But here’s what has changed recently in the world of quantum research: Today, quantum engineers can bring generative AI technology to bear. In doing so, they have the potential to accelerate quantum innovation drastically. Problems that previously required years of research and development to tackle could now be solved within months or weeks, with help from AI.
If you ask an AI engineer what the biggest barrier to continued AI technology advancement is, you’ll likely hear about computing power. Practices like enhancing model design and improving the quality of data ingested by AI tools might be factors too. But at the end of the day, compute resources remain the greatest constraint on businesses’ ability to build better models at a faster pace, and improve the speed and accuracy of inference.
There is, however, another way to supply AI with the computing power it desperately needs: Quantum devices. Indeed, businesses planning for AI success over the medium and long-term should be thinking about how the availability of “classical” compute will impact AI capabilities and scalability. But they should also be thinking about how quantum computing might transform AI in ways that create vast new opportunities or introduce huge risks.
How quantum computing could supercharge AI models
For now, quantum-proofing AI strategies remains somewhat theoretical, given that it remains to be seen exactly when quantum practicality will arrive and how organizations will leverage it. But both quantum and AI systems are sufficiently mature at present that we can certainly infer, at least at a high level, which scenarios are likely to play out. Doing so, and taking steps to mitigate the risks, is essential for businesses seeking to use quantum-powered AI to their advantage, while minimizing its potential for harm.
The good news is that it’s possible to get ahead of these challenges. Organizations should consider practices like investing in quantum-ready security controls, protecting source code from third-party access (which makes it easier to find and exploit security vulnerabilities using AI) and leveraging quantum-powered AI to block or mitigate the actions of malicious quantum-powered AI systems.
In that case, the model would presumably be able to find and exploit critical vulnerabilities in milliseconds instead of hours. It may also discover types of risks or exploits that are simply impossible to imagine at present.
Quantum-AI cross pollination
The reason why quantum could help AI overcome the scalability and performance barriers it faces at present is simple enough: Quantum computers can deliver parallel compute power at exponentially higher rates than classical CPUs or GPUs. And because AI models require vast amounts of compute, quantum will allow them to unlock capabilities that seem unthinkable at present.
At present, the chief barrier to leveraging quantum computers to scale AI is that quantum devices are not yet stable and reliable enough to power real-world systems. Critical engineering problems, like correcting for quantum errors efficiently, remain to be solved before quantum practicality arrives. And, given that quantum researchers have spent decades working on those challenges without fully resolving them, it may be tempting to be skeptical that quantum will ever become truly usable in real-world settings.
To date, the standard approach to addressing AI’s need for compute has been to toss more computing hardware (especially GPUs) at AI models. But those devices are costly and in limited supply, making it unclear to what extent AI can continue to scale.
Adapting AI strategies for the quantum era
To put this in real-world context, consider Anthropic’s Mythos model, which, in a matter of hours, famously uncovered more than 10,000 major cybersecurity vulnerabilities that traditional scanners had long missed. Now, imagine what Mythos could do if it had the ability to perform its calculator a million times faster for training and inference – a very plausible scenario in the event that quantum computers become practical enough to connect to production AI systems.
This is notable not just because it suggests that the pace of quantum innovation is likely to speed up significantly, but also because it highlights the directional acceleration effect of AI and quantum. AI is helping to enable quantum practicality, and quantum practicality is poised to enable vastly more powerful AI systems.
On the other hand, quantum practicality will probably introduce a whole new scale of AI risks. This is because organizations that are able to pair quantum and AI early on will gain massively powerful capabilities (like the ability to discover and exploit cybersecurity flaws with wholly unprecedented speed) that, when used maliciously, could threaten other organizations. In that world, AI could have the effect of creating more problems than it solves – at least for businesses that are not prepared.
The answer is two fold. On the one hand, quantum will likely make AI systems vastly more powerful than ever before, while also drastically reducing their costs. Businesses worried that high token costs make AI impractical from a financial standpoint will almost certainly find that, in the quantum age, the cost of intelligence falls faster than ever as the leading models become ever more capable.
For business leaders and strategists, the question is: How will quantum practicality affect AI transformation in the real world?
