Generative AI automates quantum optimization circuit design

IonQ and Oak Ridge National Laboratory demonstrated a generative AI method for synthesizing quantum optimization circuits, replacing manual parameter tuning. On a 100-variable benchmark, tuning time remained ~28 seconds as subproblems grew from 4 to 12 qubits, while the old method grew from 34 seconds to 11 minutes.

• AI model based on transformer architecture trained on existing quantum circuits
• Old method: time grew from 34 seconds to 11+ minutes at 4–12 qubits
• New method holds ~28 seconds regardless of subproblem size
• Accuracy doubled on large subproblems; simulations on Nvidia H200 GPUs