Short answer
Quantum-inspired AI describes models that borrow the computational logic of quantum computers but run on ordinary graphics processors in production. The approach aims to let companies gain efficiency without waiting for general-purpose quantum hardware to mature. New funding for the Chinese start-up Qingxing stands out as a sign of growing interest in the field.
Highlights
- Beijing-based Qingxing has closed two funding rounds in three months to develop quantum-inspired AI.
- The company's RiverONE model is designed to run on ordinary GPUs without using quantum hardware.
- Qingxing is working on hardware compatibility with Chinese chipmakers such as Biren Technology and MetaX.
2 min read
Qingxing Heterogeneous Computing, which has its roots at Tsinghua University, has announced that it closed two funding rounds within three months to develop quantum-inspired AI, according to The Quantum Insider, with the total reaching roughly 100 million yuan (about $14.9 million).
What does the quantum-inspired approach mean?
According to the report, Qingxing uses simulated quantum computing methods during model training; once training is complete, however, the model runs on ordinary graphics processing units (GPUs) without needing a real-time quantum computer. This gives companies a route to commercial products without waiting for general-purpose quantum hardware to mature.
Partnerships with chipmakers
The Quantum Insider reported that Qingxing is working with Chinese chipmakers such as Biren Technology and MetaX to adapt its models to different hardware, but that revenue from these partnerships has not been disclosed. The report also noted that the company had previously established a collaboration with AMD.
What should software teams take from this?
The goal of shrinking model size is also worth following for companies that want to cut cloud computing costs: a model that does the same job with fewer parameters can also reduce server costs and latency.
The report itself carries a caveat: it does not say which independent body carried out the comparison, and it stresses that the result is based on a single specialised task.
- Add approaches that run on existing GPUs, rather than those waiting for quantum hardware, to your R&D team's radar.
- Do not base product decisions on performance comparisons in funding news without independent verification.
- Track chip compatibility partnerships as an indicator of the technology's commercial maturity.
Frequently asked
- Does quantum-inspired AI use a real quantum computer?
- No. Quantum computing is simulated while the model is being developed, but the version used in production runs on ordinary graphics processing units (GPUs).
Sources
Follow UNIT Journal
What's new in search, AI and technology, in your feed every day.
