Quantum + AI: A Future
It was both great fun and a great honor to see the program unfold that I designed for the two-day IQT “Quantum + AI” (Q + AI) event in New York last month. Was I a modest person, modesty would have prevented me from saying how great the speakers at the conference were. But I’m not modest. So here goes. We had Amazon, Amgen, IBM, IonQ, JPMorgan Chase, Microsoft, Multiverse Computing, Rigetti, SandBoxAQ, SKT Telecom, Mayo Clinic, Moody’s Analytics, Quantinuum, Thales, Vanguard, VTT and others with global reach. And, of course, we had our fair share of innovative quantum + AI startups. No conference of this type would have been complete without NVIDIA (www.nvidia.com), which straddles the world of quantum and AI so effectively and at the IQT show Elica Kyoseva, NVIDIA’s Director of Algorithm Engineering, spoke to “Quantum-accelerated Supercomputing at NVIDIA.” Incidentally, Q+AI New York will be back in 2025 on October 27-29.
The Q + AI space seems full of opportunity right now and this is what this edition of “From My Desk” is about. Trying to be objective here, it could go either way. Perhaps Q + AI will turn out to be what has been called (in an entirely different context) a “luxury belief” – an idea that let practitioners show off how clever they are but ultimately not rooted in real market needs and requirements. Who knows, perhaps there is yet another AI Winter in our future and one that will pull the quantum sector along for the ride. Given the extravagant, breathless hype about AI in the media these days this sometimes seems inevitable.
That said, the talks and panels focused on Q + AI applications at the IQT event included three that I think present genuine new opportunities for Q+I. The three that I am thinking about are (1) Using Q + AI in materials design, (2) using AI to help design and optimize quantum hardware, and (3) the rising area of quantum + AI security, which at present doesn’t get the attention it deserves. Each of these three topics deserves its own “From my desk,” but for now here is some vorspeisen.
Q+ AI + Materials Design
The use of Q + AI in materials design is overshadowed today by Q + AI for designing drugs. They are closely related topics. This is understandable – curing cancer trumps new automotive nanocoatings any day on both social and PR grounds. On the other hand, one recalls that nanocoatings were the one success story from the nanotech boom of yesteryear. So still worth paying attention to.
At the IQT event in New York, Arman Zaribafiyan, Head of Product, Simulation Platforms at SandboxAQ talked about the considerable potential for Q + AI to create a “Renaissance of Materials Innovation.” So, when I got back home to Virginia, I re-read a report I wrote for a client that discusses how Q+AI was being applied to new materials design to see what I could see. Among the companies that I found that are in the report and that are applying quantum + AI to new materials were GenMat (photovoltaics, superconductors and batteries). GenMat has recently been acquired by Comstock, which is focused on repositioning the product range for decarbonization. OTI Lumionics is a company using quantum machine learning (QML) to develop advanced materials for OLED displays but it’s also working with Nord Quantique developing materials research with applications in the semiconductor and specialty industries.
There are already some big-name firms invested Q + AI + Materials Design. Accenture Ventures has invested in Good Chemistry, which has recently become part of SandboxAQ. Good Chemistry has been developing Q + AI to accelerate new materials design, based on its QEMIST Cloud platform. Quantistry has similar goals and has announced a successful €3-million funding that included a contribution from BASF.
How Q + AI Helps Fabricate and Optimize Quantum Hardware
A somewhat more esoteric meaning for Q + AI is all about using AI to speed up the development of quantum computers and quantum algorithms. AI can, for example, be used to model the error patterns in qubits. In this way Q + AI may take quantum computing beyond today’s NISQ realities.
Another way that Q + AI in this sense could prove useful is in quantum benchmarking. It can, for example, run benchmarking tests comparing the results to benchmarking metrics or formal standards.
