A battle is starting for influence over future quantum software platforms. This will have long term implications.
Break through announcements in quantum hardware continue to gather apace. While we are still a long way from a universal fault tolerant quantum computer, we are about to enter a period where noisy intermediate-scale quantum (NISQ) technology is a reality. A battle is beginning to be played out that will have long term and wide implications: what standard software platforms will emerge for individuals and companies taking their first steps into understanding and applying quantum computational techniques? The commercial battle to shape and dominate this domain is underway.
At QIP 2018 (the flagship Quantum Information Processing conference), taking place at QUTech (the Netherland’s world leading quantum technologies hub) top industry players took the opportunity to present their wares.
The big battalions
IBM first offered cloud based access to its early 5 qubit processor in March 2017. In addition to a very accessible introductory web graphical interface, their QISKit quantum computing platform now features the SDK software development kit for working with the OpenQASM intermediate level language. The IBM Q service has certainly been a remarkable success: its early adopter base claims 60,000+ cloud users across 7 continents, 1.7 million experiments run, 35+ external papers, 1500+ colleges and universities, 300 schools, 300 private institutions. 16 and 20 qubit processor options have recently been added. IBM stands to benefit from good will, lessons learnt and feedback from this base. However this also puts it on point for a pertinent question from the QIP floor: what is the status of copyright and privacy issues surrounding experimental code being developed and executed over the cloud? Commercial adopters will expect providers to get ahead in offering options to manage this.
Microsoft’s Quantum Development Kit features Q#, a new quantum focussed programming language, and supporting libraries. It is a major push to persuade the field to become part of its pervasive .NET world. It emphasises a high level language environment with programmer defined types, operations and functions. These are the well-worn advantages of a hardware agnostic approach protecting a client’s code investment in the medium term, though normally at the expense of some optimisation in the short term. By basing itself within the well establish Microsoft Visual Studio development environment it immediately offers a polished experience that will court familiarity with the conventional development community. It also points towards easy inter-operability with routines written in other conventional languages. Microsoft offers local and cloud based simulation, though not yet execution on real quantum hardware. It remains to be seen whether this close connection with the .NET world will be an advantage or a distraction during the particular opportunities and challenges of the NISQ era.
Intel didn’t present directly at QIP, but their research partner Leo DiCarlo of QuTech did cover software issues as part of his presentation. The QUTech team have done extensive and impressive work across the full emerging stack. The OpenQL compiler platform is now engaging users for a beta trial: it assumes a master programme in Python or C++, an intermediate hardware independant quantum assembly language (QASM), which can then be compiled into the hardware specific eQASM instruction set for execution on one of the prototype intel devices. As work progresses at QuTech we can expect to hear more about the quantum device microarchitecture QuMA that has been developed.
Google’s presentation by contrast didn’t offer a software platform. Starting from the widely believed observation that ‘textbook’ quantum algorithms will be beyond the reach of NISQ devices, they focus on what is needed to explore near term opportunities: heuristic problem solving approaches, augmenting the ‘human element’ in hand optimising early ‘low depth’ algorithms. An interesting example of this is the Quirk quantum circuit simulation and visualisation tool Google engineer Craig ‘the compiler’ Gidney, has created. Google also emphasised the value it sees in application area specific software initiatives. A prominent example is OpenFermion, which aims to provide specialist tools to speed-up application development in quantum chemistry (and related simulation areas).
Alibaba didn’t present at QIP, but their academic recruitment efforts were well represented at the conference as they seek to build-out the major investment they are making in quantum computing research. They have just announced cloud access to their own 11 qubit quantum device, developed in association with the Chinese Academy of Sciences.
New Model Armies
Rigetti, a full stack quantum computing start-up active in both hardware and software, demonstrated its cloud friendly quantum development environment called Forest, which supports programming with the Quil instruction set and pyQuil library. Forest emphasises accessibility and the easy combination of hybrid quantum and classical programme elements aimed at near term applications. Chris Osborn won one of the few spontaneous rounds of applause at QIP, by writing a small but non-trivial trial programme from scratch, testing it on a simulator then re-targetting it to run live on their own cloud-based 8 qubit quantum processor (access to their 19 qubit processor is also now online). Rigetti can’t aim to match the scope and resources of the tech majors, but it places great merit on being a pure play company that “has to make it work”. Engaging with their small staff you get the sense of motivation and excitement this generates.
Though not represented at QIP, D-Wave is of course another start-up champion, but with a completely alternative approach to quantum computing known as quantum annealing. Leaving aside the merits of the hardware battle, the striking thing about listening to the experiences of the D-Wave user base is the completely different approach to problem definition and solution used by this ‘adiabatic’ computational approach. D-Wave’s suite of tools helps users map their problem to a rugged landscape of peaks and valleys. The lowest point on this terrain is the optimisation that the user interprets as the solution. D-Wave now plans to widen its reach by offering a cloud based service later this year. It could be that for specific problem classes tools of this sort will prove more accessible than conventional procedural programming approaches.
Others, though not presenting at QIP, are progressing their own very interesting strategies.
Q-CTRL, is an Australian quantum software start-up focussing on the firmware and control suite needed to get maximum performance out of early quantum hardware. Its Black Opal launch product focusses on providing a library of noise control protocols with machine learning support to allow hardware to be tuned for optimal performance. Its aim is to provide a service that is hardware agnostic, across not just vendors but also competing qubit technologies. Tools for tuning qubit performance are likely to be valuable over the coming years.
1QBit is another start-up at the opposite end of the software stack. If focusses on working with blue-chip industry clients to tackle what currently seem intractable problems. Even when running ahead of today’s actual hardware capability there are advantages to 1QBit and its clients gaining experience. Innovative thinking can spur development in conventional solutions and this type of approach tends to attract and retain bright staff.
Quantum simulators illustrate yet another contribution (not to be confused with the simulation of quantum algorithms on conventional computers, quantum simulators are real quantum systems, but used in an analogue rather than programmatic mode). In the first instance they can be used to model the physics and chemistry of less accessible systems, however they also have potential application to more abstract problems. Jose Rosales and Vicente Martin continue to progress their interesting proposal to solve integer factorisation problems using a Penning ion trap. This type of approach is perhaps more like mathematics than software. The required problem must be restated so that at a difficult step in the maths its form mirrors the physical equations of the simulator system. This allows a series of physical measurements to be interpreted as the required mathematical solution. Sometimes a partial solution is all that is required to dramatically accelerate conventional calculations.
High-end theoretical academic work also continues to gather pace in the sector. The many themes of research on display at QIP are themselves a powerful illustration of this. The theory of optimising quantum compilers is not for the fainthearted. This year Earl Campbell took the conference through some of his latest results that serve to illustrate this. Unlike conventional digital logic gates, for non-trivial quantum algorithms the promise of even a ‘universal’ set of quantum gates is not exact, it’s only that we can always achieve a sufficient approximation to the required result by extending the sequence of gates executed. Naturally this has a cost. Remarkably this benefits from the use of randomisation in the compiled gate sequence, as this acts to ensure errors don’t conspire to the worst case and so fewer gate operations are required.
Communities such as QIP will continue to champion this cause. The EU’s FET flagship initiative on Quantum Technologies already identifies software as a key enabling area. Research groups such as the Netherlands QuSoft and the UK’s NQIT are seeking to raise its prominence further through their work programmes and lobbying initiatives such as the Quantum Software Manifesto.
The battle to dominate quantum software has the potential to be diverse. The history of digital tech development over the last forty years arguably shows that the advantages obtained here are even more lucrative and sustainable than those in hardware. This may be true both for companies and nations. The battle to influence and dominate this sector has certainly begun.