Yuval Boger is the Chief Commercial Officer of QuEra Computing, a leader in neutral-atom quantum computers.
Quantum computing appears on track to help companies in three main areas: optimization, simulation and machine learning. The appeal of quantum machine learning lies in its potential to tackle problems that classical ML might struggle with. Quantum algorithms could help handle complex data relationships and large-scale data, but these benefits are still mostly theoretical and depend on future improvements in quantum hardware.
Identifying Areas For Quantum Advantage
Companies exploring QML today are primarily motivated by two reasons:
• Identifying areas for quantum advantage: Organizations are experimenting with QML to pinpoint cases where quantum approaches could yield superior results compared to classical ML. Classification tasks (such as fraud detection and image recognition) and predictive analytics (such as forecasting energy usage and housing prices) are high on the list.
For instance, HSBC is investigating QML to enhance fraud detection and risk management, aiming to improve the accuracy and efficiency of its financial services.
• Readiness for the future: With quantum computing advancements on the horizon, companies want to establish familiarity with QML now to avoid being left behind. Adopting early gives them a head start in understanding which workflows could benefit from quantum enhancement as the technology matures. This is similar to how early AI adopters gained a competitive advantage, understanding the best use cases and integrating them into their business processes before the competition.
How Companies Approach QML
For businesses venturing into QML, it’s common to start small, often running experiments on simulators or hybrid quantum-classical platforms before committing to larger, resource-intensive quantum jobs. These initial explorations allow them to identify feasible use cases without incurring steep costs.
Goldman Sachs, for example, has been using hybrid quantum-classical platforms to explore applications related to pricing financial instruments.
Use Cases
Quantum ML is particularly promising for two types of problems: classification and prediction. Airbus, for example, is researching QML to address complex optimization problems in aircraft design and operations, including improving materials science and aerodynamic modeling. (Full disclosure: Our company previously worked on a project with Airbus.)
Classification tasks, such as fraud detection and image recognition, are essential across industries. For instance, financial companies are always looking for ways to improve the speed and accuracy of their fraud detection systems. Quantum algorithms process data in new ways that can enhance classification accuracy by capturing complex correlations within data that classical models miss.
Predictive analytics is another area where QML could make a significant impact. For example, energy companies need precise forecasts of energy consumption patterns to optimize grid management, while real estate firms rely on accurate home price predictions to make informed investment decisions. Quantum-enhanced predictive models could enable these companies to achieve higher forecasting accuracy by identifying subtler patterns and interactions within their data. These are the types of tasks that businesses frequently encounter and often struggle to optimize with classical techniques alone.
For instance, our company has partnered with Moody’s to use QML to predict the severity of tropical cyclones.
Integrating Quantum And Classical
QML projects often use classical computing as an integral part of the workflow, particularly in the early stages. Here are some ways classical computing complements QML experiments:
• Data preprocessing: Classical systems are often employed to prepare and filter data so that only the most relevant information is passed on to the quantum system. This preprocessing can reduce the number of qubits needed for calculations, making QML experiments more manageable.
• Data compression: In some instances, classical ML techniques are used to compress large data sets before they’re fed into a quantum algorithm. This approach helps maximize the utility of a limited number of qubits, especially given that quantum hardware is currently constrained in scale. For instance, Honda Research Institute uses such data compression for image analysis using quantum computers.
• Training: Classical ML can also help in the training phase of quantum models, such as in quantum reservoir computing. By supporting this process, classical ML contributes to making quantum workloads more efficient.
Quantum ML And Data Set Expansion
Another innovative way companies are using QML is to expand data sets or to find correlations in small data sets. For instance, Amgen works to use QML with clinical trial data. Building classical machine learning models with small data sets from early-phase trials is challenging, but QML could address these issues.
The Ideal QML Team
One of the key insights from companies with successful QML initiatives is the need for cross-functional expertise. A QML team should ideally combine quantum computing experts with business domain specialists. This interdisciplinary approach allows for better alignment between quantum capabilities and business goals. That way, QML projects are focused on solutions that deliver tangible value.
For example, in the pharmaceutical industry, a QML team would benefit from combining quantum physicists with chemists or pharmacologists. Similarly, a finance-focused team should blend quantum expertise with analysts who understand market dynamics and risk modeling.
Starting QML Exploration: The Road Map
Here are a few steps companies can take to get started with QML:
• Identify high-impact use cases. Start with applications where QML shows the most promise, such as classification and prediction. Even if full quantum advantage is years away, focusing on high-impact areas allows you to reap incremental benefits.
• Leverage hybrid workflows. Use classical-quantum hybrid approaches to overcome hardware limitations. Classical computing can handle data-heavy preprocessing tasks, which are crucial for successful quantum applications.
• Engage in simulation-based exploration. Begin with simulators to test initial concepts before investing in real quantum hardware. Simulators allow your company to experiment at a lower cost and build a foundation in quantum methods.
• Build a multidisciplinary team. Ensure quantum experts collaborate with domain specialists to align technical efforts with business needs. This will ultimately drive more relevant and impactful outcomes.
QML offers exciting use cases. Companies that invest early will be well-positioned to capitalize on its potential as the technology evolves.
Forbes Business Development Council is an invitation-only community for sales and biz dev executives. Do I qualify?
link