TheEconomistSeptember 28th 2019 711I
n an articlepublished in 2012 John
Preskill, a theoretical physicist, posed a
question: “Is controlling large-scale quan-
tum systems merely really, really hard, or is
it ridiculously hard?” Seven years later the
answer is in: it is merely really, really hard.
Last week a paper on the matter was—
briefly and presumably accidentally—pub-
lished online. The underlying research had
already been accepted by Nature, a top-tier
scientific journal, but was still under
wraps. The leak revealed that Google has
achieved what Dr Preskill dubbed in his ar-
ticle, “quantum supremacy”. Using a quan-
tum computer, researchers at the informa-
tion-technology giant had carried out in a
smidgen over three minutes a calculation
that would take Summit, the world’s cur-
rent-best classical supercomputer, 10,000
years to execute.
A credible demonstration of quantum
supremacy, which few disagree that the
leaked paper represents, is indeed a mile-
stone. It will divide the history of the field
into two eras: a “before”, when quantumcomputers were simply hoped to outpace
even the best classical kind, and an “after”,
when they actually did so. There has been
much talk, including in this newspaper,
about the latter era. Now it has arrived.Leaping forward
Google’s experiment was “circuit sam-
pling”: checking whether numbers their
machine spits out, given random inputs, fit
a particular pattern. This niche task was
chosen to be easy for a quantum computer
while still being checkable—just—by a
classical one. It does, though, confirm thatquantum computers may in time be able to
handle long-standing matters of practical
importance. These include designing new
drugs and materials, giving a boost to the
field of machine learning, and making ob-
solete the cryptographic codes that lock up
some of the world’s secrets.
Quantum computers employ three
counterintuitive phenomena. One is “su-
perposition”, the idea behind Schrödin-
ger’s famous dead-and-alive cat. Unlike
classical bits, which must be either one or
zero, “qubits” may be a mixture of both.
Google’s machine has 53 qubits, which be-
tween them can represent nearly ten mil-
lion billion possible superposed states.
The second is “entanglement”, which
ties quantum particles together across
time and space. In standard computers
each bit is rigorously sequestered from the
next. Quantum machines like their qubits
entangled. Mathematical operations on
superposed and entangled qubits can act,
to a greater or lesser degree, on all of them
at once.
A quantum calculation starts by ad-
dressing qubits individually: making one
of them mostly zero, say, and then entan-
gling it with its neighbour by a certain
amount. That done, it lets the rules of phys-
ics play out, with the qubits’ states and
linkages evolving over time. At the end (but
not before, which would ruin the calcula-
tion), the qubits are examined simulta-
neously to obtain an answer.Quantum computingSchrödinger’s cheetah
Proof that a quantum computer can outperform a classical oneScience & technology
73 Manipulativerobots
73 Medicine,geneticsandthelaw
74 A taleofliliesandmonkeys
74 An interstellar visitorAlsointhissection
72 Ice cores from Antarctica