The NSF-DOE Vera C. Rubin Observatory in Chile will generate up to 10 million alerts every night during its Legacy Survey of Space and Time (LSST), each one flagging a point in the sky that has changed - a flare, a moving asteroid, an exploding star. With a 3,200-megapixel camera, the world’s largest, the observatory produces about 20 terabytes of raw images per night and a new image every 40 seconds.
No human team can triage 10 million alerts a night. Instead the alerts are routed to an ecosystem of software systems called community brokers - including ALeRCE, ANTARES, Fink and Lasair - that use machine learning to filter and classify them and pass on the ones astronomers care about. The scale of the data is the reason machine learning has become central to time-domain astronomy: the survey is built around the assumption that algorithms, not people, will do the first pass over the changing sky.