How GPT-5.6 Sol helps run quantum computing experiments
GPT-5.6 Sol, coupled with Codex, autonomously conducts routine quantum computing experiments, freeing researchers for higher-level analysis.Summary
A graduate student in MIT's Engineering Quantum Systems Group, Beatriz Yankelevich, used GPT-5.6 Sol, integrated with the Codex platform, to test if AI could streamline her experimental workflow with superconducting qubits. These qubits are cooled near absolute zero and controlled via software, making them ideal for AI integration. The AI agent was given skills to run specific measurements. It successfully completed standard calibration sequences for a six-qubit chip, identifying frequencies and calibrating controls with minimal oversight. However, it struggled more with weak or noisy signals, indicating that while current AI excels at well-defined workflows, interpreting ambiguous physical data remains a challenge. The group now regularly uses such agents to handle routine measurements, allowing researchers to save time and focus on experimental design, result interpretation, and planning. The student reported that multiple agents can now work on different problems concurrently, significantly accelerating progress.
(Source：OpenAI)