AI-driven telescope succeeds in its first observation campaigns in Chile

An automatic planner developed by the SkAI institute has decided on its own where to point the Blanco telescope during two campaigns, in the spring and summer of 2026.

Teams from Northwestern University, the University of Chicago and Fermilab, gathered in the AI Institute for the Sky (SkAI), have entrusted an artificial intelligence system with a task previously reserved for astronomers: to choose, night after night, the area of the sky to observe. The tool was put to the test on the 4-metre NSF Víctor M. Blanco telescope at the Cerro Tololo Inter-American Observatory in Chile, equipped with the Dark Energy Camera and its 570 megapixels.

The planner has been trained on the history of the Dark Energy Survey, which allows it to arbitrate in real time between weather coverage, atmospheric quality and available observation time — a scarce and expensive resource. Two campaigns conducted in the spring and summer of 2026 showed that it was doing as well as human planners, but not exceeding them for the time being.

For Alex Drlica-Wagner, one of the researchers involved, the experiment is a step towards more autonomous observatories. The challenge goes beyond comfort: as the large surveys of the sky multiply the transient targets to be followed in a few hours, human decision making becomes the slow link in the chain.