The Emergence of Quantum Computing

When analytics and optimization problems—such as routing hundreds of autonomous cars on busy city streets or creating a personalized drug regime to combat a rare form of cancer—grow in size and complexity, we need more computer power to solve them. Despite increases in computer performance, and the availability of scalable solutions like cloud computing, novel approaches to problem solving are increasingly vital. In analytics specifically, the size and detail of datasets continue to swell.

Today, researchers estimate approximately 30 to 50 billion devices will be connected to the internet by 2020 [1]. These devices contribute to the rapid growth of the digital universe, which is the aggregate of all digital data. Researchers also estimate that the size of the data universe will grow from the 2013 level of 4.4 ZB (or 1 trillion gigabytes) to 44 ZB by 2020 [2]. Once datasets become this massive, it’s especially valuable to quickly analyze data. 

New technologies, such as scalable cloud computing and GPU processing [3], are common methods of attacking these growing analytics and optimization problems. Despite the appearance of new technologies, there is growing concern that Moore's Law [4]—which states that processor speeds, or overall processing power, doubles every 2 years as a result of increasing the number of transistors on a silicon microchip—will soon end, ushering in an era of decreased gains in computer performance. These decreases will only magnify computers' inability to tackle the growing size and complexity of problems mirrored in the world—from combatting cyber attacks in real time to optimizing plane routes to decrease fuel usage and passenger counts simultaneously. This concern is the chief motivator in the development of new computing paradigms, such as quantum annealing. 

What's Quantum Annealing?

Think of it this way: While your smartphone contains more computing power than top-of-the-line machines that took up an entire room 50 years ago, it’s still a basic calculator. Conventional computers can easily crunch and process numbers while following pre-arranged programming instructions. Quantum computing is a physics-based approach that considers the behavior of atoms and their deviations from the laws of physics we can observe. Like all quantum computing techniques, quantum annealing takes advantage of the strange properties of quantum mechanics, like quantum tunneling, superposition, and entanglement to perform computations in a very different way. Programmers theorized quantum annealing about around 2 decades ago, but only recently implemented it on real, special-purpose hardware, making some models available for commercial use via specialty clouds or enterprise-level partnerships.

The basic unit of this quantum hardware is called a quantum bit—or qubit, for short—made of a very small, and very cold, loops of superconducting wire. But unlike a classical bit, which sends information in binary via a string of 0’s and 1’s, a qubit can exist in a superposition of 0 and 1 at the same time. 

Confused yet? 

Picture flipping a coin to make a decision. In a classical computer, the “coin” only exists when it is either heads or tails. But when the coin is spinning in mid-air, it could be either heads or tails—so it is essentially both at the same time. Schrodinger’s famous thought experiment about a cat being both alive and dead as long as we don’t open the lid of the box to check is built on the same principle. Quantum computers can “hold” the coin in its spinning state—when it is in a superposition of heads and tails. Qubits can therefore perform calculations with many more variables than just a 0 or a 1, allowing for the calculations to be solved in a completely new way. For certain problems, such as optimization, the calculation takes fewer steps. 

For example, trying to arrange forty coins into five smaller groups on a table has 75,740,900,000,000,000,000,000,000 possible arrangements. A classical computer can use approximation methods to find “good enough,” but not likely optimal, solutions. But sometimes a near-perfect answer is needed. That is where quantum annealers come in. By leveraging superposition, they can run through possible arrangements in fewer steps, and therefore are potentially faster than a classical computer. If done correctly, this solution could present an optimal result, although you may need to repeat the process multiple times. Fortunately, computers can complete this in a matter of milliseconds, allowing many repetitions in a very short time. 

Quantum Annealing at Booz Allen

At Booz Allen, our quantum computing team is exploring a combination of quantum annealing and traditional computing to solve optimization problems, a strategy called heterogeneous computing. While promising, even the latest quantum computing hardware face many challenges in error-correction or long-term computer memory. Preliminary studies, including this academic paper written by members of our team, shows that our strategy of combining current, state-of-the-art quantum computers with classical computers produces results equal or better than either individually. The Booz Allen team is also exploring how to use quantum annealers to enhance machine learning algorithms and is engaged in preliminary work with multiple use cases.