How quantum computing and machine learning boost each other

Quantum computing becomes reality

Preparing for a future driven by quantum computing took on new urgency recently when Google announced it had created a quantum computer capable of performing a computation in just a few minutes that would take classical computers thousands of years. The company hailed the breakthrough as the first realization of quantum supremacy -- the moment when quantum computers can solve problems that classical computers cannot.

While some took issue with how Google structured its processing test and questioned the legitimacy of its claim, the announcement is at least a sign of progress in quantum and an indicator of what might be coming. The exponential increase in processing power that is theoretically possible with quantum computing has implications for drug discovery, cybersecurity and general AI to name a few areas.

Quantum computing and machine learning will enable models that reflect complex conditions far better than today's models are capable of doing, Langione said. This will be a particular boon in use cases like financial portfolio optimization, fluid dynamics simulations and material design.

That type of advanced R&D will be the biggest beneficiary of quantum-driven machine learning, said Ahmed El Adl, AI consulting and intelligent solutions leader at Accenture. He said traditional computing does a fine job of powering most of today's common machine learning and AI applications. In those use cases, quantum computing won't add much value.

"We don't need quantum computing for chatbots or natural language processing," he said. "But, when we talk with the R&D organizations about the invention of new products, they tell us we've reached the limits of computational capabilities today. For life-changing applications, that's where we need quantum computing."

Advanced quantum computing could even open the door to general AI, El Adl said. In his view, true intelligence is made up of three areas: learning, knowledge representation and reasoning. Today's AI is essentially synonymous with machine learning, so the first part is covered. But the abilities to generalize knowledge to new situations and understand the contextual meaning of things -- El Adl's last two criteria -- are problems far too complex for today's computers. If quantum computers are able to process this kind of complexity, the implications could be substantial across nearly all industries.

"If we want to make real breakthroughs in AI, we need more computing power," El Adl said. "Once we make this breakthrough, I do not see any industry that will not be positively affected."

Path to business use cases remains cloudy

Still, there are a number of roadblocks to effective use of quantum computing and machine learning in the enterprise. Celia Merzbacher, associate director of the Quantum Economic Development Consortium, said there is a skills gap in quantum computing that is holding back development.

"There's currently a huge mismatch between supply and demand," she said. "There's a huge need for people who have some quantum literacy."

In her view, upskilling the existing workforce is a realistic path for enterprises to build quantum capabilities. She said quantum computers aren't likely to outright replace existing computers anytime soon. Classical computers will likely continue to handle most of an enterprise's workload, with quantum computers stepping in for specialized tasks. Today's data scientists will have to learn how quantum capabilities can enhance the deep learning and machine learning they're already doing and when to tap into these emerging resources.

Langione agreed and said that a hybrid model combining the best of classical and quantum computing will be the best way to bring quantum computing and machine learning together as the technology gets more common.

"All of the important algorithms today and in the future will be hybrid algorithms," he said.