Xianggan Technology Founder Jin Yirong: The More Realistic Near-Term Path Is AI Helping Quantum, Not the Other Way Around
Based on nearly two decades of superconducting quantum circuit experience, Jin Yirong, founder of Xianggan Technology, argues that quantum computing belongs to the compute layer while AI belongs to the application layer. Because current quantum systems are still battling noise, the more realistic near-term path is to use AI for quantum calibration, parameter optimization, and fault diagnosis first, and then let quantum computing reverse-empower AI's combinatorial optimization tasks.
ImportanceMaterialEvidenceE3 inspectableWrite-upStandard
The relationship between quantum computing and AI is usually discussed in one direction: once quantum computing matures, it will help AI break through training bottlenecks. Jin Yirong, founder of Xianggan Technology, offers the opposite near-term judgment. He has worked in superconducting quantum circuits for nearly twenty years and simultaneously leads the ecosystem development team at the Beijing Quantum Information Science Research Institute. In October 2023 he spun Xianggan Technology out of that institute's superconducting quantum computing team; the company's quantum chips have now reached the hundred-qubit scale, and it is building a thousand-qubit cloud platform.
His core argument rests on a layering logic: quantum computing belongs to the compute layer, providing foundational capability; AI belongs to the application layer, built on top of computation. Only with stronger computing power can you get better AI. But quantum computing is still in a critical stage of engineering, still locked in intense struggle with noise. Precision is insufficient, error rates have not come down, and no matter how large the scale, it cannot be put into practical use. Under these conditions, quantum computing directly helping AI is unrealistic.
He believes the more realistic near-term direction is precisely the reverse: first use AI to help quantum computing with calibration, parameter optimization, and fault diagnosis; leverage AI to quickly improve quantum computing capability, thereby enabling quantum computing to reverse-empower AI sooner. This direction corresponds directly to the noise challenge described above: the core bottleneck in quantum computing's engineering stage is precision and error rate, and calibration, parameter optimization, and fault diagnosis are exactly where those bottlenecks manifest.