The partnership between Quantum Machines and Nvidia, which combines Quantum Machines’ cutting-edge quantum control technology with Nvidia’s DGX Quantum platform, was announced almost a year and a half ago. Despite a period of silence, the collaboration is already yielding encouraging outcomes, advancing the field’s quest for an error-corrected quantum computer.
The businesses showed earlier this year that a commercially available reinforcement learning model on Nvidia’s DGX platform could support qubit calibration on a Rigetti quantum device. Enhancing the fidelity of “π pulses,” which are essential for rotating qubits in quantum processors, was the main goal of this partnership. These pulses must be regularly recalibrated due to the continual drift in quantum systems, a task that is ideally suited to reinforcement learning because it demands a lot of control.
While earlier conventional computing systems for quantum control were limited, Nvidia’s potent DGX platform allows for sophisticated real-time recalibrations, which are crucial for sustaining high performance, according to Yonatan Cohen, co-founder and CTO of Quantum Machines. Sam Stanwyck, Nvidia’s product manager for quantum computing, highlighted how important the platform’s low latency is for these calculations. Optimizing qubit control pulses is essential to tackling quantum error correction, a significant difficulty as quantum computers grow in size.
According to Quantum Machines Product Manager Ramon Szmuk, even a 10% improvement in calibration might result in exponential advances in error correction for logical qubits, demonstrating the potential significance of better calibration.
This partnership is only beginning. The group made use of a simple quantum circuit, but they intend to expand to manage more intricate systems and develop open-source libraries for wider application. Stanwyck believes that the combination of quantum control and supercomputing is a crucial step towards practical quantum computing. With the release of Nvidia’s next-generation Blackwell processors in 2019, the platform’s capabilities are expected to expand, enabling additional developments in quantum error correction and control.
Source: Nvidia, Quantum Machines.