When it comes to practical problems, including things such as the traveling salesman problem, a classic in optimization, the value of quantum is still to be decided, say Richard Moulds, left, head of Amazon’s Braket quantum computing service, and Robert Liscouski, head of Quantum Computing Inc., which makes Qatalyst software to do optimization on both classical and quantum machines.
It’s easy to imagine a problem for which, if one had a computer that magically leapt across steps of the computation, your life would be much better.
Say, for example, a computer that auto-magically searches through a vast space of possible solutions much faster than you can with a CPU or GPU.
That’s the premise of quantum computing, and surprisingly, for all the hype, it’s not clear if that premise is true.
“I don’t think we’ve seen any evidence yet that a quantum machine can do anything that’s commercially interesting faster or cheaper than a classical machine,” Richard Moulds, head of Amazon Braket, the cloud giant’s quantum computing service, said in an interview with ZDNet. “The industry is waiting for that to arrive.”
It is the question of the “quantum advantage,” the notion that the entangled quantum states in a quantum computer will perform better on a given workload than an electronic system.
“We haven’t seen it yet,” Robert Liscouski, CEO of Quantum Computing Inc, said of the quantum advantage, in the same Zoom interview with Moulds.
That aporia, the as-yet-unproven quantum advantage, is in fact the premise for a partnership announced this month, whereby QCI’s Qatalyst software program will run as a cloud service on top of Braket.
QCI’s corporate tag line is “ready-to-run quantum software,” and the Qatalyst program is meant to dramatically simplify sending a computing task to the qubits of a quantum hardware machine, the quantum processing units, or QPUs, multiple instances of which are offered through Bracket, including D::Wave, IonQ, and Rigetti.
The idea is to get more people working with quantum machines precisely to find out what they might be good for.
“Our platform basically allows the democratization of quantum computing to extend to the user community,” said Liscouski.
“If you look back on the quantum industry since it started, it’s traditionally been very difficult to get access to quantum hardware,” said Moulds, including some machines that are “totally unavailable unless you have a personal relationship with the the physicist that built it.”
“We’re trying to make it easy for everyone to have access to the same machinery; it shouldn’t be those that have and those that have not, it should be everyone on the same flywheel,” he said.
The spectrum of users who will be working with quantum comprise “two important communities” today, said Moulds, those that want to twiddle qubits at the hardware level, and those that want to spend time on particular problems in order to see if they actually gain any benefit when exposed to the quantum hardware.
“There’s a lot of researchers focused on building better hardware, that is the defining force in this industry,” said Moulds. “Those types of researchers need to be in the weeds, playing at the qubit level, tweaking the frequencies of the pulses sent to the chip inside the fridge.”
On the other hand, “the other class of users is much more geared to Robert’s view of the world: they don’t really care how it gets done, they just want to understand how to program their problem so that it can be most easily solved.”
That second class of users are “all about abstraction, all about getting away from the technology.” As quantum evolves, “maybe it slides under so that customers don’t even know it’s there,” mused Moulds.
When it comes to those practical problems, the value of quantum is still to be decided.
There has been academic work showing quantum can speed up tasks, but “that’s not been applied to a problem that anybody cares about,” said Moulds.
The entire quantum industry is “still finding its way to what applications are really useful,” he said. “You tend to see this list of potential applications, a heralded era of quantum computing, but I don’t think we really know,” he said.
The Qatalyst software from QCI focuses on the kinds of problems that are of perennial interest, generally in the category of optimization, particularly constrained optimization, where a solution to a given loss function or objective function is made more complicated by having to narrow the solution to a bunch of variables that have a constraint of some sort enforced, such as bounded values.
“They are described at a high level as the traveling salesman problem, where you have multi-variate sort of outcomes,” said Liscouski. “But it’s supply-chain logistics, it’s inventory management, it’s scheduling, it’s things that businesses do today that quantum can really accelerate the outcomes in the very near future.”
Such problems are “a very important use case,” said Moulds. Quantum computers are “potentially good at narrowing the field in problem spaces, searching through large potential combinations in a wide variety of optimization problems,” he said.
However, “classical will probably give you the better result” at this time, said Liscouski.
One of the reasons quantum advantage is not yet certain is because the deep phenomena at the heart of the discipline, things such as entanglement, make the field much more complex than early digital computing.
“A lot of people draw the analogy between where we are and the emergence of the transistor,” said Moulds.
“I think that’s not true: this is not just a case of making the computers we have today smaller and faster and cheaper, we’re not anywhere near that regime, that Moore’s Law notion of just scaling these things up.”
“There’s fundamental scientific discoveries that have to be made to build machines that can tackle these sorts of problems on the grand scale that we’ve been talking about.”
Beyond the machines’ evolution, there is an evolution implicit for programmers. Quantum brings a fundamentally different approach to programming. “These are physics-based machines, they’re not just computational engines that add ones and zeros together, it’s not just a faster slide rule,” said Moulds.
That different way of programming may, in fact, point the way to some near-term payoff for the Qatalyst software, and Braket. Both Liscouski and Moulds expressed enthusiasm for taking lessons learned from quantum and back-loading them into classical computers.
“Typically, access to quantum computing is through toolkits and resources that require some pretty sophisticated capabilities to program to ultimately get to some result that involves a quantum computer,” observed Liscouski.
“With Braket, the platform provides both access to QPUs and classical computing at the same time, and the quantum techniques that we use in the platform will get results for both,” said Liscouski.
“It isn’t necessarily a black and white decision between quantum and classical,” said Moulds. “There’s an emerging area, particularly in the area of optimization, people use the term quantum-inspired approaches are used.”
“What that means is, looking at the ways that quantum computers actually work and applying that as a new class of algorithms that run on classical machines,” he said.
“So, there’s a sort of a morphing going on,” he said.
An advantage to working with QCI, said Moulds, is that “they bring domain expertise that we don’t have,” things such as the optimization expertise.
“We’ve coined the phrase, ‘Build on Braket’,” said Moulds. “We’re trying to build a quantum platform, and we look to companies like QCI to bring domain expertise to use that platform and apply it to problems that customers have really got.”
Also important is operational stability and reliability, said Moulds. For a first-tier Web service with tons of users, the priority for Amazon is “running a professional service, a platform that is reliable and secure and durable” on which companies can “build businesses and solve problems.”
Although there are “experimental” aspects, he said, “this is not intended to be a best-effort showcase.”
Although the quantum advantage is not certain, Moulds holds out the possibility someone working with the technology will find it, perhaps even someone working on Braket.
“The only way we can move this industry forward is by pulling the curtains apart and giving folks the chance to actually see what’s real,” he said.
“And, boy, the day we see a quantum computer doing something that is materially advantageous from a commercial point of view, you will not miss that moment, I guarantee.”
