Radio telescopes still seem to be the best choice for the search for extraterrestrial intelligence, SETI for short: radio signals can easily propagate in space, and radio transmitters and receivers can be operated efficiently. But mankind also knows that radio signals can be used to transmit information wirelessly over long distances. That is why the search for aliens is mainly disturbed by man-made radio interference : are they aliens or is it a scattered GPS signal? In the future, artificial intelligence (AI) could answer this question better than the methods used to date. Researchers present the deep learning algorithm in the journal “Nature Astronomy”.. The method can already point to a first partial success: It filtered out eight potential alien signals from the data mass that nobody had noticed before.
Peter Xiangyuan Ma and his colleagues selected 820 stars from the mountain of data from Breakthrough Lists. That corresponded to 480 hours of observation data from the Robert C. Byrd Green Bank telescope in the USA. They then unleashed an AI on the 115 million data snippets: a special algorithm that they had previously trained with artificially generated signals to filter out the unusual. The main task for the AI is to block out millions of terrestrial interference signals that are generated, for example, by our cell phones or GPS receivers. The AI thus reduced the data set to around 20,000 signals. The researchers write that previous analysis methods could not have analyzed this data set nearly as efficiently.