There are many tasks that AI systems can excel at, such as analyzing large databases. This also includes examining astronomical recordings, which on the one hand is a very tiring and time-consuming task, and on the other hand involves the possibility of human error. the Northwestern University An international group led by It means that it is a robot that temporarily measures the brightness of objects. In the past six years, researchers working on developing artificial intelligence have earned approx. They spent 2,200 hours sifting through tons of footage (with their eyes), looking for short bursts of supernova flares to determine which ones were worthy of further investigation. With the onset of artificial intelligence, professionals can spend this time on more useful tasks.
“This was the first time that a series of robots and artificial intelligence algorithms had observed, identified, communicated with another telescope, and finally confirmed the discovery of a supernova,” explained Adam Miller, head of the research. He added that automation allows astronomers to focus instead of on the monotonous work of, for example, clarifying the theoretical background to our knowledge of these explosions.
Palomar Mountain Observatory, operated by the California Institute of Technology.
Source: Wikimedia Commons
The current method also uses robotic telescopes, which take repeated images of one part of the sky and then, by comparing the images, bright spots seen by humans are further examined. “The automated software gives people a list of potential targets that look like explosions, and then they analyze the scene on them,” Miller added. “We can’t decide if a candidate is really a supernova unless we examine its spectrum, and scattered light from a light source reveals elements that It arose during the explosion.” There are robotic telescopes that also function as spectrographs, but this work is often done by professionals.
The supernova SN2023tyk detected by the system can be seen on the right side of the image, and almost on the left side of the distant galaxy. At 10-11am – The photo on the left was taken earlier, before the explosion, and by definition still missing light.
Source: Northwestern University
1.4 million recordings of about 16,000 objects were used to train the AI, including supernovae, temporarily bright stars, periodic variable stars, and galactic explosions. The newly trained AI was then used to examine an object that was first imaged on October 3, 2023 by the ZTF (Zwicky Transient Facility – a project that practically searches for various flares through camera recordings applied to one of the telescopes) (from Palomar Mountain Observatory in California), Then on October 5, the observatory was asked to create a scene based on a recent image of the object. Another robotic instrument at the observatory, the SED machine, did this and sent the data to an artificial intelligence-based system run by the California Institute of Technology, which was able to classify the supernova into the correct type based on the spectrum. Accordingly, the system witnessed a type Ia supernova explosion, which is a white dwarf explosion that existed in a binary system. After classification, where it was confirmed that it was indeed a supernova, the automated system I posted it too In the appropriate forum Supernova discovery on October 7, 2023.
