- Alibaba’s AI work platform helps researchers build a telescope simulation system without bespoke software development
- The project shows AI agents moving from data analysis into scientific workflows
QwenWork, Alibaba’s AI productivity platform, has helped a research team at China’s National Astronomical Observatories build a digital simulation system for a large-aperture research telescope in three days, according to Qwen.
Similar systems previously took about three months and cost tens of thousands of yuan to develop, while the new system cost less than 1,000 yuan ($149), according to Qwen.
The virtual system recreates the telescope’s operating environment, allowing an AI agent to read equipment status, plan observations, call interfaces and execute procedures as if it were operating the real instrument.
Researchers can therefore test observation strategies without consuming valuable telescope time.
AI enters research workflows
That matters in astronomy, where transient events such as supernovae and gamma-ray bursts are difficult to predict.
Meanwhile, operating large telescopes can cost tens of thousands of yuan per effective observation hour.
The agent framework has been connected to the SiTian Pathfinder (“司天”探路者) and SiTian prototype (“司天”原型机) telescope.
It has flagged eight very early supernova candidates, two of which triggered follow-up observations when weather conditions permitted, according to Qwen.
From analysis to execution
Li Yuyang (李瑀旸), an expert on the National Astronomical Observatories’ AI Promotion Committee, said AI is moving beyond data analysis into the execution of scientific research.
“Researchers can focus more on posing questions and setting goals, while intelligent agents take on observation, verification and iteration,” Li said.
The SiTian project has been included in UNESCO’s International Decade of Sciences for Sustainable Development recognition program and is planned to form an observation array of dozens of meter-class telescopes by 2030.
Agents run experiments
The broader significance of the project lies in how research infrastructure is built.
Instead of relying on customized software development, scientists used a general AI platform to create an environment where an agent can interact with scientific equipment.
Qwen said the simulation data will also be used to train vision-language-action models for telescope control, pointing toward a future where AI does not just analyze scientific data, but helps operate the instruments that generate it.
Header image credit: Alibaba Qwen
