Last December at Q2B24 in Silicon Valley, top academics, industry end users, government representatives, and quantum computing vendors gathered from all over the world to speak. They focused on six major quantum technology themes, computing, sensing & time, communications & security, quantum AI, error correction, and on-premises quantum computers. With over forty speakers, many experts shared their thoughts and insights regarding quantum innovation over the past year.
Two industry luminaries gave plenary presentations at the conference. Scott Aaronson, a computer science professor at UT Austin, presented an insightful talk about the state of quantum algorithms in 2024. He examined key results in quantum computing, people’s responses to the field’s progress, and the potential for breakthroughs. Prof. John Preskill, the director of the Institute for Quantum Information & Matter at Caltech, talked about a several aspects of quantum computing. Detailing the path to quantum value, challenges in quantum error correction, and future directions for quantum, he provided a brilliant insight into the quantum industry. We also heard from both of them during an additional "Ask me Anything" panel in which the audience was able to ask questions that sparked astute and crucial dialogue we will dive into throughout this blog.
For the eighth year in a row, John Preskill has spoken at Q2B. This time, his talk touched on a few major subsections of the industry. He began by covering the theme of Q2B, the path to quantum value. The process of achieving quantum value lies in fault-tolerant quantum computing. Current Noisy Intermediate-Scale Quantum (NISQ) devices are necessary for breakthroughs but lack demonstrated commercial quantum advantage.
Along the journey from our current NISQ solutions to eventual Fault-tolerant Application-Scale Quantum (FASQ) solutions, a “megaquop” or “gigaquop” machine can perform about a million or a billion quantum operations. In the megaquop era error mitigation will continue to be useful. Machines will be beyond classical computers, NISQ, or analog, potentially with one hundred logical qubits and a depth of ten thousand, requiring tens of thousands of high-quality physical qubits. If this happens within the next few years, the modality will most likely be Rydberg atoms with optical tweezers. It is hard to know if this computer will have practical uses, what we will learn from creating it, and when it will be completed.
Quantum error correction remains essential to the success of practical quantum computing. Improvement in error syndrome measurement and decoding is critical to advances in error correction. We have seen promising advancements in logical error rates and resistance to ionizing radiation with Google’s Willow processor. The speed of real-time error decoding remains a bottleneck in quantum computing.
There have been advancements across modalities used in the industry. Superconducting qubits, cat qubits, and dual rail encodings have seen advancement throughout the year. These options for qubit representation offer varied approaches to improving error resilience. Ion traps are also promising, however scalability and clock speed remain challenges.
Innovations in hardware, software, and error mitigation will be necessary for future progress toward fault-tolerant quantum computing. Early applications will likely focus on quantum dynamics and material science rather than high-depth circuits similar to those needed in complex chemistry. Quantum technologies will advance more rapidly if higher education focuses on training students to tackle hard problems within quantum science. Future collaboration between industry and academia is essential for building a quantum workforce.
The talk begins with Prof. Aaronson explaining where we currently stand in the space of quantum algorithms. Bounded Error Quantum Polynomial Time (BQP) represents problems efficiently solvable by a quantum computer. We currently use Shor’s Algorithm which offers an exponential speedup for factoring and discrete logarithms. Also, Grover’s Algorithm provides not an exponential, but a square root speedup for unstructured search.
NP-complete problems, the hardest problems that are verifiable, likely will not see exponential speed-ups through quantum solutions. If a solution were to exist, it would not follow what we know from Shor’s or Grover’s algorithms. However, there are likely solutions to many NP problems, problems that can be checked in polynomial time, that will have direct benefits to the industry. It is important to stay between optimism and pessimism, we should explore new quantum algorithm paradigms while acknowledging their challenges.
Recently there has been promising work that shows exponential speedups in specialized algorithms. There has been success in certain graphs for which a quantum “walk” from one vertex can reach an end vertex exponentially faster than a classical walk. There was a theoretical breakthrough for the adiabatic algorithm, in a black box setting a quantum solution reaches the global optimum exponentially faster than any possible classical algorithm. Development for the Quantum Approximate Optimization Algorithm (QAOA) has shown promise for low-depth quantum devices, but there is no proof that QAOA beats a classical algorithm. However, recent empirical evidence has shown a polynomial speed-up in some optimization problems.
There are still challenges in identifying practical tasks where quantum algorithms outperform classical ones. Forrelation was an algorithm designed to show the potential advantage of quantum computing over classical alternatives. It is hard to see if the applications of these quantum algorithms are in any way beneficial to industries like machine learning. For instance, in 2016 a quantum algorithm was proposed to design recommendation systems, however, the algorithm was able to be turned into a classical algorithm with comparable performance. Many other machine learning speed-ups have been debunked in this way.
A breakthrough by Yamakawa and Zhandry was able to show exponential quantum speedup for a specific NP search problem based on random Boolean functions. This was essential for establishing the first NP search problem based on a random oracle that quantum allows for a speedup. Stephen Jordan and his collaborators were able to achieve a better approximation ratio for a real-world Optimal Polynomial Intersection (OPI) problem. This was a milestone achievement, but builds upon key parts of Shor’s algorithm, and does not yet solve a new unique problem.
At the end of Q2B24 Silicon Valley, the audience was able to ask both of these speakers a few questions in a panel moderated by our Chief Commercial Officer. They first spoke about the quantum computing landscape and promising new technologies. When looking at hardware, rydberg atoms and superconduction qubits are promising, but there is no clear winner, and we should be open to new technologies.
There is an argument between a traditional “Manhattan Project” style initiative, but this has its issues. Prof. Aaronson spoke about government funding, and would specifically like to see more funding for undirected and basic research. Although a billion dollars of funding was approved in 2018, all of it may get captured by large companies with a specific plan, smaller amounts for students with new ideas should receive more of it. This could lead to faster results and more breakthroughs due to less focus on a singular modality.
Impressive innovations in quantum error correction have been made, including Google’s use of surface code to protect the logical qubit for longer times as the code distance becomes larger. There has also been innovation on the theoretical side, we have found codes more efficient than surface code and in how we create simulations for physicists.
Quantum computing is not going to break Bitcoin or complete other groundbreaking practical work anytime soon. There is currently hype around quantum computing, leading to exaggerated media narratives. It is important to separate scientific progress from what you may see through media. The speakers are skeptical about finding significant speedups in AI problems beyond what we have today. However, quantum computing could help AI through areas such as error correction and optimization of quantum circuits.
Regarding the future, it is important to advise young students to pursue what excites them. Computer science, physics, mathematics, engineering, and more are all topics that apply to quantum computing. A broad education that fosters curiosity and problem-solving skills is far more valuable than preparing for specific roles. Quantum computing is still in its infancy and faces challenges in hardware scalability, error correction, and finding practical applications. However, it is a promising field and we should expect innovation and improvements in the years to come.
We look forward to covering future keynotes from Prof. Aaronson and Prof. Preskill
.svg){kind=small}
