On September 23, Intelligent Quantum Inception (Beijing) Technology Co., Ltd. (IQI) held a launch event in Beijing for Baize, an AI-agent-driven neutral-atom quantum computing experimental platform.
The event was hosted by IQI and co-organized by Qosmos, iFLYTEK, and the Beijing Key Laboratory of Cold Atom Quantum Computing, with representatives from government, universities, industry, and the media in attendance.
On behalf of IQI, Qi Shi, Chairman of Qosmos and General Manager of IQI, together with Qosmos, iFLYTEK, Tsinghua University's neutral-atom quantum computing team, and the Beijing Key Laboratory of Cold Atom Quantum Computing, formally launched Baize, China's first AI-agent-driven neutral-atom quantum computing experimental platform. The platform embeds AI agents in a real quantum experimental system to conduct quantum science research intelligently, reaching an internationally leading level.
In neutral-atom quantum computing experiments, laboratory staff must spend considerable time each day calibrating equipment and maintaining the system's operating condition. Changes in laboratory temperature, magnetic fields, electric currents, and laser frequencies can all render yesterday's effective parameter settings ineffective today. Addressing this problem has become urgent for the development of neutral-atom quantum computing: on the one hand, researchers have to spend a great deal of time each day frequently switching between control software, camera images, and analysis programs, severely limiting the efficiency of quantum computing research and development. On the other hand, when experimental parameter drift is detected, experiments typically have to be interrupted so that the parameters can be adjusted to keep the system running. Future real-world applications will require quantum circuits to run continuously for days or even longer, making intelligent long-term operation and maintenance one of the key technical challenges to be addressed on the path toward practical neutral-atom quantum computing.
Baize, launched at this event, addresses precisely this problem. With AI agents embedded in the experimental control system, researchers can state their goals in natural language, and Baize can arrange experimental workflows, call experimental tools, coordinate experimental measurements, and intelligently adjust parameters based on experimental results, carrying out interactive validation to achieve the desired results and complete the experiment. When performance indicators decline, it can use experimental images and learn from expert knowledge bases, including previous experimental records and historical data, to identify the causes and make corrections. It also keeps records of its actions and results, much like an experienced experimental engineer.
Two videos of real experimental demonstrations were shown at the launch event, presenting the two core capabilities that Baize has already achieved. The first demonstrated goal-driven parameter adjustment in response to natural-language instructions and intelligent closed-loop execution of an entire experiment. The second demonstrated continuous monitoring and autonomous correction of key quantum computing parameters.
In his remarks, Hui Zhai, Professor at Tsinghua University and Director of the Beijing Key Laboratory of Cold Atom Quantum Computing, noted that the development of Baize is an inevitable consequence of the development of artificial intelligence and a necessity for the development of quantum technology. He outlined the team's journey in developing an AI-agent-driven quantum computing platform and described the strategic thinking behind using it as a starting point to comprehensively advance intelligent experimental platforms for the physical sciences in the future.
Tao Zhang, R&D Director at IQI, answered reporters' questions about Baize's technical direction, application scenarios, and future plans. Zhang said, “We hope that Baize can advance the scientific research paradigm and keep improving through working alongside human experts.” The Baize 1.0 version launched at this event is an “AI experimental engineer” capable of handling daily operation, calibration, parameter optimization, and system-state maintenance, supporting the intelligent and safe execution of experiments. The Baize 2.0 version that IQI will release next will be an “AI experimental scientist” capable of reading literature autonomously and independently conducting scientific experimental exploration. As Baize progresses from an “AI experimental engineer” to an “AI experimental scientist,” human researchers will have more energy to devote to scientific goals, physical judgment, and ideas for new experiments.
END
