You've heard plenty of people by now—including us—banging on about quantum computers, and how they’re the future of high-performance computing. Quantum computing, we're meant to understand, is set to change the world. But despite its promise, it's neither widely available nor particularly useful yet. Here's why not.
First things first, a quick reminder of exactly what a quantum computer is. Essentially, they’re computers that make use of quantum physics. Unlike a normal digital systems which rely on data encoded into binary digits (bits)—that can only ever take the form of 0 or 1—quantum computation uses quantum properties to represent data and perform operations.
A quantum computer, then, uses not bits but qubits (quantum bits). Each of its qubits can represent a 0, a 1, or—crucially—anything in between.
Imagine a table covered in coins. In a classical computer, each one is either showing a head or tail; in a quantum computer, a coin could be showing 25 percent head and 75 percent tail, or any other position between the two possible states. Once measured, of course—and we have Schroedinger to thank for this—it assumes one of the two states, heads or tails.
Because each qubit can assume such a wide range of values, a modest number of them can hold an insane quantity of information; this is what lends quantum computers their theoretical grunt. Just 100 qubits can store 1,267,650,600,228,229,401,496,703,205,375 different numbers—many trillion times the storage capacity of all computers ever made. In other words, 100 qubits can simultaneously represent all possible 100-bit numbers in their huge quantum state, as opposed to a classical 100-bit computer, which can represent just one.
It’s that vast ability to assume many states at once that—in theory—means that quantum computers can provide untold power, many times faster than any classical computer. In practice, it's rather more difficult.
First off, building a quantum computer is no mean feat. While the technology required to create computers full of traditional bits is well established, manufacturing qubits is far from straightforward.
For starters, no one is quite sure what the best way to make one is. Some techniques involve trapping ions, electrons or other tiny little particles; some propose using superconductors to create microscopic quantum circuits; others suggest it might be possible to use photons and complex optical apparatus to achieve a similar goal. What these techniques all have in common, however, is the fact that they’re currently plausible on the small scale but incredibly difficult to realize on the large. Essentially, that limits quantum computers to research machines, at least for now.
The scaling problem itself is a result of quantum decoherence—or, rather, wanting to eliminate it. The problem is that, as good ol’ Schroedinger was only too keen to point out, quantum systems need to be isolated from the rest of world in order to work. Interactions with the external world cause the system to decohere, collapsing down and taking a binary state, just like a normal computer.
But the very guts of a quantum computer—quantum gates, lattice vibrations, background nuclear spin of the qubits—can all introduce decoherence effects, too. The solution? Decide on an error rate—the amount of dechorence you’re happy for the system to put up with—and design for that.
Even that's an imperfect solution, though; to have an error rate small enough that you're still getting the benefits of a respectable quantum computer, you’d need a weighty bump in the number of qubits to provider error correction, and those qubits are extremely difficult to produce in the first place, which… well, you can see where that goes.
Let's say, though, that you're able to build a functional quantum computer (not surprisingly, Google's on it). While you might think that having all that raw, pulsating computational power at your fingertips might be a dream come true, in reality using the damn thing is close to nightmarish.
First, there’s the question of knowing if it’s even working in the first place. A widely known tenet of quantum mechanics is that merely observing the phenomenon changes the outcome of an event. So, watch a quantum particle, or a qubit, or anything quantum for that matter, and you change its behavior. That means that it’s actually very difficult to tell if a quantum computer is behaving in the way we’d expect or need it to.
In fact, the currently available so-called quantum computers aren’t actually verified to be working the way they're supposed to. They’re simply based on the right theory, some fingers crossing, and judged by their output.
That raises a whole other set of issues. Specifically, actually attaining that output is a mess. Coding a quantum computer is no mean feat; by their very nature, they give answers that are necessarily probabilistic, not concrete. For many solutions, that means that the answer isn’t necessarily bang on at first attempt; instead, the same calculation has to be repeated a number of times before the obvious correct answer emerges. In turn this means that, depending on the type of problem, there isn’t necessarily a huge amount of advantage in using a quantum computer compared to a regular one.
It's possible to exploit some of the mystical magical power of quantum mechanics to improve the speed with which solutions are reached, but so far researchers have only managed to do it for a very small set of problems, like finding the prime factors of very large numbers. That’s neat—and, it turns out, useful for cryptography—but it’s certainly limited.
Here’s the real kicker, though. Despite all that computational grunt, all the blood, sweat, and tears building and coding the damn thing, it’s not even necessarily possible to tell if a quantum computer produces the right answer in the end. Say what?
A quantum computer could, in theory, be used to calculate solutions in days, maybe even hours, that would take a normal computer thousands of years to produce. While some of the answers its spits out are verifiable—a complex cryptographic key generated by a quantum computer is testable by using it and checking it by encrypting and decrypting a message, for instance—there are others that can’t be tested. Simply put, quantum computers are often used to solve the types of problems for which we have no other confirmation mechanism. We often can't double check their work.
There may, however, be a way around this. A team of scientists from the University of Vienna have developed a technique call “blind quantum computing” that it believes could help. The idea’s quite straightforward; it involves mathematical traps, essentially intermediate steps in a calculation, which humans—or at least other computers—definitely can work out in advance. If those answers come out incorrect, then the overall answer must contain an error, too. Instead of checking the final solution, it keeps an eye out for problem spots along the way.
In Nature Physics, the researchers have demonstrated that the technique can at least work on a small scale, using a four-qubit quantum computer as a verifier to test against a second quantum computer. They even claims that it’s scalable, and could be used on computers with hundreds of qubits. But, like all things in quantum computing, there’s a snag. "Like almost all current quantum computing experiments, this currently has the status of a fun demonstration proof of concept, rather than anything that's directly useful yet,” explained computer scientist Scott Aaronson at the Massachusetts Institute of Technology, to Science.
And that, right there, demonstrates why quantum computing remains little more than a lusty dream. At least for now.