Could quantum computers solve the Big Data crunch?

Editor's note: This story was originally posted 21 Jan 2014, but it's now graduated and become part of Astronomy 101!  Please be aware that any references to events that seem current may not actually reflect events happening right now.

No, this isn't science fiction.  Quantum computers are real and, for certain tasks, they are fast.  Really fast.

Last year, NASA teamed up with Google and the Universities Space Research Association to buy one of the first commercially-available quantum computers.  Manufactured by D-Wave, the system has "just" 512 quantum bits (qubits, more on this later!), yet it's already competitive with the best traditional computers in some cases.  Google engineers have been speed testing the new machine against the best traditional computers available today, and the results are pretty impressive.  Not only is the quantum computer as fast or faster for many specialized tasks than a normal machine, adding just a few more qubits to the existing 512 will mean orders of magnitude faster performance.

 The first generation of commercial quantum computers are already demonstrating the promise of this new technology.

The first generation of commercial quantum computers are already demonstrating the promise of this new technology.

What is a quantum computer, anyway?  Described most simply (and, trust me, there is nothing simple about quantum computing), a quantum computer is a machine which leverages the principles of quantum mechanics - a fundamental branch of physics which describes the behavior of the tiniest building blocks of nature.  You might be familiar with quantum mechanic's most famous theorem: the Heisenberg Uncertainty Principle, which states that we can never know a particle's exact position and exact momentum simultaneously.  But, quantum mechanics is overflowing with additional laws that are remarkable for both their wide application and their obtuse nature.

Quantum computing makes use of these physical laws to find the result of mathematical operations.  In traditional computing, circuits manipulate "bits," which can either have a value of 1 or a value of 0.  Quantum bits, or "qubits," can have a value of 1, a value of 0, or a value of both 1 and 0 at the same time. More than just being cool, this idea (called "superposition") is at the heart of quantum computers' power.  To make a quantum calculation, the machine sets the qubits into an initial state representative of all the input information.  Because qubits don't have to represent only one value, the machine can simultaneously consider all possible combinations of information.  The problem you want to solve is described by a series of physical constraints that the system must obey.  When the qubits are allowed to change in response to these constraints, they settle into values of 1 or 0 which represent the answer to the problem (more correctly, the most likely answer to the problem - quantum mechanics is all probability!).

What I've described here is not a quantum computer which can solve any problem - those are still in the development phase.  The computer described here, like the one now owned by NASA, can be described as an optimizer.  It can't run your web browser, but it can quickly find the most efficient outcome of a situation.  It's easy to see why a company like Google might want something like this.  Every time you ask Google Maps for directions (a lot, if you're me!), you're asking Google to optimize the best route for you to take.  Just making a web search is equivalent to asking the search engine to take the initial conditions (all the information on the Internet) and your constraints (the search terms) and compute the best outcome (the first result!).

But what could NASA and the Universities Space Research Association want with such a device?  Astronomy in general is on the precipice of an unprecedented influx of scientific data.  Upcoming projects like the James Webb Space Telescope and the Large Synoptic Survey Telescope (LSST) will generate far more data than previous instruments.  LSST, for example, is designed to photograph the entire southern night sky every three days.  This will result in the equivalent of about a billion images per year from LSST alone.  Since it began operation, the Sloan Digital Sky Survey has cataloged more than 900 million additional objects.

How to deal with this torrent of data will be the next big challenge in astronomy, and quantum computers are poised to help in a number of ways.  For one, they could efficiently compare different models against these vast data sets and select the ones which best match our observations.  More interesting in the long run, though, might be a quantum computer's ability to learn.

Learning has long been one of the greatest challenges in modern computer science.  Classical (non-quantum) computers more or less follow a specific set of instructions when performing a task.  For example, if I wanted to classify spiral galaxies in my image, I might give the computer a pattern and tell it to select features which match that pattern to some tolerance.  Or, as Sloan does, I could have people classify my galaxies for me.  With a quantum computer, you could show the system a bunch of images of spiral galaxies and say "find things which look similar to this."  You literally teach or train the program what a spiral galaxy looks like.  Such algorithms (for example, neural networks) exist today and are used by things like Siri, but this is a far more natural concept in quantum computing than it is in the traditional computer world.

The bottom line here is that, while quantum computing is in its infancy, it will probably be sooner than we think that this remarkable new technology is helping make scientific gains.

Oh, and if you're looking to buy one for your home, I wouldn't hold my breath.  Quantum computers must be cooled to within a hair of absolute zero - not the sort of device you'd want your cat to knock over!