Two companies that spring to mind that fit into this category are Mind Foundry and QuantrolOx. Both have strong ties to the University of Oxford. Mind Foundry uses machine learning (ML) to calibrate qubits in quantum computers and says that its approach can result in a 100x speed increase. Then there is QuantrolOx which uses ML for automated characterization and tuning of qubits in quantum computers. QuantrolOx – we had the CEO, Vishal Chatrath, speak at the IQT Q + AI event -- is building automated control software for quantum technologies to tune, stabilize, and optimize qubits. Specifically, the kind of ML that is being used by QuantrolOx is based around Bayesian optimization.
Q+ AI + Security
In New York there was also a panel on “Quantum AI and Cybersecurity” moderated by Reza Azarderakhsh, the CEO of PQSecure with representatives from Qrypt, QuSecure, Thales, and West Point. It’s a big topic and, to be honest, it didn’t occur to me when I first put the event program together that Q + AI + Security was a special thing. Of course, Q + AI must cope with all standard cybersecurity issues, plus the current rush to comply with PQC. But it also has its own special security and privacy concerns. This is especially the case because so much of Q + AI currently is focused on banking/financial services and healthcare. These are sectors where privacy and security are at a premium.
Security issues in Q + AI can threaten both training and inference phases of the AI component Specific vulnerabilities that can be found in AI models include in the model architecture, training/testing data, encoding techniques, and trained parameters. One key motivation in the case of bad actors in the Q + AI space is to be free riders. Thus, they can benefit from not having to pay for expensive quantum computing time and the millions of “trials” in a training run. Developing state-of-the-art AI models inevitably require millions of dollars so AI model thieves can be on to a “good” thing
The free rider aspect of security in Q + AI also extends to time. Bad actors might steal models and data not just to save large amounts of money, but to get a leg up in terms of getting products to market. Thieves of this species kind can end route around the significant amount of time — potentially months to years — that it takes to construct an AI model. The vulnerabilities of Q + AI also extend to intellectual property (IP), which may be of especial interest at this early stage of the game when startup valuations depend so much on IP. QML IP may relate to various aspects of the model and its relationship to quantum computing including such things as entanglement strategies, the number of parameters, the neural network layer depth, and the measurement basis.
Practitioners should also worry about their cloud providers. Cloud providers that are inexperienced with particularities of Q + AI can make mistakes. And yet another concern in the world of Q + AI is is ransomware attacks where malicious adversaries can encrypt sensitive data or computational resources with demands for payment for access restoration.
Where to from Here?
Or perhaps we are on a path to something wonderful. Recently coming out of stealth mode is
Nirvanic Consciousness Technologies (https://www.nirvanic.ai/). This company appears to be in business of taking Q + AI to the next stage in terms of simulating human consciousness with a better approximation than the current generation of AI to displaying intuition, empathy, and nuanced judgment. Assuming it is successful I can see how what Nirvanic is aiming at will add an entirely new dimension to AI products and I wish them very good luck.
It is worth noting however that what is being proposed by Q + AI may well be something less an actual consciousness. Thus an article in Quantum Insider references both computational consciousness and the Penrose-Hameroff theory of quantum consciousness. However, I note that Sir Roger Penrose one of the great physicists of the late 20th century, has been at pains to point out that his quantum consciousness theory is specifically not a computational theory of consciousness. As far as I can see, however, there is nothing on the Nirvanic Website about Penrose-Hameroff or their thoughts. (See https://www.newscientist.com/article/mg25634130-100-roger-penrose-consciousness-must-be-beyond-computable-physics/)
As I have already noted, IQT’s extravaganza will be back in 2025 with all the big names. For those that are a little more philosophically I also recommend the MindFest conference down in Florida (https://www.fau.edu/future-mind/files/mindfest/mindfest-2025-for-the-website.pdf). Hameroff and David Chalmers were there in the flesh this year, so a very different look at feel to my IQT event, but well worth attending.
In the next “From my desk” I will share some thoughts on the role of hype in new technologies such as AI and quantum. There are some good things to say about it.