IonQ and Nvidia unveil AI method to optimize quantum computing

A research collaboration between IonQ, Oak Ridge National Laboratory, Nvidia, and the University of Tennessee has introduced DQAOA-GPT, a hybrid framework designed to optimize quantum computing processes using generative AI.

Traditionally, quantum approximate optimization algorithms (QAOA) require an iterative process of running circuits, measuring outputs, and adjusting parameters to find solutions. This cycle is computationally expensive and introduces noise on current noisy intermediate-scale quantum (NISQ) devices. The new DQAOA-GPT method utilizes a transformer-based generative model to synthesize efficient quantum circuits in a single step, bypassing the need for thousands of iterative loops.

By training the AI to understand the relationship between problem structures and optimal circuit parameters, the team has automated the circuit design phase. Testing on complex Higher-order Unconstrained Binary Optimization (HUBO) instances showed significant cost reductions and improved efficiency, providing a potential pathway for scaling hybrid quantum systems to solve larger industrial and scientific problems.