“Adding Quantum Value” is a series of blog posts that explain why adding value to your business with quantum computing is so difficult. It is intended to help the reader understand the different subtleties, and to pinpoint where their own pain points might lie. If you have such pain points, reach out and see how we might be able to help you.
In this fourth blog I approach why the lack of standardization of our quantum compute consumption hampers the adding of value.
Hybrids
There are a few major challenges around how we consume quantum computing. And they are barely addressed in popular blogs. You might say; “I’ll just use the IBM cloud quantum computer”, but in practice there is a catch. IBM (no longer) allows you to tune your own instruction set to a pulse level. They stopped doing this when they started to develop their own hybrid quantum computing capabilities.
IBM is not the only quantum hardware company recognizing that hybrid classical-quantum algorithms have the future. IonQ’s VP of Product Development Matthew Keesan states:
“There are lots of things that classical computers are better, or faster at, especially with our current generations of hardware. By letting the quantum computer do what it’s good at, and the classical computer do what it’s good at, you can get more out of both,”
So, what are these hybrid algorithms that these hardware giants refer to? Take one that has a major impact on the finance sector, pricing of options currently done by classical Monte Carlo simulation. Why this impact is so large, is a question for a whole presentation or lecture, but a simple google search will confirm that it is one of the major commercial applications currently driving most investments in quantum computing by financial institutions.
In essence, the quantum alternative of option pricing means that the probability distributions, originally generated by Monte Carlo simulations, are all kept in a relatively small quantum memory. One that benefits from that exponential scaling of superpositions. You can probably fit the entire option portfolio of BlackRock in a 32-qubit quantum oracle and compute it’s value at risk in a few seconds. But there is a catch. It takes a very long time to upload or generate all the information into that quantum oracle. So that 32-qubit oracle must be completely fault tolerant. And beyond that we cannot keep rebuilding that oracle over and over. I.e. we cannot observe the oracle without destroying the value it adds. We could, but we’d achieve no speed up over the classical alternative.
Amplitude estimation comes to the rescue, it gives an estimation of a value from the oracle separately, than observes and classically computes this to alter the oracle so that we can slowly extract information from the oracle without destroying it. This is a fully hybrid quantum algorithm because to function the CPU needs to be able to interrupt the QPU, and vice versa.
Let’s interrupt for a second!
So, the key to unlocking this value are these interrupts, and this is what is meant with hybrid quantum computing capabilities; the ability to interrupt QPU codes.
And here comes our immediate problem, if I consume my quantum computer through the cloud, I cannot do interrupts unless I have all the classical compute and data available right next to the quantum computer to minimize the latency between the communications of the QPU and CPU. This is especially valid when I look at minimal error correction. Low latency equals value. And currently that is not offered anywhere. There is no quantum-HPC hybrid cloud provider out there yet. Or at least none that understands the demands of these clients.
And that is one major problem with choosing your solution, you cannot just consume cloud as you are used to (yet). You could of course opt for purchasing your own quantum-HPC hybrid, but the challenges that come with that are greatly different. After all you want absolute control of your own system to optimize the return of a powerful application like option pricing. A 0,1% improvement can save you millions of dollars. Which means you are stuck with an expensive deal that the full-stack providers really don’t like, or an open quantum system with its own plethora of challenges.
Quantum is a barely cloudy day
Somewhere in the future I’m sure smart quantum hardware companies will see the light and offer the Quantum-HPC hybrid capacity in reliable cloud offerings. But until that day, you are stuck travelling from system to system. At least for the next 2 years.
Understanding this is pivotal in your search for the right solution that can add the value that is available on paper. Don’t disprove the power of quantum because you have a bad system integrator. Some applications are just to valuable to door a poor integration job on. If you have such a challenge, reach out, and we’ll help you get going.
