To understand the enormous potential quantum computing holds for artificial intelligence (AI), it’s first helpful to understand – in a very general way – how quantum computing works.
Unlike classical computing, which stores information in “bits” — essentially zeroes and ones — quantum computing makers use of “qubits,” which can store two values at once. A qubit can be both a zero and a one at the same time.
This characteristic of quantum physics – alongside a phenomenon known as quantum entanglement – makes it possible for a set of qubits to store vastly more information. Instead of needing 1012 bits to store a terabyte of data, you’d need just 40 qubits.
At the same time, quantum computing will process this information in a uniquely different way. Conventional computing is linear: One action or process follows on another. Quantum computing, on the other hand, can perform many computations simultaneously, providing a “quantum speedup.”
It’s easy then to see then how quantum computing can be a powerful accelerator for AI in two ways – increasing storage capabilities for input data, and more efficiently analyzing that data to see patterns.
Many in federal government may be most familiar with quantum computing’s cybersecurity implications. One day, quantum computing will be able to break the encryptions that have traditionally kept data safe. This is a vital issue, which will require significant preparation.
Yet, quantum computing will also create immense opportunities.
<<< Start >>>
The data informing federal missions is growing exponentially; our ability to process that data needs to grow alongside it.
<<< End >>>
Quantum computing can supercharge AI, helping federal agencies better make sense of increasingly large and complex data sets. This can lead to more efficient processes and more robust insights for meeting mission needs.
Furthermore, quantum-powered AI isn’t theoretical – though still evolving, services are available today through major cloud service providers.
What quantum-based AI can do for federal
An important clarification is that quantum computing will not now – or likely ever – fully replace classical computing. Many complex, large issues can and still will be solved by classical computers.
The future of the technology stack will be a hybrid of classical and quantum computing. The problems best fit for quantum computing fall in three categories:
- Sampling problems
- Machine learning
Optimization is well-suited for a quantum speedup. The “Traveling Salesman Problem” is a classic scenario that demonstrates the inherent complexity of optimizing at scale. The problem asks: If a traveling salesperson wanted to visit a certain number of cities, what would be the shortest route that visits each city exactly once and returns the salesperson back home? For even 10 cities, there could be hundreds of thousands of possible routes. And this problem is only considering one factor – distance between cities.
As optimization problems grow in complexity, with many factors and inputs, classical computing will eventually fail to deliver a satisfactory answer in a reasonable length of time. For federal agencies with complex and large-scale logistics, quantum can enable a new level of insights and capabilities. Quantum’s ability to compute simultaneously means it can more efficiently tackle these types of problems.
A quantum approach could likewise apply to sampling problems, a statistical means of gathering insights on an entire population based on a subset or sample. This kind of statistical work depends on the ability to randomize data, and quantum is by its nature random. The fundamental randomness of quantum processes makes it extremely difficult to introduce any kind of bias, meaning sampling and simulation in a quantum environment will likely be far more random than when tackled via classical computing.
<<< Start >>>
<<< End >>>
Since machine learning is based on both sampling and optimization, improvements in these techniques will in turn drive better algorithms, or the engines of machine learning that turn data sets into models. Sampling in quantum can provide more distributed, reliable input data for the algorithms, while quantum optimization could be used to improve the algorithms over time.
Where to begin
Federal agencies can take steps today to prepare for quantum-powered AI:
- Learn what quantum can do and decide where it applies to your business.
Now is the time to start thinking about which processes will be best served by a quantum approach. Classical computing will continue to support many basic algorithms. In support of a hybrid future, agencies can start now to identify the bigger problems – the oversized and esoteric use cases – that will benefit from the power of quantum computing.
- Build a quantum innovation roadmap.
Quantum isn’t at its end state yet: Today’s processes still generate a lot of noise and require correction to make their outputs useful. It will take at least two or three more years before we see more fault-tolerant quantum processes become more widely available. It’s worth diving in now, though, to get ahead of the curve and take advantage of the capabilities already on the table. An innovation roadmap can help chart that course.
- Evaluate quantum hardware and software and start experimenting.
Agencies can start to put those quantum techniques to work today with as-a-service cloud offerings such as Azure Quantum, Amazon Braket, and IBM Quantum. By running experiments on the cloud vendors’ platforms, they can gain confidence and better understand which use cases can see the most value from quantum-powered AI today.
Quantum is no longer a hypothetical opportunity. Quantum-powered AI is already robust enough to start creating new efficiencies and insights in areas from cybersecurity to supply chain to finance and more.
Any time you’re looking at patterns or systems that cannot be quickly understood by humans, or through classical computing — when you’re tackling bigger, weirder problems, or looking for a smaller needle in the haystack — quantum-powered AI may be able to offer a powerful path forward.