Quantum Matters: Quantum & AI – Early Days For A Killer Combination

Guest Post by Karina Robinson

Artificial Intelligence (AI) suffers from two massive blocks: colossal, costly energy use and a transparency deficit. The technology can be technically feasible but is still too expensive for most organisations to consider it. That was the conclusion of a recent paper from MIT Future Tech, which looked at computer vision tasks as an example of an AI-enabled technology. But the MIT group isn’t alone in highlighting the problem.

The computing cost of Deep Learning is exploding. Sam Altman, CEO of OpenAI, made it clear last year that training ever larger Language Learning Models (LLM) is not the way to advance AI, not least because teaching GPT-4, its latest product, cost over $100m, while by 2027 it is estimated that the AI industry could consume as much power as a country the size of the Netherlands.

On the openness front, the models are far too opaque – usable in consumer applications but a legal minefield for companies to consider rolling out. Their tendency to ‘invent’ plausible facts is also not helpful.

Quantum is the route through which AI’s limitations can be lifted.

Two perception issues are delaying the advance. Firstly, there is an impediment to AI/Quantum cooperation based on misapprehensions that quantum is only about hardware – creating a quantum computer with enough power to break current encryption. That is unlikely to happen for several years. It may take up acres of media space, but much more advanced are quantum sensors, some of which are already in the market, while in quantum communication the Chinese are apparently more advanced, and quantum software/quantum-inspired software is advancing at pace. All of these are based on quantum physics and applicable to AI in different ways.

The second issue is the silo mentality of many of the companies involved in these fields, who have separate divisions for AI and Quantum, or only concentrate on one. Nevertheless, more visionary firms are breaking through the barrier.

Scott Faris, CEO of US firm Infleqtion says, “The convergence of AI and Quantum is one of the most powerful combinations that we are starting to unlock. The convergence will have both immediate and long-term implications.”

Karina Robinson is Senior Advisor to Multiverse Computing and Founder of The City Quantum & AI Summit

The firm, which manufactures quantum products and parts, ranging from sensors to computer hardware, counts NASA as one of its clients. Faris points out that quantum-enabled technologies are “quickly demonstrating their utility in addressing the crushing data infrastructure scaling challenges driven by AI. Scaled networks of quantum sensors will create vast new data sets of unparalleled precision and value which will be unlocked by parallel advancements in AI.”

A case in point is CompactifAI, the product launched late last year by Multiverse Computing*. Europe’s largest quantum software and quantum-inspired software firm, which counts Bosch and the Bank of Canada among its clients, uses its technology to compress the data from a Large Language Model (LLM) by up to 70%, thus using much less computing power, and achieve results that are comparable in quality.

“Left on its own, AI is going to burn the world by consuming intolerable levels of energy,” says CEO Enrique Lizaso. His firm, shortlisted as one of three finalists in the European Future Unicorn Award, is using its AI and quantum-inspired capabilities in fields ranging from forecasting weather catastrophes – on the increase with global warming – to helping car manufacturers in the training of their Machine Vision.

This is done on the premises of the industrial site, rather than data processing centres, with faster retraining of the multiple streams of data, reduced processing power requirements and added security.

Lizaso is adamant that the cross over between AI and Quantum is environmentally helpful in other ways. He notes that optimising routes for shipping, for instance, or optimising the amount of fuel tankers need for a journey, cuts back on the carbon footprint of the ship.

Nvidia, best known as the AI chip market leader, is also keen on quantum.

“AI is accelerating quantum computing today. We’re starting to see researchers tap into the mature infrastructure of AI and accelerated computing to leverage things like Large Language Models (LLMs) to develop new quantum algorithms and improve the performance of quantum computers,” says Tim Costa, who leads the HPC and Quantum Computing Product Team.

“We are just starting to scratch the surface of how Gen AI can improve quantum computing,” he adds, noting however, that in the near term researchers are already investigating quantum machine learning (QML) and quantum-inspired methods for financial applications like fraud detection and forecasting.

Recently, researchers from the University of Toronto, St. Jude Children’s Research Hospital and NVIDIA developed a new quantum algorithm called the “GPT-Quantum Eigensolver.” It uses the framework of generative AI models to generate quantum circuits with desirable properties, in this case to calculate the ground state energy of molecules of interest. Versions of this generative quantum algorithm can be applied towards important problems in drug and new materials discovery, as well as a host of other applications, some of which we cannot even imagine.

As for the problem with the unclear thinking process of AI, and its fantasising, quantum is also part of the answer.  To use it more widely, the interpretability of the system is key – in essence understanding why a system makes the decisions it does so it can be held accountable. Only last week Quantinuum, formed from the merger of Cambridge Quantum and Honeywell Quantum, announced a first public step in creating AI that is “interpretable and accountable” via the development of a framework for compositional models of AI using a type of maths called category theory. Their academic paper has yet to be peer reviewed.

