DARPA searched for fields quantum computers really could revolutionize, with mixed results
DARPA searched for fields quantum computers really could revolutionize, with mixed results
It's not all hype, but more work is needed before solutions are feasible or affordable
The US Defense Advanced Research Projects Agency has published the results of an exercise that assessed whether quantum computers will deliver on the promise of solving problems that stump classical machines – with mixed results.…
In 2021 DARPA created a Quantum Benchmarking program "with the goal of reinventing the metrics critical to measuring quantum computing progress and applying scientific rigor to often unsubstantiated claims about quantum computing's future promise."
That effort saw DARPA create eight interdisciplinary teams, which compiled "more than 200 potential applications from which they created 20 candidate benchmarks that could quantify progress in using quantum computers to solve hard computational tasks with economic utility."
A second phase of work saw benchmarks selected for detailed study in three broad categories: chemistry, materials science, and non-linear differential equations. Five teams refined those benchmarks using what DARPA last week described as "rigorous, utility-driven criteria, and then expand[ed] those benchmarks' applications, incorporate scalable and robust testing, evaluate real-world utility, and create tools for estimating resources and performance needed to run end-to-end instantiations of the applications on realistic quantum hardware."
The result of that effort is seven pre-press papers – all available here – that DARPA believes demonstrate it "is plausible that quantum computers will provide advantage for economically valuable applications in certain chemistry, quantum materials, and materials science applications."
But the news isn't all good. One of the papers is titled "Feasibility of accelerating incompressible computational fluid dynamics (CFD) simulations with fault-tolerant quantum computers" and finds that, while quantum systems have plenty of promise, "future quantum computers are unlikely to provide utility for incompressible CFD applications unless significant algorithmic advancements or alternative quantum approaches are developed."
A similar conclusion is reached in another of the papers, catchily titled "Quantifying fault tolerant simulation of strongly correlated systems using the Fermi-Hubbard model" – an unsolved approach to assessing the properties of materials thought to be a workload ideally suited to quantum computers.
One approach to the problem uses a technique called "time evolution circuits" and would require "a calculation that takes over three years and uses 400 logical qubits."
Given that the hardware required for even that pace of computation "is at the extreme end of today's classical hardware, this points to a need for improvement along a variety of axes."
A paper on nuclear magnetic resonance spectroscopy is more positive, suggesting quantum computers could model phenomena that are currently hard to analyze with classical machines.
Another of the papers, "Applications and resource estimates for open system simulation on a quantum computer," found that quantum computers could lead to savings of $2 million in the cost of materials needed for each test at the Los Alamos National Laboratory's High Magnetic Field Laboratory. That lab is also mentioned in another of the papers – titled "Potential Applications of Quantum Computing at Los Alamos National Laboratory" – in which another six fields of research suitable for quantum computers are described.
One of the papers considers quantum computers themselves. Titled "Fault-tolerant resource estimation using graph-state compilation on a modular superconducting architecture," the document considers how to build quantum computers to tackle scientific tasks.
The document's authors conclude that "a fault-tolerant quantum computer based on a distributed superconducting architecture, consisting of two modules could, in principle, house 2,000,000 physical qubits and could serve scientifically interesting applications in a reasonable run-time."
None of the documents express certainty that quantum machines will revolutionize the fields they consider. But they also assume fault-tolerant machines will work on the problems considered – and such machines are nascent.
Three more pre-press papers are in the works:
Quantum enthusiasts may take heart from knowing that the papers are yet to be peer-reviewed – meaning they could be wrong, and a land of quantum unicorns and rainbows awaits us all mere weeks after AI solves all the problems quantum computers were going to address. ®