NEWSSCIENCE sciencemag.org 10 JULY 2020 • VOL 369 ISSUE 6500 131GRAPHIC: C. BICKEL/
SCIENCE
states give a quantum computer its power.
But they’re also fragile, and the slightest
interaction with their surroundings can
distort them. So scientists must learn to
correct such errors, and Kuperberg had
expected Google to take a key step toward
that goal. “I consider it a more relevant
benchmark,” he says.
If some experts question the significance
of Google’s quantum supremacy experi-
ment, all stress the importance of quantum
error correction. “It is really the differ-
ence between a $100 million, 10,000-qubit
quantum computer being a random noise
generator or the most powerful computer
in the world,” says Chad Rigetti, a physi-
cist and co-founder of Rigetti Computing.
And all agree with Kuperberg on the first
step: spreading the information ordinar-
ily encoded in a single jittery qubit among
many of them in a way that maintains the
information even as noise rattles the un-
derlying qubits. “You’re trying
to build a ship that remains the
same ship, even as every plank
in it rots and has to be replaced,”
explains Scott Aaronson, a com-
puter scientist at the University
of Texas, Austin.
The early leaders in quantum
computing—Google, Rigetti, and
IBM—have all trained their
sights on that target. “That’s
very explicitly the next big mile-
stone,” says Hartmut Neven,
who leads Google’s Quantum
Artificial Intelligence lab. Jay
Gambetta, who leads IBM’s
quantum computing efforts,
says, “In the next couple of years,
you’ll see a series of results that
will come out from us to deal
with error correction.”
Physicists have begun to test
their theoretical schemes in small
experiments, but the challenge is
grand. To demonstrate quantum
supremacy, Google scientists had
to wrangle 53 qubits. To encode
the data in a single qubit with suf-
ficient fidelity, they may need to
master 1000 of them.THE QUEST FOR QUANTUM com-
puters took off in 1994 when
Peter Shor, a mathematician at
the Massachusetts Institute of
Technology, showed that such a
machine—then hypothetical—
should be able to quickly factor
huge numbers. Shor’s algorithm
represents the possible factoriza-
tions of a number as quantum
waves that can slosh simultane-ously through the computer’s qubits, thanks
to the qubits’ two-way states. The waves
interfere so that the wrong factorizations
cancel one another and the right one pops
out. A machine running Shor’s algorithm
could, among other things, crack the en-
cryption systems that now secure internet
communications, which rely on the fact that
searching for the factors of a huge number
overwhelms any ordinary computer.
However, Shor assumed each qubit
would maintain its state so the quantum
waves could slosh around as long as neces-
sary. Real qubits are far less stable. Google,
IBM, and Rigetti use qubits made of tiny
resonating circuits of superconducting
metal etched into microchips, which so far
have proved easier to control and integrate
into circuits than other types of qubits.
Each circuit has two distinct energy states,
which can denote 0 or 1. By plying a circuit
with microwaves, researchers can ease itinto either state or any combination of the
two—say, 30% 0 and 70% 1. But those in-
between states will fuzz out or “decohere”
in a fraction of a second. Even before that
happens, noise can jostle the state and alter
it, potentially derailing a calculation.
Such noise nearly drowned out the sig-
nal in Google’s quantum supremacy ex-
periment. Researchers began by setting the
53 qubits to encode all possible outputs,
which ranged from zero to 2^53. They imple-
mented a set of randomly chosen inter-
actions among the qubits that in repeated
trials made some outputs more likely than
others. Given the complexity of the inter-
actions, a supercomputer would need
thousands of years to calculate the pat-
tern of outputs, the researchers said. So
by measuring it, the quantum computer
did something that no ordinary computer
could match. But the pattern was barely
distinguishable from the random flipping
of qubits caused by noise. “Their
demonstration is 99% noise and
only 1% signal,” Kuperberg says.
To realize their ultimate
dreams, developers want qubits
that are as reliable as the bits in
an ordinary computer. “You want
to have a qubit that stays coher-
ent until you switch off the ma-
chine,” Neven says.
Scientists’ approach of spread-
ing the information of one
qubit—a “logical qubit”—among
many physical ones traces its
roots to the early days of ordi-
nary computers in the 1950s. The
bits of early computers consisted
of vacuum tubes or mechanical
relays, which were prone to flip
unexpectedly. To overcome the
problem, famed mathematician
John von Neumann pioneered
the field of error correction.
Von Neumann’s approach re-
lied on redundancy. Suppose a
computer makes three copies of
each bit. Then, even if one of the
three flips, the majority of the
bits will preserve the correct set-
ting. The computer can find and
fix the flipped bit by comparing
the bits in pairs, in so-called par-
ity checks. If the first and third
bits match, but the first and sec-
ond and second and third differ,
then most likely, the second bit
flipped, and the computer can
flip it back. Greater redundancy
means greater ability to correct
errors. Ironically, the transistors,
etched into microchips, that
modern computers use to encodeBit-fip error
Exchanges 0 and 1, fipping
the qubit in latitude10Qubit
statePhase-fip error
Pushes the qubit’s state halfway
around the sphere in longitudePhaseEqual mix
of 1 and 0Mixture
of 1 and 0Mapping a qubit
Whereas an ordinary bit must be either 0 or 1, a qubit can be in any combination
of 0 and 1 at the same time. Those two parts of the state mesh in a way described
by an abstract angle, or phase. So the qubit’s state is like a point on a globe
whose latitude reveals how much the qubit is 0 and how much it is 1, and whose
longitude indicates the phase. Noise can jostle the qubit in two basic ways that
knock the point around the globe.