Quantum Computing Takes Emerging Tech Baton from Generative AI

As generative AI becomes an everyday tool in organizations’ arsenals, quantum computing is waiting in the wings to step in as the next transformative emerging technology.

This is happening as the National Institute of Standards and Technology (NIST) is set to release its four post-quantum cryptographic algorithms and researchers are discovering ways to eke out more logical qubits from fewer physical ones, a key step towards scaling and practical quantum computing.

In this Q&A, Konstantinos Karagiannis, director of quantum computing services at global consulting company Protiviti, offers insight into the potential use cases for quantum computing, the quantum internet and the pending “big sea change” impact of NIST’s standards, including on IoT devices.

Enter Quantum: What are you hearing from your clients about the opportunities quantum computing offers?

Konstantinos Karagiannis: Last year, generative AI arrived in a big way and it's pretty clear that that affected a lot of decision-making in what was considered, like emerging funding or approaches or teams, so a lot of projects last year were very AI-heavy.

But now I think generative AI is becoming a standard tool for most companies and less of an emerging tech. I believe that next year, we're going to see more companies interested in true bleeding-edge use cases, which is where quantum lies.

Related:How Will Quantum Computers Help Solve Complex Problems?

There are huge differences between the two techs, but quantum has use cases that are focused on AI—quantum machine learning—and a lot of companies are really interested in those. Quantum machine learning includes things like fraud detection or any binary classification. When you're dealing with massive amounts of data, it's hard to pick out the bad data from the good data, and quantum computers are pretty good at that, such as for credit card fraud or network anomaly detection.

The other area I expect to see explode soon is optimization use cases. If you're a manufacturing supply chain optimizing how a warehouse or supply chain works, anything involved with vehicle routing problems, saving money or not sending trucks out of their way—the traveling salesman problem—traditionally excels at that sort of thing.

Then there's quantum simulation for building better materials, protein folding, modeling and future chemicals. There are a lot of areas there that classical computers choke and quantum can help.

What do you make of the progress being made with noisy intermediate-scale quantum (NISQ) computers?

I wasn't expecting to see logical qubits this year, but for the first time, we have error-corrected qubits. We thought the ratio would be horrendous at first. IBM used to say it would be 1,000 physical qubits to get one logical qubit, which would mean you need tens of millions of qubits to do anything. But some of the results this year are a lot more aggressive than that, down to as low as 7.5 physical qubits per logical qubit, which is amazing. It shows a way better path forward for scalability.

Related:Quantum Computing Can Learn Sustainability Lessons from AI

You have a background in telecommunications. What’s your opinion on what’s being done using quantum in that sector?

There’s a misunderstanding that a quantum internet will fix things and be better than the regular internet but it’s the same paradigm as for quantum computers. It's not going to replace the internet. Any quantum internet will be purpose-built to allow quantum computers and quantum sensors to interact with each other without destroying superposition.

The magic there is you can have a few really good quality quantum computers and network them together in a way that they act as one and we’ve never had that before. We've had grid computing of classical computers but they have a massive overhead, so you might end up just using a small percentage of each computer that's in the grid. Whereas in quantum computing, if you have four 1,000-qubit machines, you have a 4,000-qubit machine.

But that’s also kind of worrisome because we might more quickly be able to do things like crack encryption, which is why post-quantum cryptography is so important.

NIST is expected to release its four post-quantum cryptographic algorithms soon. What can organizations expect?

All the unofficial word I've heard is that they're completely drafted, they just need to go through the signing processes. We knew all along that NIST would be putting them out this year and July has been the month that typically they've been hitting us with previous rounds.

The White House has telegraphed what would be coming. A couple of years ago they released the famous National Security Memorandum 10 on promoting U.S. leadership in quantum computing while mitigating risks to vulnerable cryptographic systems.

It describes what federal agencies are going to have to do once the NIST finalists are published. However, you can read between the lines and know that regulators and the private sector are not going to spend tens of millions of dollars to research what they should do. They're just going to pretty much cut and paste.

Organizations are going to have to have a technical inventory and have to have plans for migration by a certain year. They’re going to have to realize that the path to deprecation also begins then. Whether a company believes that a quantum computer that can crack encryption is 10 years away, or 100 years away—I promise you it's not that far away—once ciphers become deprecated, it's your problem.

You’re going to fail penetration tests, you could fail audits, you're not going to be compliant. It's not even a matter of will quantum computers become powerful enough by that date, it's just going to have to be the migration that everyone does. A big sea change is coming.

What will the impact be for IoT devices?

Some IoT devices have lower performance and lower bandwidth. The first IoT devices were a joke security-wise, they had hard-coded passwords in the firmware, crazy stuff like that. They've gotten better since then, but they still are, by design, lightweight.

You've got SCADA infrastructure and shadow networks that are alive and well and I don't know how easy it will be to upgrade those. They could introduce all sorts of potential attack vectors in the future.

But typically, the PQC that we're looking at right now is going to be pretty performance friendly. If you look at CRYSTALS-Kyber, the performance numbers are not that bad, and they seem to outperform what we have now, so that was a pleasant surprise.