Quantum Computing, Machine Learning Perform Better Together

AI and machine learning are enjoying a surge in technological advances and funding but quantum computing is coming hot on their heels.

In its role supporting businesses to best adopt emerging technologies, professional services multinational Deloitte has recognized a significant opportunity for the technologies to work hand-in-hand.

In this Q&A, Deloitte global quantum computing lead Scott Buchholz discusses how quantum computing and machine learning can best work together and whether businesses are better prepared for quantum computing than they were for AI.

What have been your observations about the convergence between quantum and machine learning?

We're seeing increasing levels of capability in what we call quantum machine learning. There are a handful of things that can already enhance the machine learning algorithms that some organizations run today.

As we look to the future, what we're seeing in the quantum machine learning space is that when we're creating machine learning models, they seem to be able to train to higher accuracy with less data than the comparable classical models.

That becomes interesting because there are problems that people care about in areas of advanced machine learning, which that enables in different ways. One of those instances is in the space of generative AI, which includes large language models but also a lot of other things like generating images for medical applications and a variety of other synthetic data.

Related:Deloitte launches AI Institute to advance innovation in the enterprise

There's an approach called a generative adversarial network, which is a big word for saying I have two machine learning algorithms that I pit against one another. One's job is to figure out if something is real or fake. The other’s job is to do a better job of generating fake data.

You run the two of them in parallel for a while and the generator gets better at generating fake data over time so you can use that over time to generate synthetic data. It turns out that doing this classically is really difficult, but because of the way quantum works, people believe, as near as we can tell today, that it's going to be much more effective to do in quantum computers.

That becomes exciting because that might mean we can apply some of these things in new and different places. Then the question becomes, are we going to run our large language models — ChatGPT cloud AI, you name it — on quantum computers anytime soon? Probably not.

On the other hand, there is a universe of machine learning that we use to detect fraud or identify and then prioritize what to triage that may well be enhanced.

Related:Quantum-Hybrid AI Platform Funded for Japan

A recent Omdia report found that governments seem to be backing quantum efforts after having missed the boat on AI. Does that reflect what you hear?

I believe so. Two things are true about AI. One is that what we call artificial intelligence keeps moving every couple of years. I remember a decade ago, the idea of being able to have a conversation with your phone was science fiction and now we call it Siri or Google or Alexa or what have you. It's no longer AI because it's everyday functionality.

AI has been around since the 60s as a concept, but the bar for how we define it keeps moving. Part of the challenge with AI is that people haven't necessarily had something that focuses the mind in quite the same way. When big things happen, people go, oh, we need some of that but we’re relatively behind. That's part of what you're seeing in AI.

In quantum, there are a couple of things that focus minds and as a result, it's easier to see. You've got the dynamic of we feel like we missed out on the last one and there's a bigger outcome here. That combination of things and the perceived fuzziness drives different behaviors.

This is an extract from an article first published on AI Business’s sister site Enter Quantum.