The universe is estimated to contain hundreds of billions of galaxies, and there are roughly that many planets in each galaxy, so the probability of Earth being the only place with life is incredibly low. A new artificial intelligence system has analyzed millions of radio signals from space, identifying those with a potentially artificial source, and discovering eight strange-looking signals.

If the aliens were to scan the Earth with the right equipment, they would pick up our radio chats and other electromagnetic signals that we’ve been transmitting for most of the century. With this in mind, the Groundbreaking Listening initiative aims to change that and search for artificial radio signals from other planets in our galaxy. The team calls these signals “techno signatures.”

The problem is that the universe is a noisy place: stars, black holes, magnetars, quasars, FRBs, supernovas, gamma-ray bursts, and a host of other objects and events can generate radio and other signals. And then there are all the distractions from our own technology, like cell phones and GPS satellites. Tuning the background noise to detect possible alien techno signatures is an enormous task.

Or at least for humans. Artificial intelligence is adept at sorting through large chunks of data to find patterns, which means it’s an excellent job for it. So for a new study, University of Toronto student Peter Ma developed a new machine learning algorithm designed to filter out the most promising technological signature candidates.

The algorithm works according to a two-stage process. The first step is to involve an autoencoder trained on simulated intelligent extraterrestrial signals so that it knows what to look for. In fact, it should be a narrowband signal with a noticeable aberration rate that appears only during observations of certain areas of the sky. These simulated signals are added to the real data pool until the autoencoder learns to reliably select them.

As soon as you do this, the artificial intelligence will start the real work. Every signal in the dataset goes through an algorithm called a random forest classifier that separates the significant from the noise. In this case, the team fed more than 150 TB of data collected by Green Bank Telescope (GBT) into an artificial intelligence system.

The AI ​​identified 20,515 interesting signals from the 3 million signals in the dataset. The researchers then had to manually test each one – and curiously enough, eight of these signals had the right characteristics to be techno signatures and could not be attributed to interference.

“The eight signals looked very suspicious, but after we looked at the targets again with our telescopes, we didn’t see them again,” Ma said. “It’s been almost five to six years since we got the data, but we still haven’t seen the signal again. Do whatever you want.”

As intriguing as the signals are, they are far from the answer to the deepest question of whether we are alone in the universe. The team acknowledges that we don’t know exactly what a real extraterrestrial techno signature would look like, and that using a simulation could teach AI to focus on the wrong things. All in all, this is a valuable exercise, and applying AI to other datasets can yield more potential techno-signatures.