- Quantum computing has come a long way since its theoretical birth in the 1980s, with the works of Paul Benioff, Yuri Manin, Richard Feynman and David Deutsch. We still don’t have functional, large-scale, universal1 quantum computers, but it might not be too much longer before we do.
- Currently the domain of large companies such as IBM and Google, and physics research labs in universities, the search is on to find the best approach to building one. Leaving aside the work of Canadian company D-Wave Systems, which uses quantum tunnelling effects to solve problems, the two most successful methods for performing quantum computations are through the use of superconductors and trapped ions.
- Superconducting computers
- Make use of a Josephson junction: two superconducting electrodes with a barrier down the middle, which exhibits quantum effects when cooled to near absolute zero. Trapped-ion computers, on the other hand, suspend charged particles in magnetic fields to create their quantum gates and induce the desired effects.
- These quantum effects include being able to enter a superposition of states. A silicon computer bit can be on or off (0 or 1), but a quantum bit, or qubit, can be 0, 1 or both at the same time. It’s mind-blowing, but it works, and when fed the right algorithm it could achieve in an afternoon what it might take a classical supercomputer a billion years to compute.
- That’s pretty revolutionary, and the term “quantum supremacy” was introduced by Caltech’s John Preskill to mark the moment quantum computers will exceed the processing power of conventional silicon. That point comes when we have a processor operating at around the 45-50 qubit mark, and some pretty big names think it might be approaching fast.
- Earlier this year, Google announced it was planning to run a 50-qubit computer by the end of 2017, and IBM has plans to hit that mark soon too. Its latest machine, a superconducting model weighing in at 17 qubits, is still in the lab, but a five-qubit machine is running and available to the public, with a 16-qubit computer in beta testing. The IBM Quantum Experience has more than 50,000 users who have executed code more than 300,000 times, publishing their results in 17 scientific publications. There’s even an API and code on GitHub to help you get started.
- “Something you need to keep in mind is they should be perfect qubits,” says Dr Stefan Filipp, a quantum computing scientist at IBM’s research facility in Zurich, Switzerland. “That’s qubits without any influence from the environment, without any noise properties. There’s a grey zone around how many qubits you need to outperform classical computers, but 50 qubits is the first threshold. What we want to make is a universal2 quantum computer, and that requires perfect qubits, but we’re realistic enough to know that you don’t find perfect qubits.”
- What’s needed is some form of error correction. “We know now that if we have 100 or 1,000 imperfect qubits, we can distil from them one perfect qubit.” says Filipp. “So if you want to have 50 perfect qubits, depending on how imperfect the real ones are, we have an overhead of 1,000 or even more qubits.
- “It’s still the case that we need to build a system that’s actually capable of outperforming a classical computer. We are quite certain that we can do this, but it’s not clear whether this will be this year, next year or in five years3.”
- Trapped Ions
- Not everyone sees this model of quantum computer as the best way forward, however. Winfried Hensinger, professor of quantum technologies at the University of Sussex, has published a plan for building a quantum computer today, using existing trapped-ion technology.
- “Trapped ions are a very attractive candidate [for quantum computing] because they can work at room temperature,” he says. “To solve really interesting problems, you need millions or billions of qubits4, so imagine being able to cool all those quantum bits down to such a low temperature.” We’re talking 0.01K, or -273ºC, remember.
- “The method to implement quantum gates with trapped ions has been to use pairs of laser beams,” Hensinger explains. “They have to be focused to a precision of one-hundredth of the width of a human hair. That can be easily done if you have only a few ions, but imagine you want to build a quantum computer with millions and billions of qubits. The engineering requires you to have millions and billions of laser beams.
- “We’ve been thinking about this for a long time, and developed an entirely different approach involving applying voltage to a microchip to do exactly the same thing. We can now replace millions and billions of laser beams with voltages. In a way it’s exactly how a classical computer works – the transistors in a microprocessor work the same way. You apply a voltage, and that executes logical operations.”
