An AI discovers strange radio signals from previously analyzed data. It is now clear to everyone that we are in the midst of a rapid evolution of artificial intelligences. In recent years, many incredibly capable AIs like Midjourney, DALL-E 2, and ChatGPT have shown rapid progress in machine learning. They are increasingly being used in all areas of science to help researchers with routine classification tasks.
AIs in the search for intelligent life
Even scientists involved in the search for evidence of intelligent life are not left behind. Researchers have built an AI (ref.) that outperforms classic algorithms in signal detection activities. This artificial intelligence was trained to search radio telescope databases for signals that cannot be generated by natural astrophysical processes. When the AI was provided with a previously studied data set, the artificial intelligence reported eight signals of interest that the classic algorithm had not identified.
To be clear, these signals do not come from extraterrestrial intelligences and are probably rare cases of radio interference. However, the discovery published today in Nature Astronomy highlights how AIs will continue to play a continuous role in the search for extraterrestrial intelligence.
AIs do not understand or think. They excel in recognizing patterns and are very useful for classification activities, but they do not have the ability to solve problems. They only perform the specific tasks for which they were trained. So, it must be specified that artificial intelligence programs are not intelligent and will not find direct evidence of extraterrestrial intelligence.
Training an AI in the search for Technosignatures
Astronomers have been searching for radio technosignatures for some time. These signals would indicate technological presence and, therefore, the existence of a civilization that is utilizing it. To expand this search, scientists are combining traditional algorithms with AI to find a true technological signature among radio interference. Surprisingly, the AI works better than they hoped.
The technosignature search is comparable to the search for a classic needle in a cosmic haystack. Radio telescopes produce massive amounts of data, and there are a multitude of interferences from sources such as phones, Wi-Fi, and satellites. The search algorithms used must be able to quickly separate real technological signatures from false positives.
The AI used in this study was designed by Peter Ma, a student at the University of Toronto and lead author of the article. As a training set, Ma entered real signals, and as the AI processed the data, it learned to identify salient features in the data. After training the algorithm, more than 150 terabytes of data (480 hours of observation) from the Green Bank Telescope in West Virginia were provided to the AI and it identified 20,515 signals of interest.
The eight identified signals
Eight signals identified were manually inspected and had the characteristics of techno-signatures and could not be attributed to radio interference. In checking these signals, researchers went back to the telescope to observe the signals of interest, but unfortunately, they were unable to detect any of them again.
To resolve this issue, the team has recently implemented a powerful signal processor on the MeerKAT telescope in South Africa. Thanks to interferometry, the radio telescope will be better able to determine where a captured signal originates, reducing false positives due to radio interference. If astronomers were able to detect a technological signature that cannot be explained as interference, it would be a decisive signal that humans are not the only creators of technology within the Milky Way. It would, without a doubt, be the most important discovery one can imagine.
At the same time, if we do not detect anything, it does not necessarily mean that we are the only technologically capable species in circulation in the galaxy. A lack of detection could also mean that we have not searched for the right type of signals, or that our telescopes are not yet sensitive enough to detect weak transmissions from distant exoplanets.