Queue up for IBM’s quantum computer

Today, IBM is releasing its first working model of quantum computing to the public by way of a Web interface to a small quantum computer the company has sitting inside a dilution refrigerator (at temperatures lower than outer space) on its own premises in Yorktown Heights, New York. Quantum computing is based on quantum mechanics, the odd behavior of matter at submicroscopic levels. In the quantum world, quantum bits or qubits, can hold both their zero and one states at the same time.

Although the theory of quantum computing has been around since the 1970s, programming with it is pretty new. And yet in some ways, the path quantum computing is heading down seems strangely familiar. With quantum computing, what goes around, comes around — in more ways than one.

First of all, the way people will access the quantum computer sounds like something from the 1960s, at least in certain respects. For now, there is only one operational computer, and users will have to queue up to use it, reminiscent of the old batch-job days when engineers stood in line with their decks of punch cards, waiting to run their jobs on an organization’s only mainframe. This queue, of course, will form in virtual space. Access to what IBM is calling the IBM Quantum Experience will be arbitrated online via a combination of first-come-first-serve and a kind of digital chit or virtual currency, which is, for example, awarded to beginners for taking the quantum computing training or lavished on experts who may advance the state of the art. Sorry, these chits can only be spent on the IBM Quantum Experience; there is no other market for them. 

On the positive side, only IBM Research could even do something like this today, and, by making quantum computing available to potentially anyone via the cloud, the company is arguably doing some sort of public service, even if it may benefit from what people learn while experimenting with the new system. This is science at its best. But I digress.

Second (we’re listing things coming around again), at the scale of this particular computer — which is composed of five qubits — regular serial, digital computers of the sort we use every day can check the results of the quantum computer. Back in the days of the first digital computers in the mid-1940s, scientists working on the Manhattan Project checked the work of the new digital machines with rooms full of human “computers,” beavering away on slide rules to check the digital computer’s results. (The job description “computer,” which often filled the classifieds in newspapers at the time, ultimately gave the name — and the job — to the digital machines.) Just the way today’s digital computers can make calculations far beyond the checking capacity of humans with slide rules, so tomorrow’s quantum computers will exceed the ability of even high-performance digital computers to verify.

Third, quantum computers and algorithms actually look — at least in conceptual space — like spheres. Quantum bits literally go around and come around at the same time. Jerry Chow, Manager of Experimental Quantum Computing at the T.J. Watson Research Center in Yorktown Heights, explained how the system works:

The Web interface presents something that looks for all the world like musical notation: a staff of five lines. Like musical notation, the progression from left to right on the staff represents passing time. The five lines represent the five qubits in the machine.

IBM

At the bottom of the page are symbols for gates, which can be dragged onto the staff to create a program. Three green square symbols, which are named X, Y and Z, derive their names from the idea that they represent the three dimensions of space in a qubit sphere. Taking a closer look at the description of one of the gates, a “Pauli X gate is a π-rotation around the X axis and has the property that X → X, Z → -Z. Also referred to as a bit-flip.” Got that? It’s not saying what the bit is, only that it will go to the opposite state. It is Heisenbergian to know where you’re going but not where you are.

Another curious gate is symbolized by a + sign and is called a Controlled-NOT gate. This one is “a two-qubit gate that flips the target qubit (i.e., applies Pauli X) if the control is in state 1. This gate is required to generate entanglement.” Entanglement is a fascinating property of qubits that lets the programmer make two of them move together. They can be either 00 or 11, but of course, this being quantum, you don’t know which, just that they are in the same state.

In a simple example, entanglement can be used to solve the question of whether a coin is “fair” or not in a single operation. A fair coin has both a head and a tail. A fake coin could have two heads or two tails. A normal computer would take two operations: look at one side, note what it is, and flip it over to ascertain that it’s whatever the first side was not. Using entanglement, you give up some information to gain other information. You don’t know whether you’ve got a fake coin with two heads or two tails, just that both sides are the same and therefore the coin is not fair.

You were just about to ask, I know it, why five qubits? The answer has to do with the instability of qubits. Part of the reason IBM’s quantum computer is sitting in an environment at a frigid 0.015 degrees Kelvin is that qubits are inherently unstable. Chow puts it this way: “A qubit wants to connect to the whole world, to give its information away to something else. We need to limit the exposure of a quantum object to other parts of the environment.” So, there’s a lot of “decoherence” in the system. In the current configuration, one qubit in the center can help keep the other four coherent, sort of like a checksum or parity operation.

Okay, so you drag your gates onto the staff. What then? Drop in your chits, join the queue, and wait for your program to run. You can run it once, 1024 times, or more, obtaining statistical results. Since the qubits are a little wobbly, some of the individual results will be off, but the pattern will be clear.

For now, five qubits can represent 2⁵ or 32 configurations. That’s small enough so that a regular computer can emulate the operation to check it for validity. Ultimately, a huge universal all-singing, all-dancing quantum computer will have 100 qubits. At that level, there would be 2¹⁰⁰ or 1,267,650,600,228,229,401,496,703,205,376 configurations, a number probably larger than there are stars in the universe.  Even at 40-50 qubits, no simulation could be run on a regular computer.

Assuming that IBM and all the professors, researchers, learners, tinkers, partners, amateurs, students, thinkers, and hangers-on can — using the IBM Quantum Experience and its follow-ons — figure out how to program a quantum computer for useful work, what would this thing actually do?

Well, for now, the application set is fairly limited: some chemical research and optimization problems might yield to even midsize quantum computers with simple error mitigation. But the ultimate machine — the one with lots of qubits and great error detection and correction — could do massive database searches instantly, perform uncanny superhuman machine learning and break any encryption known to man in a finger snap.