Inside the high-stakes race to make quantum computers work

Scientific breakthroughs don’t always go exactly to
plan. On Friday, a draft research paper erroneously uploaded to
Nasa’s website accidentally tipped the world off that Google had
reached a quantum computing milestone: quantum supremacy.

It’s a goal that Google – and its competitors – had
in their sights for years. In 2017, the firm predicted it’d reach
quantum supremacy by the end of that year, but that deadline came
and went without any breakthrough. In the intervening years IBM and
Intel nipped at Google’s heels, testing quantum computers with
ever increasing numbers of qubits – the units of information that
are the reason that quantum computers are so potentially
powerful.

Now it appears that Google has reached this
particular milestone ahead of its competitors. The draft paper
details how Google researchers used a quantum processor called
Sycamore containing 53 functioning qubits to solve a random
sampling problem that would have taken the world’s best
supercomputers 10,000 years to work out. It took Sycamore just
three minute and 20 seconds. Google, which partnered with Nasa for
this project, did not respond to requests for comment.

But this breakthrough doesn’t mean that useful
quantum computers are just around the corner. Not by a long shot.
Instead, Google has just kicked open the door to the next era of
quantum computing. And that’s where things start to get really
interesting.

Quantum supremacy isn’t quite as exciting as it
sounds, says Simon Benjamin, professor of quantum technologies at
the University of Oxford. It just means the moment that a quantum
computer completes a task that conventional computers find
impossible. In Google’s case, that meant telling the quantum
computer to run a random set of instructions and then measure the
results. The researchers then tried to get a supercomputer to
predict what the quantum computer would produce, to make sure that
the results really were only achievable by a quantum computer.

This task is – practically speaking – pointless.
It’s good at sorting out quantum computers from their classical
kin, but that’s it. And that means that, in some ways, quantum
supremacy is pointless too. “It’s [a milestone] that must be passed
to get to the really good stuff, but it is not going to have any
immediate consequences at all,” says Benjamin.

Some researchers, including Benjamin, reckon that
the term quantum supremacy, which was popularised by the
theoretical physicist John Preskill, is a little overblown.
Benjamin has a suggested alternative – quantum inimitability –
which he says conveys the technical aspects of supremacy a little
more accurately, without inviting any Terminator-esque
comparisons.

But just because the actual task Google used to
demonstrate supremacy is pointless, it doesn’t mean that this isn’t
a big deal. “[The supremacy test] aims to be something that gives
you very high confidence, more than we’ve ever had before that this
machine is doing something computationally that cannot be
replicated on a classical computer,” says Toby Cubitt at University
College London’s Quantum Science and Technology Institute.

That’s big news, and a massive stamp of approval
for quantum computing. In October 2017, IBM demonstrated
that it could simulate quantum computers with 56 qubits
on non-quantum supercomputers. And if you can simulate quantum
computers, why bother going through the trouble of creating
room-size device that needs to be cooled to within a whisker of
absolute zero?

Google’s paper waves a massive flag reminding
everyone that there is a whole category of calculations that can
only be solved by using quantum computers, and opens the door to
people who might actually want to start designing problems to be
solved by those computers.

But what comes next is much more exciting. Google’s
experiment puts a line in the sand between two eras of quantum
computing. Welcome to the supremacy era. Ahead of us is something
called noisy intermediate-scale quantum (NISQ). If the end goal for
quantum computing is to be able to run calculations that no
classical computer could dream of running – complex molecular
interactions or cracking cryptographic codes – then NISQ is like
quantum’s awkward infancy.

The problem is that even the most promising quantum
computers are disastrously error prone. Imagine asking a computer
to perform a calculation and once every 1,000 times or more it’d
just spit out a completely random answer. And that’s per
calculation – a programme could be made up of millions of
individual calculations, so Google’s Sycamore processor would be
hopeless at tackling any kind of useful quantum problem.

The NISQ era is all about making quantum computers
that just about work, despite all this noise. The key, Benjamin
says, is working out how to squeeze value out of imperfect
machines. That could mean engineering programmes that are simple
enough to run on quantum computers while also yielding useful
results. And the other part will be improving hardware so that
quantum computers produce useful results while still only using a
relatively low number of qubits.

Benjamin has a name for the moment we crack the
challenge of making noisy quantum computers that are actually
useful: quantum advantage. “[It’s about] how we can squeeze value
out of imperfect machines,” he says. “At least then we’ll be in an
era when quantum computers have started to become useful
tools.”

And if we get past quantum advantage, that’s when
we can start thinking about the big challenges that only quantum
computers can solve, such as factorising very large numbers, or
modeling quantum mechanics. This will require a different kind of
quantum computer altogether, one that has ironed out all of its
errors and uses millions of qubits to make computations.

While Google might have the lead now, there’s no
way of knowing whether its quantum computer – based on
superconducting electronic circuits – will be the basis of future,
useful, quantum computers. “It’s hard to predict which hardware
will lead to a million qubit computer,” says Natalia Ares at the
University of Oxford. In the early years of quantum computing ion
traps were the most popular kind of hardware, but that was later
leapfrogged by the superconducting machines favoured by Google and
IBM.

For Ares, the Google news is exciting, but it’s
important not to expect useful quantum computers any time soon. The
largest quantum computer to date is Google’s 72-qubit
processor called Bristlecone, and that hasn’t even demonstrated
quantum supremacy. We’re a really long way off the millions of
qubits we’ll need to crack difficult quantum problems. “We are far
from doing anything very useful but it shows that we are making
progress,” Ares says. Google may have inched ahead for now, but the
marathon is only getting started.