Quantum computers promise to be able to solve tasks that would be impossible using conventional machines.

But those benefits are still theoretical at present, with quantum computers lacking a sufficient number of processing units, known as qubits, and enough stability to do useful work.

Companies are going to huge lengths to build quantum computers, cooling devices to a few micro kelvins above absolute zero. Even then challenges remain, while IBM has a 50-qubit prototype machine and Google a 72-qubit chip, each has their own roadblocks that prevent them from being truly useful devices at this moment.

Here is the expert view on what quantum computers will and won’t be able to do, and the challenges we still face.

SEE: Ebook–IT leader’s guide to the future of quantum computing (Tech Pro Research)

1. Quantum computer won’t replace classical computers

“Quantum computers will never be able to run the if/then/else type of logic that we’re familiar with with our traditional Von Neumann architecture computers, [where they are] sequentially going from step to step,” said Andy Stanford Clark, IBM CTO for UK and Ireland.

2. Quantum computers excel at optimization problems

“Quantum computers are really good at solving those problems where you’ve got an exponential number of permutations to try out,” said Stanford Clark.

“If, for example, you’re optimizing the lengths of aircraft routes, or optimizing the layout of spare parts for a rail network, something where there’s 2n possibilities and you’ve got to try each out in order to find the optimal solution.

“If you had a 2100 problem, which would be basically impossible to solve on a classical computer, with a 100-qubit quantum computer, you’d be able to solve it in one operation.”

Stefan Filipp, technical leader for superconducting qubit quantum computation at IBM, said: “We know a few algorithms where we can get exponential speed up. For example, quantum chemistry or material science questions, calculating properties of molecules, that’s definitely a thing where a quantum computer can help.”

3. Quantum computers will augment classical computers

“We’re not going to see people throwing away all their classical computers and replacing them with quantum computers,” said said Stanford Clark.

“We’re going to see, in the same way as you have a maths co-processor and a GPU on your classical computer…you’ll have a quantum computer co-processor alongside your classical computer.

“When you come to a point where you’ve got to solve some massive exponential problem, you’ll package it up, throw it over to the quantum co-processor, it’ll burp out its answer and then you’ll carry on with the answer to your computation in your classical algorithm.”

4. We need 50-60 qubit computers to do useful work

“The tipping point as to where classical computers give way to quantum computers is in the 50-60 qubit mark,” said Stanford Clark.

“In terms of years, I don’t know if it’s 5 or 10 [until we reach that point], probably that sort of order of magnitude where we have something at the 50-qubit level that is actually doing useful stuff in that it stays up long enough to do useful computation.”

5. Building a working quantum computer is about more than qubits

“We’ve got a prototype 50-qubit computer at the moment but the problem is one of quantum coherence, this 100-microsecond window during which it’s stable, means you can’t get very much useful done with your 50 qubits, so we’re a little way off that yet,” said Stanford Clark.

IBM’s Filipp adds: “The challenge is getting hardware to the point where we can use it and run practical algorithms.

“That means we have to not only increase the number of qubits, but also have to increase the coherence. We have to improve this to get to the point where we can solve practical algorithms.

“We have a roadmap that goes in the direction of improving coherence, but it is still a significant challenge in getting to real practical quantum computers.”

6. We have no idea how to write useful quantum software

“At the moment we have no real idea how to write big complicated algorithms for quantum computers as we’re so used to doing with classical computers,” says Stanford Clark.

“We just haven’t had the experience and exposure to try to try to solve problems with quantum computers that we have with classical computers.”

7. Quantum computers need error correction

“There’s also fault tolerance inside the qubits as well,” says Stanford Clark, adding that quantum computers will need an equivalent to the error-checking parity bits found in conventional computers.

“You’ll need the same sort of equivalent technology inside a quantum computer, so that we can detect when a bit has inadvertently flipped or come out in the wrong state, and therefore can be error corrected for.”