Technical innovations are often overinterpreted in the short term and underestimated in the long term.
Artificial intelligence (AI) and quantum computing represent two such innovations, particularly for businesses. The potential for AI far transcends the current applications of chatbots like ChatGPT and other language models.
It is the “fusion” of quantum computing and generative AI that ultimately presents a revolutionary approach to solving complex problems across various industries, Christopher Savoie, CEO at Zapata AI, told PYMNTS.
“What we’re generating here is that we are learning a process, an industrial process. And then we’re simulating, if you will, other alternative ways to do things,” Savoie explained.
This approach can lead to more sophisticated and nuanced outputs, including real-time insights that are applicable across industries such as defense, manufacturing, automotive and finance.
“Large language models,” Savoie said, “are basically a glorified word prediction like on the iPhone. … It’s a statistical model that’s trained on the entire internet. But in industrial use cases, you need a much more tailored model.”
“The latest Taylor Swift lyrics is not going to do a great job at doing FP&A for a CFO,” he added, stressing that LLMs trained on general internet data might struggle to provide accurate financial forecasts or optimized industrial processes.
The application of quantum-enhanced AI is already making impacts in real-world scenarios.
Explaining a project Zapata did with BMW, Savoie said his team was given the challenge of optimizing factory processes under a multitude of constraints, including labor laws, union rules and varying supply chain dynamics across different regions.
“Optimizing this per plant, per process is a very difficult thing to do,” Savoie said. But by using quantum generative math to simulate and generate better ways to optimize factory operations, the AI was able to propose more optimal solutions in 70% of cases.
The power of quantum computing lies in its ability to handle statistical models with unprecedented accuracy and efficiency. Generative AI, which relies heavily on statistical models, benefits immensely from quantum algorithms.
“The statistics we use in quantum land are better at generalizing and expressing over those statistical models,” Savoie explained about quantum-enhanced AI. “The ability to accurately simulate these complex scenarios is important.”
He clarified that while quantum technology holds tremendous promise, the current state of quantum hardware is not yet at a stage where it can outperform classical computers for most tasks. “Eventually, we’ll have hardware that does this natively,” Savoie said, “but it’s going to take time until we have fault-tolerant, perfectly computable systems.” However, Zapata AI is already leveraging quantum mathematics on classical hardware, using GPUs to tackle problems that were previously insurmountable.
Looking ahead, Savoie emphasized the need for smaller, more tailored models that can be trained on domain-specific data to achieve better results. Quantum-enhanced AI, with its superior ability to generalize and model complex scenarios, offers a promising solution to this challenge.
“We’re not looking for one omniscient human to run an entire Fortune 100 company, and it’s the same with AI. You don’t want the CFO doing the engineer’s job, or the engineer doing marketing. … We tend to differentiate and specialize in human organizations, and that is where AI is going, too,” he explained.
“You want to orchestrate the ensembles. … We’re going to have ensembles of smaller, more capable models for specific tasks that then can interrelate with each other and communicate with each other and be used in concert with each other to get better business outcomes.”
And as AI continues to evolve, the ethical and societal considerations surrounding its advance are becoming increasingly important. Savoie, who chairs the Quantum Technical Advisory Committee for the Quantum Economic Development Consortium, stressed the need for a multidisciplinary approach to addressing these concerns. “We need to do a lot more thinking,” he said, “not just from computer scientists but also ethicists, sociologists and economists.”
Ultimately, while potential for AI to be misused is a real concern, Savoie remains optimistic about the future. He argued that innovation should not be stifled, as it is crucial for developing defenses against potential adversarial AI. “We don’t want to stunt our ability to innovate,” Savoie warned, “because the innovation’s going to help us protect ourselves.”
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
We’re always on the lookout for opportunities to partner with innovators and disruptors.
Learn More