ISSUE 378|COMPUTERSHOPPER|AUGUST 2019 119119
THE D?WAVE COMPUTER
While researchers in academia were just
starting to demonstratequantum computers
with double-digit numbers of qubits back in
2011, Canadian company D-Wave announced
the world’s first ever commercial quantum
computer,the D-Wave One.Ithad 128 qubits.
Experts continue to be divided on whether
this machine or its successors, including the
company’s latest –the 2,048 qubit D-Wave
2000Q –are genuinely quantum computers.
However,D-Wavedrew our attention to papers
in two prestigious peer-reviewed academic
journals that describe the architecture as
employing quantum effects.
D-Wave says it still has work to do to reach
benchmarked quantum advantage –bywhich
it means obtaining proofthat the architecture
achievesaspeedadvantageoveraconventional
computer –and it says it will be an important
milestone when it, or anyone else,does.
Akey factor to recognise when considering
D-Wave’s 2,048 qubits compared to around
50 elsewhere is that we’re not comparing like
with like.The quantum computers discussed
in the main part of this article are generalpurpose,inthe same waythat today’s PCs are
general purpose.Byway of contrast, D-Wave
says that its machines employthe technique of
quantum annealing, which is dedicated to
certain types of problems. Included here are
optimisation, machine learning and image
analysis, with applications in areas such as
logistics, artificial intelligence,materials
sciences, drug discovery,cyber security,fault
detection and financial modelling.
Three organisations that apparently do
believe in the potential of this approach are
NASA, Google and the Universities Space
Research Association, which in 2013 jointly
invested in a512-qubit D-Wave Two. This has
since been upgraded to 2,048 qubits.
NASA says it’s using this system to
investigateareas where quantum algorithms
might somedaydramatically improve the
agency’s ability to solve difficult optimisation
problems in aeronautics, earth and space
sciences, and space exploration. It goes on to
refertothe keyareas where this type of
machine can excel, namely machine learning,
optimisation and pattern recognition.today’s mainstream microprocessors
might have a64-bit architecture,they
contain many more than 64 binary bits
when all the internal registers are taken
intoaccount. We asked Filipp whether
that 50-qubit figure relates to the width
of aregister or the total number of
qubits, or something entirely different.
Interestingly,his response suggested
that the concept of aregister doesn’t
apply to quantum computers.
“For aquantum computer we do
not talk about registers but about
the total number of quantum bits,”
Filipp explains.
“Each qubit has to be connected to
at least one other qubit so that, by
applying quantum gates acting on
single or two qubits, one can create
entanglement. This entanglement can
spread over all qubits. Moreover,things
are abit more complicated when we
start talking about quantum error
correction. Since the qubits are not
perfect and will not be,because of their
interaction with the environment, a
universal quantum computer –the
ultimategoal in the field –will have to
rely on error-correction schemes. For
these error-correction schemes one
needs to encode each logical, error-
corrected qubit intomany physical
noisy,imperfect qubits with typical
numbers of 1,000 and more physical
qubits needed to get one logical qubit.”THEWAY AHEAD
The reference to needing 1,000 physical
qubits to implement asingle logical
qubit, and the fact that the coherence
time of IBM’s qubits is just 90
microseconds, is quiteaneye opener.
It’s quitedifferent from what we’re used
to in today’s computers, where bits
remain in the statetheyare set. Clark
notes that “qubits are tremendously
fragile”, by which he means that they
easily lose that stateofsuperposition.We’ve seen that deliberately
observing aqubit causes it to decohere,
but it transpires that this is just the tip
of the iceberg.
“Any radio frequency noise or
unintended observation of them can
cause data loss,”Clark explains. His coin
analogy –inwhich aspinning coin
represents astateofsuperposition,
heads and tails simultaneously –helps
explain this, and also why the problem
increases with the number of qubits.
“Consider ahandful of coins,”he
says. “Imagine you and three or four
friends are trying to get as many coins
as possible spinning on atabletop
without any falling over.The longest
you would likely get aquarter to spin is
about eight seconds. Sayyou get
another coin spinning every second, the
most coins you could have spinning
simultaneously would be seven coins,
and that seven-coin ‘system’would only
be active forone second in total beforeTOP:The D-Wave
2000Q has 2,048
qubits, farmore
than researchers
at Intel or IBM
have achieved, but
it uses adifferent
architecture for
optimisation
problemsIm
ag
e:D-
W
aveSy
stems