Will quantum computers ever see the light of day, or are they the ultimate vapourware?

Scan back through yellowing copies of any computer magazine and you’ll find repeated reference to quantum computers being ‘the next big thing’ for the past decade or more. So why are there no quantum computers on the shelves at PC World yet?

Quantum computers do exist, but you’d have to visit the laboratories of corporations such as Google or IBM – not your local high street – to see them. If you did, what you’d encounter would very different from the rectangular box that hums quietly beneath your desk.

“Quantum computers are an entirely new technology which has really very little to do with conventional computers,” says Winfried Hensinger, professor of quantum technologies at the University of Sussex. “Quantum computers are going to be used for things where you cannot solve the problem any other way. So even the fastest supercomputer in the world may take millions or billions or years to calculate something [that] a quantum computer can calculate.”

So, putting aside the misleading headlines about quantum computers taking over the world, what is the real future of the technology?

The big difference

Explaining how quantum computers work is difficult, because it’s so counterintuitive to everything we know about computing – or physics itself. Conventional computers use bits that can possess one of two values, namely one or zero. The computer writes strings of these ones and zeros into memory, processes them in sequence, and outputs an answer. Crucially, a conventional computer can only make those computations one-at-a-time, modern computers just do it incredibly fast.

READ NEXT: What is quantum computing?

A quantum computer does the same thing – with one key difference. Rather than bits, a quantum computer uses quantum bits, or qubits, as its basic unit of information. Unlike a standard bit, a quantum bit can be both one and zero.

“In quantum physics… an atom can be in two different places at the same time,” Hensinger explains. “Many years ago, I did an experiment where I made an atom move forward and backward simultaneously. So, imagine that you’re sitting in your car, and as you hit the car in front of you as you get out of the parking lot, you also hit a car behind you. Both at the same time.”

(Prof Winfried Hensinger, Credit: University of Sussex)

This effect is known as quantum superposition. At the level of a single qubit, it’s not very useful. If you add another qubit, however, things become interesting. “When you have two quantum bits you can simultaneously program zero-zero, zero-one, one-zero, one-one,” Hensinger says. “You can store all these possible combinations into these two quantum bits, and now, when these two quantum bits go into the processor, the quantum processor does all the calculation simultaneously.”

Consequently, a quantum computer’s potential processing power increases exponentially with the number of qubits added. The threshold where quantum computers can outperform regular computers at certain tasks – a point known as the quantum supremacy – is around 50 qubits. At that point, a quantum computer would be capable of performing over a quadrillion computations simultaneously.

Reaching the milestone

That sounds impressive, but getting to the 50-qubit mark is tough – one of the reasons why we’re still talking about quantum computers in labs rather than server stacks. Quantum states are extremely fragile. Any interactions with the environment can easily destroy a quantum state. Simply observing a qubit in action can force it to occupy a single state – destroying the superposition. It’s like working with a socially averse, genius mathematician who can only work if left alone. If she so much as senses someone else in the room, she’ll physically collapse and be unable to function.

Avoiding this problem requires intense and ingenious engineering. Currently, there are several different approaches to building a quantum computer. The first of these is superconducting quantum computing, which is the method employed by Google, IBM and Intel. This basically involves cooling the microchip that carries the circuitry right down to the edge of absolute zero (-273.15°C). This enables current to flow with almost no resistance, the ideal environment for quantum states.

The second method is known as ion-trap quantum computing, which is the area Hensinger specialises in. This method also involves cooling the qubits to absolute zero, but instead it’s performed by shining two lasers at the atoms which form the qubits and tuning the frequency so the lasers cool the atom directly. It also has another effect. “There’s something which is more exciting about trapped ions, which is the ion actually levitates,” says Hesigner. “They’re not coupled to anything. And this is why it’s so easy to manipulate trapped ions to make quantum gates.”

(A trapped ion quantum computer uses two lasers to cool the qubits to almost absolute zero)

Both approaches are perfectly viable for building small-scale quantum computers. The current record is around 17 qubits, and both Google and IBM are working on 50-qubit superconducting quantum computers (in fact, IBM recently tested a prototype of a 50-qubit quantum processor). But hitting quantum supremacy is only the first step. To make a quantum computer that can do anything practical requires many more qubits, running into the billions depending on the problem you’re trying to solve. For these kinds of machines, the current architectures are simply not practical. Such machines would be mind-bogglingly huge and enormously expensive to build and power.

