leading to a loss of the quantum state and information leakage. To mitigate such loss, it is key to implement error correction to protect quantum information. "Error correction is the most important feature that we need to develop," says Savona. "Fortunately, we are witnessing constant progress."
Correcting errors tends to require a higher number of qubits. The complexity of quantum computing technologies involving superconductivity, ultralow temperatures, high vacuum and complex photonics, currently limits quantum computers to just over 1,000 qubits. While some optimization problems can get results from these small-sized computers, many commercial applications will require between 10,000 to several million logical qubits. More breakthroughs are needed to achieve high-quality computers of this size.
At EPFL, researchers are working on hardware to develop quantum simulations with trapped ions, scale up fluxonium superconducting qubits, and couple qubits to new resonators and mechanical systems.
A new way of interfacing
Another area of intense development is quantum algorithms. The language in which these algorithms are written is not just another programming language but a completely new way of interfacing with the computers. The design of efficient quantum algorithms is essential to producing practical applications and achieving a real advantage with respect to classical algorithms. "We do not know yet how to take a given problem and produce a quantum algorithm that can speed it up," explains Quek. "There is a global scientific effort underway to search for more quantum algorithmic primitives so that we can find more use cases for quantum computers."
"In the last ten years, EPFL has significantly strengthened research and development in the area of quantum technologies and quantum computing," says Savona, who adds, "EPFL is emerging in Switzerland as the academic hub for quantum algorithms."
Artificial intelligence helps find complex patterns in data faster and more accurately. Quantum computing uses the principles of quantum mechanics to process information in ways that classical computers might only dream of. What if we could combine the advantages of both fields?
Some people envision quantum computers as the next step in the evolution of AI. They argue that quantum principles, such as superposition and entanglement, may enable faster training of complex AI models or help process data more efficiently. In other cases, quantum computers could help AI overcome limitations to tackle problems that classical computers are not capable of solving in areas such as drug discovery and materials science.
However, quantum computing is still in its infancy and some of these claims are still more an aspiration than a fact. "There is no special reason to believe that quantum computing is going to be especially helpful to efficiently solve tasks such as classifying images or generating text," says Giuseppe Carleo, leader of the EPFL Computational Quantum Science Laboratory. "Probably, quantum computing is not going to be a revolutionary tool for machine learning."
On the other hand, AI and machine learning have already proved to be key elements in helping quantum computing evolve in several areas. For instance, AI helps characterize quantum devices, evaluating quantum computer performance - essentially it's how we "look inside" a quantum system to understand what it's doing. Another application where AI stands out is identifying quantum errors and quickly correcting them. In this particularly important task, deep learning models perform much better than classical algorithms. "These developments are not speculative; these are very well established as a field," comments Carleo.
EPFL is a key player in developing this type of synergy between AI and quantum computing. For instance, the Laboratory of Quantum Information and Computation, led by Zoë Holmes, is pushing the limits of the synergies between quantum computers and machine learning algorithms. Meanwhile, the group led by Carleo is developing machine learning models to accurately simulate quantum systems, "in the spirit of the AlphaFold program from Google, but rather than applying it to proteins, we're applying our machine learning models to solve the Schrödinger equation, the founding equation of quantum mechanics, to accurately simulate physical systems," says Carleo.
Quantum for AI or AI for quantum? Working together within EPFL's Quantum Science and Engineering Center, Holmes, Carleo and their colleagues are committed to making the best out of both worlds.
