REBUTTAL.txt 6.94 KiB
We thank the referees for their encouraging comments on our proposal.
We are further encouraged by the fact that the few concerns can be straightforwardly allayed.
We now address the factual inaccuracies that form the basis of these concerns.
----
Concerning:
"However, it should be stressed that the non-incremental aspect of
this proposal is not entirely ensured. The fact is that this proposal
basically aims at implementing the ***already existing AZX concept*** in
(more or less) practical systems. This gives some degree of predictability to the
proposal but at the same time does not demonstrate a fundamental
breakthrough which would constitute a major advantage with respect to
other proposals in this call."
Let us emphasise that, contrary to this claim, AZX has not been developed yet. More precisely, while the ZX-calculus, which is aimed at theoretical work, is
a mature system, defining a practical language for quantum computing like AZX
-- including the automation, hardware-driven optimisation, and other
advanced features we propose -- will require significant technical
work. The tasks described in WP3 and WP4 will produce a substantial
amount of new theory and techniques with broad application in other
areas of computer science. Elsewhere in the report, the proposal is
described as "transformational", "radical", "novel and not
incremental", and "truly significant".
----
"it is not entirely clear from what is described in the proposal what
the dissemination tasks will bring to future programmers. Will they
use the outcome of this project as a black box of which they know
nothing about [?]"
Just as with existing compilers for classical programs, the AZX
toolchain could be used as black box, but knowledge of the compiler's
inner workings may assist in writing better programs. However the
more immediate consumers of the AZX outputs will be: (i) programming
language implementors who can use AZX to add hardware support to their
languages, and dually (ii) hardware designers who can use AZX to
provide a ready-made software layer. To achieve dissemination in
these communities our consortium and board of external advisors
includes leading quantum programming language designers (Valiron,
Staton, Selinger), hardware teams (Delft, Zurich, Vienna), as well as
the NQIT project and private companies (ATOS-Bull, Rigetti). Further,
our commitment to open source software and publicly documented file
formats and APIs (See T1.4 and T4.5) ensures that the wider community
may adapt and use our work as they see fit.
----
"A concern is that this approach is counter to
convention (the gate approach) and this will inevitably slow down the
impact"
We note that AZX is not intended to replace the gate approach, but
rather provide a strict superset quantum gate language. Users more
comfortable with the gate model can continue to use it and still see
benefits from this project. Indeed, quite early in the project,
Milestone 1 will provide a circuit to circuit pipeline. However we
believe that the extra power and flexibility provided by the AZX
tensor language will provide compelling reasons to take advantage of
the full language.
----
"It is not clear how the results of this proposal can percolate
correctly to industry if this is to become a standard in quantum
computing."
Formal standardisation is premature, although the implementation and
its public APIs (T1.4, T4.5) may serve as a de facto standard.
Our proposal offers a direct channels to industry via our partner
ATOS/Bull and the existing infrastructure established by the NQIT
project. As noted elsewhere in the reviews "Direct links with NQIT are
highly beneficial to the project", citing their expertise in ion
traps. NQIT also has 20+ industrial partners and hosts a biannual
Industry Forum, where AZX results will be presented (see p.12 of
proposal). The three workshops planned during the project will also be
aimed at attracting members of industry (see p.11 of proposal).
----
"It is unclear that there would be much in the way of side-benefits"
The project entails the integration of many distinct novel
contributions, which will be beneficial in their own right. For
example: our open test suite and benchmarks (T1.4) will useful to the
broader quantum computing community; our circuit-to-circuit optimiser
(Milestone 1) can be exploited by anyone using the gate model of
quantum computation; due to the embedding of circuits in AZX, our HPC
simulator (T2.3) will provide benefits to those that do not have
access to quantum devices; Tasks T2.4 and T2.5 will develop greater
understanding of the characteristics of superconducting and ion trap
qubits; T4.3 and T4.4 develop effective methods for programming
quantum architectures, which are beneficial even without the AZX
platform; all the tasks of WP3 involve developing new techniques for
working with diagrammatic languages, a widely used paradigm in
computer science and other disciplines.
OLD STUFF, NOTES:
Our proposal does have one significant channel to industry, via the
NQIT project. The NQIT architecture is one which we propose to target
(T2.5), and for which it is reasonable to expect that AZX will be
particularly well-suited. The strong connections between our
consortium and the NQIT project will allow us to ensure that our work
in this direction will be directly relevant and useful to NQIT's
internal work on architecture and compiler technology. Conversely, the
advantage to the NQIT project in using the tools which we develop will
directly motivate NQIT to present of those same tools as an API. In
particular, the AZX tools will be presented in conjunction with NQIT’s
internal work in its biannual Industry Days, in planned Application
Workshops, and in ad hoc interactions with specific industrial
partners who are interested in quantum technology.
(Augment or
supplement the above with remarks about NQIT interaction with standard
bodies??
Dom says:) In terms of dissemination through NQIT, they are plugged into the
various industry standards bodies (I knew turning up to that boring
standards session at the first NQIT conference would come in handy). I
can dig out my notes this evening and put together some names of
agencies, people, and the types of pathways to standardisation that
were discussed in the context of NQIT software, which we could use.
The Oxford site of the consortium has recently developed a separate
industry partnership with Cambridge Quantum Computing (CQC), who have
developed an interest in ZX as a technology for simplification of
quantum circuits. This is one of the central aims of the AZX toolset,
and it can be expected that the partnership with CQC will both enhance
the productivity of the project as well as a direct way in which our
results will be disseminated to industry.
(Are similar remarks
possible for Delft, Rigetti??)
This positions AZX technology in a
symbiotic role with specific industrial users, and (one or more??)
major quantum technology projects with an international reach in their
interactions with industry.