- Sad as we are to see the back of quite so many lasers, at least the finished computer might be impressively large. “A quantum computer, because of the way it operates, can’t be very small because it’s unbelievably complicated to isolate the ions in such a way that the quantum effects aren’t being destroyed by things around them, and that’s where you get headlines about computers the size of football pitches5,” says Hensinger. The machine will need to be modular, with many smaller processors linked together using another innovation – connectivity using electrical fields, a sort of incredibly fast quantum Bluetooth. “We could start building it today, but it could still take ten years to build a large machine6.”
- Different interpretations
- Yet even if such a machine were built, what would we do with it? IBM’s Quantum Experience has seen experiments geared toward figuring out what works on a quantum computer and how you write an algorithm for it. In the future, we could see applications in quantum chemistry, as Filipp explains: “To describe an electron system, you need two complex numbers. For 100 electrons you need 21007 such numbers, and it would take all the data storage available at the moment to even store these. But a quantum computer can handle it, because you don’t have to store this information in a bitstream, but in real physical objects. And these quantum objects are described by quantum mechanics8, so they have these numbers already intrinsically in them.”
- Hensinger has a slightly different view. “A quantum computer isn’t a fast conventional computer,” he says. “One interpretation of how it works is it makes use of computations across parallel universes9 – and you can already see how mind-bending that is.
- “It can solve certain problems in maybe a few hours that even the fastest supercomputer in the world would take billions of years to calculate. At the moment an example is breaking encryption, but for every problem you want to solve, you need to write a new algorithm that makes use of this strange ability to do things in multiple parallel universes. We’re not going to take some software10, run it on a quantum computer and it will run very fast. That’s a misconception.”
- Hensinger draws an analogy with the Colossus computer created by Alan Turing and Tommy Flowers to crack Germany’s teleprinter codes during World War II. “In terms of conventional computers, quantum computers are right now in the 1940s,” he tells me. “We’re very impressed by them, but we don’t yet know what they can do.”
- And what’s run on them needs to be tailored to their abilities. “There are about 50 people in the world11, at most, writing quantum computer algorithms,” says Hensinger. “The key problems are those a conventional computer could never solve; it would take forever. It’s disruptive technology – it can change an entire sector of business by adding capability that was previously not available.”
- This certainly sounds remarkable, but Hensinger’s huge machines and Filipp’s absolute-zero cooling systems don’t sound very consumer-friendly. Will we all end up owning one? Filipp isn’t sure. “I think when we have quantum computers that are capable of replacing a laptop, the cooling requirements will be solved, but it’s not in the near term that quantum computers will replace desktop or laptop computers,” he says. “Our vision of a quantum computer can do anything a normal computer can do, in principle, but you wouldn’t use it for that because it’s too complicated at the moment.”
- It’s currently doubtful whether there will ever be such a thing, or even the need for one.
- Quantum computers now seem to be envisaged as being built to solve specific problems that can’t be solved in an acceptable timeframe by classical computers, but these latter are to be left to do the routine stuff as they will be smaller and cheaper.
Footnote 3: Footnote 4:
- We don’t – see previous Footnote.
- Surely this is a mistake? What is a “really interesting problem”?
- To create a fully-functional universal machine analogous to top end classical machines you might need billions of qubits, but that’s not – or ought not to be – the aim for quantum computing.
- Doesn’t Nick Bostrom consider computers the size of (the surface of) planets to perform his Matrix-style simulations?
- But, again, this only applies to the “millions of qubits” universal machines.
- Again “large” is not needed in the sense used here.
- The article had 2100, but that’s a typo!
- Why bring in “parallel universes”? This is just a (very contentious) interpretation of quantum mechanics.
- Hensinger’s views are sensible-enough without invoking parallel universes: they don’t depend on this view.
- This is a very important point. Quantum computers aren’t going to be universal machines you can run classical programs on.
- The algorithm – and maybe the machine – has to be crafted for the problem to be addressed.
- So, you can’t just run the program that would take 1,000 years to run on a classical computer and expect it to take 5 minutes. You need to write a completely different program to solve the same problem.
- No doubt this would ramp up – and maybe already has – once there’s hardware to run them on.
- But, it might be very difficult, so that few people are smart enough to do it. A bit like classical programming was before the hardware and software improved to make it easy for anyone to do it.
Text Colour Conventions (see disclaimer)
- Blue: Text by me; © Theo Todman, 2020
- Mauve: Text by correspondent(s) or other author(s); © the author(s)