kentoh - Fotolia
During its Build 2020 annual software developers’ conference, Microsoft unveiled how it sees quantum computing fitting into its Azure public cloud. The company launched Azure Quantum, which it said would provide early adopters with a scalable path to quantum computing.
The idea is that organisations can begin to build so-called “quantum-inspired” algorithms today, which allow them to start to gain the benefits of quantum computers without needing to use them directly.
A Quantum Developer Kit (QDK) and new language Q# fill out the Microsoft quantum computing portfolio and are available on the open source GitHub repository.
As Computer Weekly has previously reported, quantum computing is a technique that promises to solve problems that cannot be programmed using a traditional algorithm run on a classical binary computer design.
Whereas traditional or classical computer architectures are very good at dealing in binary decisions, and solve problems by making discrete “yes” and “no” decisions, the complexity of some problems rises exponentially. This effectively means the problem cannot be solved in a traditional way.
Giving an update to the company’s strategy, Ben Porter, director of business development at Microsoft, said: “Having spoken to customers across every industry, there is a need to study algorithms to solve complex problems.”
But developing novel quantum algorithms is just the first part of Microsoft’s strategy. The company aims to build out an open ecosystem to solve problems that cannot be run on classical computers. It aims to provide pre-built problem solves and algorithms that can run at an industrial scale.
Describing a traffic optimisation problem that Jij developed for Toyota Tsusho, Porter said: “If you can optimise the timing of traffic lights, you can not only reduce the idling time of vehicles, but also improve the driving experience and reduce emissions.”
He said Jij mapped waiting times to waiting cost, enabling its programmers at to express the problem as a type of optimisation called polynomial unconstrained binary optimisation (Pubo).
“This is a type of problem where each variable can take one of two values,” said Porter. “The goal of the optimiser is to find some combination of variables that minimises the cost.” In traffic simulation, each variable can interact with many other variables, which increases the complexity, he added.
“It’s the hardest class of problems out there,” said Porter. “We have Azure Quantum-inspired optimisers that are specially designed to tackle these Pubos, which Jij has used to great effect.”
According to Microsoft, this allowed Jij to achieve a 20% reduction in waiting times compared with conventional optimisation techniques.
Better Oled displays
Another example is OTI Lumionics, which has developed a fast materials design approach tailored to making Oled, based on machine learning, computational chemistry simulations, optimisation, closed-loop synthesis and rapid feedback. Instead of synthesising and testing thousands of materials in the laboratory, OTI developed software tools to simulate the properties of materials.
According to Porter, this means the materials are designed rather than developed by chance. The slowest and most expensive part of the workflow is the computational pipeline – the bottleneck on available hardware when running extremely large simulations, which scale exponentially with size. Also, some simulations are so compute-intensive that they are literally unsolvable with today’s classical computers.
So the trade-off between simulation accuracy and compute-intensity is a major bottleneck in using a computational approach for commercial-size problems.
To beat this bottleneck, OTI Lumionics has been investigating quantum computing as a potential candidate to help accelerate computational chemistry simulations of new materials. Because many structure-property relationships of materials are governed by quantum physics, quantum computing, which uses quantum mechanical effects to perform computations, is a natural candidate to simulate these systems more accurately.
Read more about quantum computing
However, to simulate one molecular model requires 42 QuBits, something that cannot be produced with the degree of accuracy required for the simulation, according to OTI.
Scott Genin, head of materials discovery at OTI Lumionics, said: “Quantum computing has the potential to revolutionise materials design by enabling highly accurate simulations that could otherwise not be solved on classical hardware. Unfortunately, current gate-based quantum computing is far from being powerful enough to simulate commercial-sized problems.
“We have developed new methods that allow quantum computing algorithms for computational chemistry simulations to be represented as binary optimisation problems. Running our quantum computing methods with Azure Quantum optimisation solutions, we are getting results that are more accurate than other algorithms.”
Instead, the company has been able to use quantum-inspired algorithms running on classical Azure hardware.
With its algorithms now running on Azure Quantum, OTI Lumionics said it has been able to demonstrate meaningful results on commercially relevant sized problems.