Are quantum computers actually becoming useful?
Probably not yet … but we’re making progress and gaining confidence they will, one day
Earlier this month, Q-CTRL released a preprint describing research that used an IBM quantum computer to calculate the behaviour of a quantum mechanical system, claiming to show a significant speed up in the time taken for the calculation compared to a conventional computer. While bearing in mind that Q-CTRL have a commercial interest in providing useful software for quantum computers, this particular claim is worthy of discussion about what it really does and doesn’t show.
What’s the problem?
While quantum computers won’t magically speed up current computing by doing calculations faster/better/cheaper, they are expected to be able to enable certain types of calculations that conventional computers struggle with. One well known example is breaking certain types of encryption, but another is in simulating quantum mechanical systems themselves.
The problem is that today’s quantum computers are relatively small scale and noisy – hence they are often called NISQ (Noisy Intermediate Scale Quantum) computers. A lot of funding and effort is going into ways to scale them up to be more useful, while reducing the errors in their current operations to reach “fault-tolerant” operation. In the meantime, it is an open question whether we can do anything useful with them today by working on smaller scale problems and using error suppression techniques to reduce the impact of noise.
What did they show?
This research looked at a one-dimensional “Fermi-Hubbard” model – an idealised model of how quantum mechanical systems interact in certain types of materials, that physicists are interested in studying. It turns out to be difficult to calculate what happens in this model in many scenarios, and trying to do the calculations on a conventional computer becomes very time-consuming and expensive as the size of the system grows.
In this case, a quantum computer was programmed to simulate what happened over time when one electron was removed from the system. The researchers found a way to implement this efficiently on a NISQ computer, while also being able to suppress the errors to a manageable level, so that the final answer was reliable and accurate. They showed that the quantum computer could give the same results as the conventional computer, but in a much quicker time – at the largest reliable simulation scale the quantum computer completed in around 2 minutes while the conventional computer (a high performance 32 CPU compute cluster) took around 100 hours.
Is this quantum advantage?
Experts in the field often talk about achieving “quantum advantage” to justify the billions of dollars being invested into developing quantum computers. This definition is contested, but in its purest form, this means showing a quantum computer can do something better than the best possible conventional computer ever could. This is very difficult to prove today, as current quantum computers are still developing, and we will actually need very big quantum computers to show a major advantage over conventional ones. Meanwhile, the hardware for conventional computers and the algorithms that are run on them continue to improve.
However, a more relevant question today might be whether there are cases where the best results we can get on a quantum computer are better in some sense (either faster or more accurate) than a classical computer. While the authors of this paper are careful to avoid overhyped claims (they note that future advances in conventional computer performance and/or algorithms could allow it in the future to outperform the quantum one), this research does appear to show an advantage in terms of the actual time taken to calculate the result using a contemporary NISQ computer, compared to the best approach using a conventional computer that we have today.
How useful is it?
Although the Fermi-Hubbard model is a theoretical construct, it is something that researchers are interested in simulating. Researchers are sufficiently interested that they spend some of their limited supercomputing resources on calculations of it. This is a step forward from previous claims of “quantum advantage” such as Google’s well-publicised claims of speeding up a calculation from greater than the age of the universe to a few seconds. That was for a problem called random circuit sampling, which even the authors admitted was contrived example to show the quantum computer could something that was hard for classical computers, rather than being a problem of practical interest.
We do need to be cautious – the Fermi-Hubbard model is only really of interest to academic researchers – so whether this really is “practical quantum advantage” as the authors claim is open to debate. However, the speed up that they obtain provides confidence that that with larger quantum computers we could gain an advantage in modelling bigger, more complex systems. This could open up new applications such as in materials science or battery technologies.
If you remember nothing else…
Getting to large scale useful quantum computers may be a long road ahead, and is likely to require long term investment combined with patience to ride out the inevitable disappointments along the way. This research helps to increase the confidence that if we can get there, we will find practical, valuable uses for large scale quantum computers beyond just breaking encryption – most likely in simulating complex quantum systems. However, the first benefits are likely to be for fundamental research, with even bigger systems enabling commercial applications further down the line.
MDR Quantum provides quantum consulting and technical due diligence for startups, enterprises, government organisations and investors. If you’re interested in evaluating potential uses cases for quantum computing or planning pilot projects, please get in touch at www.mdrquantum.com.