AI Finds 'Square Structure' Inside 'Triangle' on Ceres, Failing Experiment to See If It Could Be Used to Find Aliens

In an experiment designed to see if artificial intelligence could be used to discover extraterrestrials, it detected an unusual "square structure" inside a "triangle" on Ceres, a dwarf planet that sits in the asteroid belt between Mars and Jupiter.

In discovering this "object," which researchers say is "probably just a play of light and shadow," the experiment shows that AI is just as fallible as humans when it comes to searching for signs of alien life.

Gabriel G. De la Torre, Associate Professor of Psychology at the University of Cadiz in Spain, was hoping to find out if AI was as vulnerable to visual trickery as humans. He was inspired by "bright spots" detected on the surface of Ceres in a crater called Occator. While these bright spots are most likely volcanic ice and salt emissions, when they were first photographed there was a large debate about what they could be.

Bright Spots
Picture of the Vinalia Faculae region of Ceres obtained by NASA's Dawn spacecraft on July 6, 2018 at an altitude of about 58 kilometres. AI detected a square and a triangle in the image. NASA/JPL-Caltech/UCLA/MPS/DLR/IDA

One of these bright spots, Vinalia Faculae, appears to show a series of geometric shapes. By comparing how humans and machine interpret these shapes, De la Torre hoped to discern whether or not AI could be used in the future to pick out "technosignatures," or "technomarkers," that offer signs of past or present technological activity. The idea being that it could be used to search for alien life.

The study, published in Acta Astronautica, De la Torre involved 163 volunteers with no background in astronomy and an AI system, called convolutional neural networks (CNNs), that was able to recognize geometric images in satellite images.

CNNs have previously been used in facial recognition technology and for analyzing documents.

Bright Spots Ceres
The researchers have observed the structure that appears in the central part, enlarged on the right, where the geometries that were most frequently detected by people are also indicated (below, indicated with numbers). Original photo: NASA/JPL-Caltech/UCLA/MPS/DLR/IDA/PSI

The participants and AI were presented with an image of Occator and asked them to pick out any geometric shapes they could see. Both humans and the AI identified a square shape in the image, but the AI also saw a triangle shape around the square. When the participants were shown the triangle, most claimed to be able to see it.

De la Torre says the findings suggest AI may not be a useful tool in the search for extraterrestrials. He said that while AI has "a multitude of applications," it could end up confusing us with "false" detections. He added that what the square structure actually is is unknown, "but what artificial intelligence has detected in Vinalia Faculae is most probably just a play of light and shadow."

The purpose of this paper, he said, was to see if AI could interpret new patterns that may be overlooked by the human eye. But his findings suggest this is not the case, especially as any alien technology out there may be beyond our imaginations.

"We humans conceive and model reality to fit our own convenience, experience and concepts—extraterrestrial intelligence being no exception," he wrote in the study. "When we, including scientists, talk about extraterrestrials, we tend to see them as somehow akin either to us or to robots, using radio waves and numbers, sending blueprints as an act of goodwill or even living around Dyson sphere-like megastructures," he added. "The truth could be quite different."