Quantum computing is an evolving technology that promises to enhance an array of business operations. Based on quantum mechanics that focus on the smallest dimensions of nature — molecules, atoms and subatomic particles — quantum computers are set to provide faster solutions to complex business problems, through testing multiple possible solutions for a problem simultaneously.
The basis for quantum computing is a unit of information known as a ‘qubit’; unlike bits, which can only have the values zero or one, can come in the form of anything in between, which allows for this new approach to become possible, and is called a ‘superposition’. Combined, multiple qubits can produce many outcomes at the same time. Every extra qubit doubles the search space, which therefore grows exponentially.
Many companies are looking into how quantum can bolster industries and provide new use cases for businesses. One organisation that’s exploring this space is Reply, which has been developing solutions for optimisation in logistics, portfolio management and fault detection, among other areas.
Possible business value
Discussing how Reply is helping to provide possible use cases to its clients, quantum computing expert Johannes Oberreuter said: “We work on a level which translates the problem into a quantum language that is as universal as possible, and doesn’t go too deep into the hardware.
“The first thing we’ve found that’s delivering value now is the domain of optimisation problems. An example is the ‘travelling salesman problem’, which has lots of applications in logistics, where complexities and constraints also need to be accounted for, like during the pandemic.
“Very often, problems, which are found too complex to be optimised on common hardware, are tackled by some heuristics. Usually, there’s a team or a person with experience in the domain, who can help with this, but they don’t know yet that there are better solutions out there now. Quantum computing allows for problems being presented in a structured way similar to a wish list, containing all business complexities. They are all encoded into a so-called objective function, which can then be solved in a structured way.
“Companies have used all sorts of algorithms and brain power to try to solve optimisation problems. Finding the optimum with an objective function is still a difficult problem to solve, but here a quantum computer can come to the rescue.”
Pushing parameters
According to Oberreuter, once a quantum computer becomes involved in the problem solving process, the optimal solution can really be found, allowing businesses to find the best arrangements for the problem. While current quantum computers, which are suitable for this kind of problems, called quantum annealers now have over 5,000 qubits, many companies that enlist Reply’s services often find that problems they have require more than 16,000-20,000 variables, which calls for more progress to be made in the space.
“You can solve this by making approximations,” commented the Reply data scientist. “We’ve been writing a program that is determining an approximate solution of this objective function, and we have tested it beyond the usual number of qubits needed.
“The system is set up in a way that prevents running time from increasing exponentially, which results in a business-friendly running time of a couple of seconds. This reduces the quality of the solution, but we get a 10-15% better result than what business heuristics are typically providing.”
Through proofs-of-concepts, Reply has been able to help clients to overcome the challenge of a lack of expertise in quantum. By utilising and building up experience in the field, a “shoulder-to-shoulder” approach helps to clarify how solutions can be developed more efficiently.
Quantum machine learning
Machine learning has risen in prominence over the last few years to aid automation of business processes with data, and help organisations meet goals faster. However, machine learning projects can sometimes suffer from lack of data and computational expense. To combat this, Reply has been looking to the problem solving capabilities brought by quantum computing.
Oberreuter explained: “What we’ve discovered with quantum machine learning is you can find better solutions, even with the limited hardware that’s accessible currently. While there will probably never be an end-to-end quantum machine learning workflow, integration of quantum computing into the current machine learning workflow is useful.
“Some cloud vendors now offer quantum processing units (QPUs). In a deep learning setup for complex tasks, you could easily rent it from the cloud providers by individual calls to experiment, if it improves your current model.
“What we’ve found interesting from our contribution towards the quantum challenge undertaken by BMW and AWS, is the marriage of classical machine learning models with quantum models. The former is really good at extracting attributes from unstructured data such as images, which are then joined by a quantum representation which provides an advantage for classification.”
How organisations can drive value from AI on the edge
Quantum for cyber security
Additionally, quantum technologies are being explored for cyber security, with the view that soon quantum computers can solve problems that are currently insurmountable for today’s technologies. A particular algorithm that’s been cited by Reply, that could be solved by quantum computing, is the one used for RSA key cryptography, which while trusted to be secure now, is estimated to need 6000 error-free qubits to be cracked in the space of two weeks.
“Quantum technology for cyber security is now on the shelf, and we’re offering this to our clients to defend against this threat,” said Oberreuter. “Quantum mechanics have a so-called ‘no-cloning theorem’, which prevents users from copying messages sent across a communication channel. The crux is that in order for this to work, you need a specialised quantum channel.
“We have experts who specialise in cyber security, that have been leading the effort to craft an offering for this.”
Reply is a network of highly specialised industry companies, that helps clients across an array of sectors to optimise and integrate processes, applications and devices using the latest technologies. Established in 1996, the organisation offers services for capabilities including quantum, artificial intelligence (AI), big data, cloud and the Internet of Things (IoT). More information on the services that Reply provides can be found here.
This article was written as part of a paid-for content campaign with Reply