The D-Wave alternative

There is another approach to quantum computing. You may have heard of a company called D-Wave, which has constructed quantum computers with a far greater number of qubits than anyone else in the field, most recently claiming to have built a machine of more than 2,000 qubits. But D-Wave’s machines are very different beasts to those being built by Google and IBM. D-Wave’s computers are what’s known as quantum annealers, and they operate via a method known as adiabatic quantum computing.

The key difference between quantum annealing and what’s known as “gate-model” quantum computing is that it harnesses the natural evolution of quantum states. “We usually describe it as doing computation in a low-energy state of an interacting quantum system,” says Dr Elizabeth Crosson, an expert in adiabatic quantum computing at the Caltech Institute for Quantum Information and Matter. “These qubits… their interactions determine their energies and the idea of adiabatic computing is to be in the lowest state.”

This means quantum annealers can scale up much more quickly because the environment the qubits exist in is only changed very gradually. Yet although they can be as powerful as other quantum computers, the qubits in their natural state encounter a lot of “noise”, making it hard to do meaningful computation on them. “In terms of the performances of the D-Wave machine, it’s competitive with modern CPUs with its 2,000 qubits,” Crosson says.

READ NEXT: Quantum computing comes of age

This doesn’t mean that quantum annealers should be disregarded. At the very least they’re a good proof-of-concept, and they’re useful for solving specific types of problems. It’s possible they’ll achieve quantum supremacy too, albeit with many more qubits than a universal quantum computer. What you probably won’t see with quantum annealers is the massively exponential increase in computational power that universal quantum computers will provide.

“People often say that the adiabatic model is actually an analogue computer,” Crosson explains. “So even though you have bits as your data-type, because you’re smoothly changing the interactions in the system and it’s smoothly changing your low-energy states, that actually makes it a form of analogue computing. If you look at the history of classical computers, in the early stages, analogue computers were very important and useful. And then in the long term they got outpaced by digital computers.”

What will run on them?

The hardware is reaching a point where it’s comparable in performance to current computers. But the hardware is only half of the story. Like any computer, quantum computers require software to function. As with the hardware, quantum software is very different, both in how it helps the quantum computer work and its practical applications.

A curious problem with using a quantum computer is that because the computer performs many computations simultaneously, it also delivers multiple answers. The user, of course, normally only wants one answer. But if you try to read-out that answer in the wrong way, the quantum state will break down and the computer will spit out a random answer. “If you want to use it for computation you need to add another ingredient to the mix,” says Harry Buhrman, professor of computer science at the University of Amsterdam and director of the research centre for quantum software, QuSoft.

(Harry Buhrman. Credit: University of Amsterdam)

This ingredient is known as interference, a phenomenon whereby particles act like waves. “In the swimming pool you have waves, and then if you make two on two sides of your pool, then where they meet they interfere,” Buhrman explains. “They can interfere constructively and you get higher waves, or they can interfere destructively and you get no waves. The same thing is what we want to do with quantum computing, and that is you make this superposition of all these computations, but then nature allows you to have them interfere with each other. If you do this in the right way, you interfere the computations that you don’t want to see away, and you amplify the ones that you do want to see.”

This is very tricky to achieve, as the nature of the interference pattern you want changes depending on the type of computation you’re trying to perform, and even then it only works for certain kinds of problems.

So, what exactly are quantum computers useful for? For the current and near-future quantum computers – those achieving quantum supremacy with 50 to 100 qubits – the answer is not very much at all. “The sort of problems that we’re looking at are very mathematical in nature,” says Michael Bremner, professor in quantum computation at the University of Technology Sydney, who specialises in these specific, current-generation problems. “They’re more about developing milestones and benchmarking devices. The sort of algorithms I’m working on right now, they’re randomised computations, which, if you look at the output, it would take you a very, very large amount of processing to understand there’s anything other than complete randoms coming out of the device.”

