USING the rules of quantum mechanics to carry out computations far faster than any conventional machine can manage is an idea that goes back decades. It was proposed in the early 1980s, but was confined to the blackboards of theoreticians until the late 1990s, when experimentalists gave it life by building simple machines which proved that the equations on those blackboards worked in practice. Now it has bloomed into a corporate project. Google, Microsoft, Hewlett-Packard and IBM each have dedicated quantum-computing research groups.
What quantum computing has not done, though, is make much impact on the outside world. And in some part that is because those quantum computers which do exist are still confined to laboratories. Only researchers have been able to tinker with them. Until now. For, on May 4th, IBM announced that it would connect one of its quantum computers to the internet and make it available for anyone to play with.
Quantum computing is exciting because it offers the promise of computers that can crunch through some kinds of mathematics (though not all) far faster than any classical computer that could ever be built could manage. This power comes from two counterintuitive phenomena: superposition and entanglement.
Superposition turns the fundamental unit of classical computing, the bit, into the qubit. A bit represents the smallest possible dollop of information: on or off; yes or no; 1 or 0. A qubit, though, is a mixture of both, superimposed upon each other. A classical computer with, for example, four bits can represent 16 different states. This machine can, however, exist in only one of those states at any given time. Its quantum equivalent, by contrast, can exist in a superposition of all 16 states at once.
But it is entanglement, which binds the fates of particles together, that really makes quantum computers sing. Entanglement makes it possible to manipulate groups of qubits all at once—so, as the number of qubits grows, the number of states a machine can occupy rises, quite literally, exponentially. A 300-qubit computer would have more possible states than there are atoms in the universe.
The result could manipulate prodigious amounts of information with ease. It could thus crunch through many tricky problems, from cracking cryptographic codes to simulating chemical reactions accurately at the molecular level. That is something ordinary computers find intractable, but which would prove useful for all manner of industrial processes.
A 300-qubit machine is far in the future. IBM’s current offering is a five-qubit processor built on a chip from loops of superconducting metal (see picture). It is suspended at the bottom of a large helium fridge at the firm’s research centre in Yorktown Heights, New York. This chills it to within a whisker of absolute zero—the lowest temperature possible—so that the chip’s delicate innards remain undisturbed by any stray puffs of heat. The chip is programmed by squirting carefully calibrated doses of microwaves into the fridge, with each qubit responding to a different frequency.
Qubit by qubit
None of that fiddly technical stuff, however, will be visible to users. Instead, they will be presented with something that looks rather like a musical staff: five horizontal lines, each representing one of the qubits. A collection of symbols, one for each quantum operation, can be dragged onto the staff. When a program, which IBM calls a “score” in a nod to the musical analogy, is ready, the user can press a button and the chip will execute it.
The Economist watched as Jerry Chow, who is in charge of the project, made the machine run Grover’s algorithm, a quantum algorithm designed to search through unsorted piles of data, and to do so faster than any classical machine. A classical computer could take as many tries to find an item as there are things to be searched through (so up to 52 attempts to locate the ace of spades in a 52-card pack, for instance). A quantum computer can do it in a number of steps equal to slightly less than the square root of the number of items to be searched. In the case of the cards, it needs just six attempts.
By itself, though, a five-qubit chip is not going to set the world on fire. A run-of-the-mill laptop can simulate quantum computers with as many as 40 or 50 qubits, says Scott Aaronson, a quantum-computing researcher at the Massachusetts Institute of Technology. According to Dr Chow, IBM plans to upgrade its quantum chip as the technology improves. But the main idea is not, yet, to produce a commercially valuable machine. Instead, he hopes to introduce the principles of quantum computing to as wide an audience as possible, to encourage understanding of those principles by potential programmers. The best way to do that, he thinks, is to give people “hands-on” time with a real machine.
Unlike classical machines, quantum computers answer questions probabilistically rather than definitely. A given result has only a certain chance of being correct. Exploiting their power is tricky. To get an answer it is necessary to measure the machine. That causes its quantum superposition to vanish, which leaves it in a single state, just like a classical computer. To ensure that state is the one containing the answer needs careful management, so that the probability of getting the correct answer is reinforced, while the chances of getting a wrong answer are suppressed.
Quantum computers thus need a great deal of coddling. Their superposed states are delicate and the slightest intrusion from outside—a stray electromagnetic wave or a tiny change in temperature—can make those states vanish prematurely. To work properly, then, a quantum computer must be isolated as much as possible from the rest of the universe. Bulky shielding and cryogenic cooling mean they are unlikely ever to fit on a desktop or into a smartphone. This means that big, commercially useful quantum computers will probably, like the prototype just unveiled, live in remote data-centres, wired up to the internet, waiting to be called upon by ordinary, classical machines whenever their specialised talents are required.
Powered by Facebook Comments