IBM Q division aims to offer commercial systems of approximately 50 qubits “in the next few years”

In 1954, Albert Einstein, struggling with the

weirdness of quantum mechanics, wrote that God does not play dice with the

universe. A little over 60 years later, IBM and a handful of other

organizations are not only playing with those dice, they’ve learned how to

consistently roll 7s, metaphorically speaking.

IBM earlier this week **set
up a new business**, called Q, to commercialize quantum computing. IBM

has had an experimental 5 qubit system for more than a year that the company

said 40,000 researchers around the world have already used for over 275,000

different experiments.

With the establishment of IBM Q, the company

promised two things: easier access to the quantum computer it has now, and a

commercial system of approximately 50 qubits. IBM said it intends to start

selling a 50-qubit system “in the next few years,” but that’s as specific as

the company will get on the timing.

A computer bit (“binary digit,” you no doubt recall)

is to a quantum bit (qubit) what the black and white color palette of the very

first filmed cartoons (e.g. Gertie the Dinosaur) is to the spectrum of color of

motion pictures filmed in high dynamic range (e.g. The Lego Movie). Where a bit

can be either a 1 or a 0, a qubit can represent 1, 0, and also a range of

values in between. The range increases exponentially with the number of qubits;

a 250-qubit computer could contain more bits than there are particles in the

universe.

As befits all quantum phenomena, qubits behave weirdly.

Based on many of the extant papers, a significant percentage of the work on IBM’s

5 qubit model is focused on simply characterizing the behavior of qubits and

the functionality of quantum computers. One of the results of all this research

is that the weirdness *can* be

characterized. It requires some sophisticated math to do it, but algorithms have

been developed to take advantage of qubits for computing purposes.

A 50 qubit computer should represent a profound jump

in power not only from 5-qubit models, but also from today’s common digital

processors. Such a device will enable people to solve problems they literally

could not hope to solve otherwise. For at least two years IBM has been using

the example of a caffeine molecule, which has so many possible quantum states

it is simply beyond the capability of standard computers to model them all –

and caffeine is a relatively simple molecule.

The potential for developing new drugs obviously

follows, perhaps even drugs designed for specific individuals. Other

applications IBM Q envisions for quantum computing include supply chain

management (think of the USPS in the weeks before Christmas), computer

security, and as an adjunct to machine learning.

To entice more researchers to familiarize themselves

with quantum computing in advance of producing a commercial product, IBM Q has

introduced an API (application program interface) that will allow anyone to

connect a standard to computer to IBM’s existing 5 qubit quantum computer via

the IBM Cloud Platform. The company also released a simulator that will enable

users to see what it might be like working with a 20-qubit version. The company

promised a software development kit (SDK) for working with that 20-qubit simulation

will be available by summer.

Beyond that, IBM has been parsimonious with details

about its quantum computer; it did not return calls for this article.

For example, it’s not clear what superconducting

material IBM is basing its quantum computer on, although Jerry Chow, manager of

the experimental quantum computing group at IBM Watson Research Center, said it

requires cooling near absolute zero in a 2015 TED talk. The company has publicized

work with niobium-aluminum alloys (which go superconducting near absolute zero)

on silicon substrates but it has never said explicitly that these are the

superconducting materials used in its quantum computers.

Jerry Chow’s TED talk:

There are questions about what it takes to scale up

the number of qubits, which might provide some indication on the timing of

IBM’s promised introduction of a quantum computer with roughly 50. It took the

company several years to double the count from 2 qubits to 4, which it

accomplished in 2015. At that time, Chow said IBM was experimenting with an

8-qubit model, but two years later the company is still at 5.

It’s also not precisely clear what the distinctions

are between extant quantum computers, although that might be more easily

extrapolated. The other most famous quantum computer is from D-Wave Systems

Inc., which is barely more forthcoming on details than IBM.

In 2015, D-Wave introduced a commercial quantum

computing system with 1,000-plus qubits, which would appear to represent an eye-popping

lead over IBM, but the two seem to be building different beasts. You can tell

people that a goat and a moose are both ungulates, but there’s not much more

the two have in common.

There are **three different categories**

of quantum computer: quantum annealer, analog quantum, and universal quantum.

IBM calls the quantum annealer the least powerful

and the most restrictive, though the easiest to build. Annealers are

restrictive inasmuch as they’re suitable only for solving optimization

problems, at least in IBM’s opinion.

IBM characterizes the analog quantum as a system of

between 50 and 100 qubits, with a more general applicability to a wider variety

of problem types, though still a limited set.

IBM says the most powerful quantum computer is the

universal quantum. Such a device would consist of more than 100,000 physical

cubits and would be useful for solving nearly any problem type. IBM says such a

device will be difficult to build, however, in part because of “a number of

difficult technical challenges” that will first have to be solved.

D-Wave and its partners, which include Google and

NASA, are explicitly focusing on annealing systems.

It’s not entirely clear where IBM is headed, but it

dismissiveness of quantum annealers suggests its ambition is greater. When the

company talks about the problems it is hoping to solve, those problems map against

the list of applications that an analog quantum would be appropriate for;

certainly the number of qubits is in the same order of magnitude.

That said, some of those same applications appear on

IBM’s list of applications a universal computer would be good for, and

occasionally when IBMers talk about quantum computing they refer to universal computers.

IBM also said that any organization

interested in collaborating to explore quantum applications can apply for

membership to the IBM Research Frontiers Institute, a consortium that develops and shares a portfolio of new

computing technologies and evaluates their business implications. Founding

members of the Frontiers Institute include Samsung, JSR, Honda, Hitachi Metals,

Canon, and Nagase.

IBM is

making the specs for its new Quantum API available on GitHub (https://github.com/IBM/qiskit-api-py) and providing simple scripts (https://github.com/IBM/qiskit-sdk-py) to demonstrate how the API

functions.