Even when we arrive at practical quantum computers, which could take another 20 years, they will only perform exponentially faster than a conventional computer for a limited number of applications. Perhaps the most significant of these is their ability to simulate other quantum systems, namely atomic structures and chemical reactions. Electrons around an atom also exist in superposition, and with current computers, it’s very hard to calculate how they behave. “Every time you add an electron, the possibilities double, or even more than double, and with just a very few number of electrons, the possibilities are so big we cannot compute any more what happens on our classical computer,” Buhrman explains. “But on a quantum computer, which is inherently quantum mechanical, you can simulate these reactions.”

Being able to simulate chemical systems with such accuracy may enable us to design better medicines, better materials, and overhaul our ability to understand how the fundamental building blocks of the universe work, which is a pretty big deal.

The other commonly touted function of practical quantum computers is their ability to crack public-key encryption, such as that used by modern web browsers. When quantum computers arrive on the scene, this form of encryption will no longer be secure. This isn’t necessarily the most exciting use of quantum computers, but it is arguably the most relevant.

Indeed, Buhrman is keen to emphasise the significance of this ability, and how it is urgent we respond to it now, even though large-scale quantum computers don’t yet exist. “What will happen is that people can intercept this information. They cannot read it, they cannot decrypt it, but they can store it and save it for later, and then once the quantum computer is available, they can decrypt what [is] being sent now.”

There are new methods of encryption currently being explored to counter this future vulnerability. One simply involves more complex encryption on classical computers, such as Google’s New Hope program. “The problem here is that we’re never sure that this actually cannot be broken by quantum computer,” Buhrman observes. “Maybe someone comes up with a fast algorithm soon and breaks them.”

The other method involves sending information encrypted in a quantum-state as photons along a fibre-optic cable. This would work because the moment anyone tried to intercept the information, the quantum state would be disturbed, alerting the sender to the interception. At the moment, this only works over short ranges of around 300 kilometres. “The problem is that if you send photons through fibre, which is how the qubits are coded, then actually the fibre will observe the photons.”

These are the functions we know quantum computers excel at. But the truth is there may be many future applications that we simply cannot predict, just as the pioneers of conventional computing in the 1940s could not predict that we would be using them today for worldwide communication and sharing of information. It’s entirely possible that quantum computers will change the world in ways we cannot possibly imagine.

Back to reality

But let’s stop the star gazing. People have been making wide-eyed predictions about the impact of quantum computers for decades. Is there any hard evidence these things are anything but the ultimate vapourware?

The immediate next step is quantum supremacy. It is highly likely that a quantum computer will perform a very contrived computation faster than a conventional computer in the next year or two, but it won’t be a straightforward process. “The next thing is someone will say ‘Well I can run that on my supercomputer, and here’s the data’”, Bremner says. “What I’m interested in right now, what I’m developing as a next step, is to show how you can do this in an unambiguous way.”

Once supremacy has been confirmed, then it’s about ramping up the qubits to the point where quantum computers can solve practical problems. Typically, this is estimated to be up to 20 years away. But recent research by Hensinger and his students may reduce this estimate dramatically.

Last year, Hensinger and his team published a paper which demonstrated that, instead of using lasers to trap ions, the same result can be achieved by applying a voltage to a microchip. “We’ve kind of simplified the problem in a way that’s the same as a conventional computer, which is you have transistors in a conventional computer processor, and that’s basically applying a voltage to execute a logical gate,” he says.

This led to the development of a blueprint for a large-scale quantum computer. “When I say large-scale, what do I mean? I don’t mean 50, or 70, or 85, or 250 [qubits], I mean a billion.” Hensinger and his team are currently constructing a prototype of the computer in his Brighton lab, which he estimates will be completed in the next 18 months to two years.

Quantum computing is coming. “For myself, someone who’s worked on the theory side, it’s kind of amazing that we’re getting to the point where we’re gonna test some of the things I’ve worked on,” says Bremner.

But what if, after all the effort, they reach the point where they switch the thing on and it doesn’t work? “The point we are now in quantum computing, either it’s going to work the way we predicted it to work, or we’re going to learn about new physics we didn’t know anything about,” Bremner says. Here’s hoping he doesn’t get the consolation prize.

Lead image: Prof Winfried Hensinger and Dr Seb Weidt with a quantum computer prototype. Credit: University of Sussex