Astronomers develop AI system for detecting, confirming, and announcing new supernovae

Astronomers have developed an AI system that can detect, confirm, and announce new supernovae. The recent surge of interest in AI may make the information you see and read online less reliable, but AI's role goes beyond that. Scientists are increasingly using AI systems to sift through massive datasets that would take years for humans to analyze completely. For example, researchers at Northwestern University in the United States used an AI model to detect, confirm, classify, and announce the discovery of a new supernova. Supernovae mark the end of a star's life, and the remnants they produce depend on the size of the star, which can be a white dwarf, neutron star, or something even more exotic like a black hole. The bright flashes of supernovae can outshine millions of other stars, but these beacons of stellar death are incredibly difficult to detect. They occur in distant locations from Earth and disappear from the background in a short period of time. According to Northwestern University, human astronomers have spent 2,200 hours over the past six years searching for signs of new supernovae in images. The new AI tool, called the Bright Transient Survey Bot (BTSbot), can autonomously perform this tedious task, giving astronomers more time to focus on more important work. After being trained on 1.4 million images from 16,000 different sources, BTSbot has learned how to identify supernova candidate stars and classify them as supernovae archived as SN2023tyk. The galaxy in which this supernova occurred can be seen in the image above. The research team applied BTSbot to data from the Zwicky Transient Facility (ZTF), a robotic observatory that scans the entire northern hemisphere of the sky every night. The facility detected a bright flash on October 3, which BTSbot confirmed as a supernova candidate two days later. The AI communicated with another robotic device called the Spectral Energy Distribution Machine (SEDM) to collect additional data. This allowed BTSbot to classify SN2023tyk as a Type Ia supernova, an important explosion type that helps scientists measure the expansion of the universe. The close-up shot of the supernova flash makes the Milky Way appear unbalanced. Source: National Science Foundation/Northwestern University Adam Miller, the lead author from Northwestern University, said, "Ultimately, removing humans from this process provides research teams with more time to analyze their observations and propose new hypotheses about the origins of the cosmic explosions we observe." This automated system shared the discovery with astronomers on October 7, four days after the event appeared in the telescope. AI can free up astronomers' time to work in multiple areas. Google has already demonstrated that similar AI models can detect exoplanets in a matter of hours.