JPMorgan Chase has been stepping up its efforts to be quantum ready – in other words, prepared to take advantage of quantum computers when they become powerful and reliable enough for banks to use.
“The financial industry would benefit 100% from quantum,” said Marco Pistoia, global head of quantum computing at JPMorgan Chase, in an interview. “We think it’s going to be the first industry sector to benefit from quantum.”
One reason for this is because in finance, unlike some other domains, time is of the essence, he said: “We need to do things in real time because the market changes constantly and very quickly.”
Other banks have also been working on quantum computing readiness.
These firms are putting resources and effort into a technology that may be three to seven years from practical reality. They are betting that they will be poised to reap benefits the technology promises and cope with threats it poses.
Quantum computing in banking
The first use case for quantum computing in finance, in Pistoia’s view, is optimization, such as in portfolios. It could be applied to natural language processing to automatically create document summaries, maximizing the sentences with the highest meaning and reducing the sentences that are redundant.
Other optimization problems quantum is suited to solve include anomaly detection – for instance, to detect cybersecurity attacks and fraud attempts – and delivery pricing.
A big fear that many people have is that cybercriminals will get their hands on quantum computers and use them to break the RSA encryption that’s universally used to protect sensitive data in servers all over the world – including in banks’ computers.
“The next-generation risk in terms of hacking to the system could be AI,” said Brett King, founder of Moven, Breaking Banks host and futurist, in an interview. “But actually the bigger threat is quantum computers. We’re still five to seven years away from a hundred thousand qubit quantum computer that has enough error correction in it to be reliable. But at that point, we anticipate Q Day,” the day quantum computing advances to the point that it can break the encryption methods safeguarding most of the Internet.
“There is no doubt that this is a risk and it’s actually real,” said Pistoia. “Some people disagree on when this will happen, but nobody disagrees that this will happen.”
JPMorgan Chase is working on adopting solutions to this problem.
“If you’re leaving for a vacation, you’re definitely going to lock your front door,” Pistoia said. “But why wouldn’t you lock the windows upstairs just in case somebody uses a ladder? Why wouldn’t you set the alarm system if you have one? So it’s very important to have layered defenses.”
One solution the bank is putting in place is post-quantum cryptography, in other words, new cryptographic algorithms designed to resist quantum computing attacks. It is also working on quantum key distribution – the use of quantum technology to allow two people to exchange the same cryptographic key in a way that is impossible to break.
Getting ready for Q Day now “gives us time to solidify our infrastructure to become resistant,” Pistoia said.
Fraud detection is another strong use case, given the vast amounts of transaction data that have to be analyzed and identified as fraud or not fraud, in real time.
“When somebody’s executing a transaction with their credit card, you cannot be notified that this was fraud 24 hours later, you have to know immediately that the transaction has to be blocked,” Pistoia said.
The reason it’s so hard for classical computers to catch transaction fraud is because most transactions are benign, swamping the data with innocent purchases, Pistoia said.
“When you do machine learning, you don’t have enough data to learn what is a fraudulent transaction,” Pistoia said. “The data is imbalanced, so it doesn’t have enough evidence of what constitutes a fraudulent transaction. You have a lot of information about what is a good, benign, genuine transaction.”
This helps create false positives, where a card or account is blocked even though a transaction was not fraud, and false negatives, where fraud is not blocked.
Quantum computing allows engineers to increase the feature space used in machine learning, the set of all possible values for a chosen set of features from that data. Features of transactions include the location, the time of the day, the day in the year, the address where an item is being shipped.
“By putting together all these features, you can now start to understand what is a fraud,” Pistoia said. “In a quantum computing setting, we’ll have enough space to include all these features and then we will actually start to understand better what becomes a fraud because we have more ability to encode all these features.”
Another use case for quantum computing will likely be Monte Carlo simulations, mathematical models that predict possible outcomes of an uncertain event, such as the performance of a collateralized mortgage obligation.
“They’re pretty notorious for taking basically all night to run,” said Konstantinos Karagiannis, director of quantum computing services at Protiviti, in an interview. “It would be much better to walk in in the morning and run one with new data in 20 minutes. So those kinds of speed-ups are very exciting.”
Getting quantum ready
Quantum computers are not ready for finance yet. For a financial problem like portfolio optimization or delivery pricing, 56 cubits are not enough; it would require hundreds of qubits and a low error rate, Pistoia said.
But JPMorgan Chase has a team of engineers who have been developing, maintaining and documenting quantum algorithms for financial applications. The bank aims to put them into production as soon as quantum computers are powerful enough to run them.
“We have to be patient, because we’re waiting for the moment to come, but the fact that we see results happening, is a big encouragement that gives us a lot of confidence in the future,” Pistoia said in an interview.
“The folks at JPMorgan have a front row seat to understand this,” said Jenni Strabley, head of the hardware team at Quantinuum, in an interview. “And those people in the front row seat are going to be poised to take advantage of it much, much quicker than people who have chosen to take a more wait-and-see approach.”
In June, the bank and quantum computer manufacturer Quantinuum demonstrated the use of a quantum computer handling a random circuit sampling problem, a computational task that involves sampling from the probability distribution of outcomes for a given quantum circuit. According to the two companies, the Quantinuum machine performed 100 times better than Google’s Sycamore quantum computer did in 2019 when it completed a task in 200 seconds that would take a state-of-the-art supercomputer 10,000 years to finish and declared “quantum supremacy.” Quantum supremacy is when a quantum computer can complete a task that would be impossible for a conventional computer to handle in a reasonable amount of time.
Google’s declaration of quantum supremacy was controversial. IBM technologists critiqued Google’s analysis and said Sycamore’s performance did not actually surpass that of classical supercomputers.
The term “quantum supremacy” “sends the wrong signal,” said Scott Crowder, vice president, quantum adoption and business development at IBM, in an interview. “Quantum computers are not going to reign supreme over classical computers.” Many deployments are and will continue to be combinations of classical and quantum computers.
But some see the Chase-Quantinuum work as a breakthrough.
“Anyone who doesn’t praise JPMorgan Chase for what they’re doing has some kind of ax to grind,” said Konstantin Karagiannis, director of quantum computing services at Protiviti. “I can’t think of a major company that is that active. They’re a leader by far.” The fact that Quantinuum’s system uses 56 fully connected qubits is significant, as is its high reported accuracy, he said.
“That combination means we’re now in uncharted territory,” he said.
IBM and D-Wave are also doing interesting work, Karagiannis said. IBM published some of its quantum computing results in the journal
Another advantage these banks are looking forward to is the lower power consumption of quantum computing. If a classical computer had done the random circuit sampling in the JPMorgan Chase/Quantinuum test, it would’ve consumed 30,000 times more power than the quantum computer took, according to Strabley.
The testing and prep work JPMorgan Chase is doing is important, Karagiannis said.
“It’s stuff that more companies could be doing now,” he said. “Not everyone has that kind of budget.”