FutureThink: Hybrid Quantum Computing
When we speak of hybrid quantum computing today, we assume we’re discussing the blending of a classical and a quantum computer. These two systems can be configured, thanks to orchestration, to work together to solve a computation. The assumption is that by working together, the hybrid quantum “system” will deliver better results than either can on its own.
If quantum computers are so powerful, why would we need hybrid quantum systems? One reason is that classical systems process certain types of information much faster and effectively than a quantum computer ever will. A hybrid quantum system lets classical computers continue to do what they do well, at a much lower cost than quantum computers.
The other highly relevant reason is simple.
Quantum computers do not scale to process the large data sets required by complex computations, for today and in the near future.
Certainly, larger and larger quantum computers are on the horizon. In fact, many would say the hardware vendors are accelerating their scale much faster than was expected. This may be one reason we tend to visualize them as massive supercomputer-like processors. But that’s not true.
Which brings me to my question for this post.
Why are we viewing today’s quantum computers as if they were the next generation versions of large scale supercomputers?
Seems to me the time has come to see these next generation computers for what they are today, and to find ways to apply and use them to deliver value. Not as massive supercomputer wannabes, but using the capabilities they deliver today.
A Potential for Hybrid Quantum Systems
What if we looked at hybrids not just as a single classical and quantum computer together?
What if, instead, we looked at a hybrid quantum system that includes multiple quantum computers, as well as classical, working together to solve problems in a way that delivers more value than a single QPU can provide today? Using different quantum computing paradigms and architectures such as superconducting, photonics, ions and more?
I know, you’re thinking that each quantum computer type requires specific low-level coding for software and algorithms to work. That would mean you couldn’t share software or computations across different vendors’ systems.
Here’s the thing. Qatalyst eliminates that barrier. Its micro-services run and compute across diverse QPU and CPU hardware, without any need for specific vendor code – meaning no vendor lock in. The same exact problem submission(s) are processed across diverse QPUs and CPUs, seamlessly.
Since I can submit the same problem to many QPUs and CPUs, why wouldn’t I use these diverse processors in a way that better matches their current capabilities? And that can deliver even more value in the future?
An Example
Let’s say you have a logistics optimization problem with 40 delivery destinations. When that problem is computed, it results in ~ 815,915,283,200,000,000,000,000,000,000,000,000,000,000,000,000 options that must be analyzed. That’s a pretty big computation.
Now, imagine you break down that computation via distillation or other mathematical methods, to create “model” problems that today’s QPUs can process.
You then submit each of these models to a variety of QPUs and CPUs if you want. Each processes the computations, iterating and sharing best results to continuously improve the optimization.
You then collate and analyze the results across all of the systems, identify the best possible diversity of results and return them.
You have diverse processing approaches, since different QPUs and CPUs tend to process different problems in different ways, so you get a range of computational methods and results.
You can optimize each QPU’s computations with results from other QPUs, improving the computational results thanks to the heterogeneous nature of processing.
You can send more model problems to more machines, which means you get a better set of results thanks to that diversity and expanded modeling.
Simultaneously, you can review the diverse QPUs to learn which architecture best maps to the specific problem computation you’re running. You get insights into how you’ll get your best possible results for those types of problems going forward.
The Bottom Line
I am not suggesting that we all rush to build hybrid quantum systems. I am suggesting that in quantum computing, as in any advanced technology market, it always pays to think differently.
Hybrid quantum systems are one place to apply even more innovative thinking as we explore ways to accelerate our exploration of quantum value.