Humankind’s biggest challenges will not be solved tomorrow by the combination of Quantum & AI. Nevertheless, at the risk of creating a hostage to fortune, I would predict many will be solved in the next decades by those firms at the forefront of the Quantum & AI journey.

*Karina Robinson is Senior Advisor to Multiverse Computing and Founder of The City Quantum & AI Summit which takes place on Monday, October 7th.

Will quantum computing, not AI, define our future?

The interaction between quantum and AI is the most exciting development in modern-day technologies. The hype cycle is focused on AI, but in the words of Lord (William) Hague, speaking recently at Imperial College: “Quantum computers can provide AI with the computational firepower needed to unlock their full potential. This means that if the country can lead on quantum computers, we can secure a lead on AI. And if we fall behind on quantum, we will likely fall behind on AI.”

The essence of quantum is that it isn’t binary – even today’s supercomputers are – but instead allows for multiple possibilities at the same time, making it ideal for optimisation problems. And the UK has the second largest number of quantum start-ups in the world, plus a £2.5bn quantum strategy.

There are two misapprehensions about quantum. The first is that it is prohibitively expensive because of the construction costs of a quantum computer and the environment in which it needs to be kept. In fact, companies can access time on one via the Cloud, with Microsoft Azure and AWS, among others, offering it as a service used by banks like HSBC and JPMorgan.

Second, quantum is only about computing, with the goal of creating a quantum computer within five to ten years that can break current encryption. In fact, quantum and quantum-inspired algorithms are already being used on classical computers, and the rather basic quantum computers – think of your first iPhone – that we have today.

Additionally, there is a large product suite associated with the technology. Quantum sensors, for instance, are one of the most exciting developments in the field of quantum physics. They are a magnitude more powerful than current ones in identifying minerals underground or coin-sized holes in undersea gas pipes and are likely to be commercialised by companies like Bosch within the next two years. And there are other quantum technologies being worked on in parallel, such as a quantum internet, all of which provide venture capitalists with investment opportunities.

Quantum in the finance sector

The financial services industry is already experimenting with quantum and machine learning.

In CB Insights’ recently published list of 100 Top 100 AI Companies, only two quantum companies are mentioned: Sandbox AG and Multiverse, one an American spin-out from Google, the other a European medium-sized start-up with UK offices.

Multiverse, which I advise, has worked with the Bank of Canada on cryptocurrency scenarios; with Credit Agricole CIB on counterparty credit rating downgrades and valuations of derivatives; with Spanish bank BBVA on maximising portfolio returns for a given level of risk, and a host of others.

Marco Pistoia, the managing director of JPMorgan Chase’s Global Technology Applied Research Center, is very clear on why the banking behemoth is seriously investing in quantum: it promises dramatic speed-up and accuracy improvements in optimisation, simulation and machine learning. Finance is awash with exponential complexity, and classical computers cannot deal with big datasets, let alone with the time constraints usual in the industry.

Quantum machine learning will also be useful in predicting financial crisesHSBC, meanwhile, recently announced quantum computing projects in cybersecurity and fraud detection, and became the first bank to join BT and Toshiba’s quantum-secured metro network (QSMN) to protect against cyberthreats. The QSMN connects the Canary Wharf headquarters with the global bank’s data centre in Berkshire, and will be used to trial experiments such as secure video communication and financial transactions.

Quantum machine learning will also be useful in predicting financial crises. Financial firms and regulators have access to large amounts of data, but analysing them intelligently within a usable time frame requires more computing power and AI than is currently available.

Consultancy Boston Consulting Group estimates that at maturity (projected to be around 2035), quantum technology can create US$450bn to US$850bn in net income for end users through a combination of new revenue generation and cost savings.

William Hague recently wrote a couple of reports with Tony Blair, urging the Labour and Conservative parties to make leadership in science and innovation, and AI especially, the New National Purpose. As he points out, quantum is a keystone technology. “You don’t need to understand how quantum works to understand how quantum will revolutionise our world,” were his parting words at the event at Imperial.

The mind-boggling possibilities of AI using quantum as its underpinning technology is one of the reasons I founded The City Quantum and AI Summit at the Mansion House. The annual event, now in its third year, brings together the City and the quantum community, with three principles: free for all; no jargon, no lingo, only understandable language; gender-balanced panels.

We welcome all of the CISI community to the Mansion House on Monday 2 October in a bid to ensure the City leverages its world-leading role by adopting this world-changing technology.