Demystifying What Quantum Computing Can(not!) Do, Part II
Addressing the highly-practical question of "What makes a good use case for quantum computing?".
In a prior post, we explored the theoretical limits to what quantum computers could do, by looking at various computational complexity classes and the relationships among them. We saw how the set of problems efficiently solvable with quantum computers exceeds the set efficiently solvable with classical computers. In addition, we learned that quantum computers cannot efficiently solve all problems.
For refresher, see here:
After reading that post, you might be wondering how to make it practical – that is, what can quantum do for your problems? This is a great and often-asked question, though the answer might surprise you.
If you ask me “Travis, is quantum computing good for use case X, Y, or Z?”, the answer I tend to give is “Probably, though the exact ways quantum could impact your use case are going to depend on a variety of factors.”. The reason for this answer is simple: quantum computing is not (yet) a widget or a solution, it’s a tool which can be used to re-architect and re-imagine business. The far-reaching implications of this tool aren’t yet fully understood, meaning well-defined bounds on the question of “Is this a good use case for quantum?” don’t really exist.
The far-reaching implications of this tool aren’t yet fully understood, meaning well-defined bounds on the question of “Is this a good use case for quantum?” don’t really exist.
Phrased another way, having access to quantum computers is a capability businesses can deploy as part of an effort to transform how they operate. In this sense, quantum probably isn’t all that different from AI, hybrid cloud, or any other advanced, cutting-edge technology you are looking at. And if you consider the early days of the introduction of computers into the workplace, how we use computers today to do business probably looks quite different than what people originally envisioned.
So if you’ve been tasked with figuring out what all this “quantum stuff” is about, what should you do? Instead of asking “Is quantum computing good for X?”, ask “What goes into formulating a compelling use case for quantum?”.
Allow me to suggest 4 desiderata for you to consider.
Desiderata for a compelling use case in quantum
Adopting quantum computing isn’t akin to going out and purchasing some “as-a-service” solution. Because of how early we are in the quantum computing industry, figuring out how quantum impacts a business takes time and investment. These desiderata can help you make a good investment.
1. The use case is critical and challenging to your business.
You might think this is obvious. However, even though the barrier to entry for adopting quantum is lower than ever before (and decreasing), it still exists. So before investing in figuring out how quantum could be incorporated into your business, you must identify use cases which actually matter to your business, so you can formulate how you’ll quantify return on that investment.
At this stage in the industry, “return on investment” doesn’t mean “increased revenue/profits/etc.”. We have yet to see the first true commercial application of quantum computing where this technology has been deployed in a production or enterprise-scale environment. As a result, you’ll need to figure out what kind of ROI would be meaningful to the stakeholders who need to sign off on this. Consider that your quantum journey isn’t a days or weeks sort of investment – it’s on the order of years. You’ll need buy-in to see this through.
2. The key performance indicators/metrics/etc. for the use case are known.
You need to know how “performance” for the use case is currently defined. Knowing these helps guide the team working on this to correctly balance various tradeoffs when considering possible ways to tackle the use case with quantum. It also establishes the framework against which to evaluate a quantum-based approach.
For example, “Faster time to solution” might suggest you look at quantum algorithms with either a provable speedup, or see where there are quantum-based heuristics which generate comparable solutions to your current methods in a shorter time frame. Similarly, “More accurate solutions” suggests prioritizing accuracy over runtime.
Knowing the performance indicators helps you figure out what things you are willing to trade off against in order to make progress against that which matters.
3. Your current approach performs poorly, as measured by those metrics.
This might be a surprising consideration, but look at it this way: if you are seeking incremental improvements for your use case, you most likely don’t need a technology such as quantum computing. You can probably eke out a small performance gain through other means.
What’s more – and going back to point 1 above – is achieving small gains really going to transform your business, and drive stakeholder buy-in? Most likely not. Investigating the use of quantum computing makes the most sense when you are working with a problem or data where, despite really pushing the envelope using non-quantum means, you keep coming up short.
Now, one factor which might alter this conclusion is if your particular problem comes with a variety of instances or data sets of variable hardness or complexity. Then it makes sense to use quantum on both the “easy” and “hard” instances as a way of benchmarking the relative strength of quantum for the use case. In addition, this approach could allow you to identify future data sets where using quantum makes sense.
4. The workflow for the use case has been charted & key technical obstacles are well defined.
You’ll need to understand the various steps and pieces by which the use case is currently handled by your business. The more specific, the better. Why? Because that will give you a variety of possible ‘hook points’ into which you can investigate the use of quantum. Further, this will help whatever quantum team you are working with get a handle on the end-to-end picture for the use case, and can spark interesting ideas on their end. What’s more, this level of specificity is going to help you see what bottlenecks exist, which can also inform discussions about where to use quantum.
Within each step or part of the workflow, you’ll want to identify what technical obstacles prevent improving performance against the previously-identified metrics. Those obstacles should be as well-defined as possible. For example, “We cannot train our classifiers well enough” is a poor definition; a better one could be “We need to increase synthetic training data instances for outlier-type phenomena by 5x.”. By being as specific as possible, you get to the root of what needs to be addressed. What that root actually is (that is, what kind of problem it describes) impacts how you’d bring quantum into the mix.
Consider the difference between “We need to increase synthetic training data instances for outlier-type phenomena by 5x.” and “We want to train our models on 10x the data, but don’t have the compute capabilities to do it.”. The former suggests quantum generative models as a place to look, whereas for the latter, it’s honestly not so clear whether quantum could help (perhaps you need to buy more classical GPUs?).
Wrap-up
Admittedly, this might read like a 4-step program to adopting quantum computing in your business. That’s not quite my intent here. Instead, these 4 desiderata should highlight how figuring out what to do with quantum necessarily incorporates both business and technical acumen. The difficulty and time involved in Step 4 in particular is generally under-appreciated, but it’s a really vital step. I’ve been involved in projects where this step wasn’t done up front, and quite a bit of time in my early engagement with those clients was spent getting clarity on the subject of workflows and approaches. Without this step, you can’t identify a concrete obstacle or issue to go look at applying quantum computing toward. At which point you are simply spinning your wheels.
Additional Resources
One of my current activities at IBM Quantum is tracking the research output of the IBM Quantum Network, a collection of companies, universities, laboratories and startups all doing work in quantum. Most projects from Quantum Network organizations lead to papers, and most are available for free. A database tabulating these papers is available here . If you need help getting a sense of what other companies have been doing with quantum, this is a great place to look (from a technical perspective).
If you need a report from a consulting firm to get a business perspective on use cases, see this.
Travis, well written. This was a great piece for me! And thank you for the new vocabulary word!!! Diserata was new to me. 😁