Study warns AI may trick humans with false evidence of alien life

Humanity has long wondered whether there is advanced intelligent life in the universe around us, but evidence of such species remains absent. As human technology advances, our species has increasingly embraced these tools as potential ways to detect evidence of other advanced extraterrestrial technologies. One such tool is artificial intelligence, but a new study warns that it may trick humanity with 'evidence' of advanced alien life.

One of the biggest efforts to find evidence of alien life comes from the SETI Institute, which details on its website the use of artificial intelligence and machine learning to detect unique or odd patterns involving radio bursts and similar things. These oddities are referred to as 'techno-signatures' — things that provide evidence of technology and, if they originate from far out in the universe, potentially evidence of advanced alien life.

A newly published study warns that using AI to detect these techno-signatures may be risky, however, with the potential to make humans think they have discovered evidence of alien life when no such evidence actually exists.

The new study points toward Ceres, the dwarf planet that briefly dazzled humanity with then-mysterious bright spots that some hoped would be evidence of alien life (spoiler: they turned out to be shiny salt deposits). Images of the planet included a unique square-shaped structure and, later found by AI, a seemingly triangle-shaped structure.

Such unusual formations may point toward past or present advanced extraterrestrial intelligence...but it may also just be the result of things like unique angles mixed with deceptive shadows. The new study involved showing participants an image of the planet and asking them what they saw. Many people pointed out the unique square-shaped structure.

The researchers also fed the imagery into an AI that was trained to spot squares and triangles. In addition to identifying the same square spied by humans, the AI also spotted a unique triangle shape. Once humans were made aware of the triangle shape, a number of participants claimed that they, too, were able to make out the same triangle.

The findings indicate that while AI may be able to spot things humans miss, it also risks confusing the data by presenting humans with mistakes or things that aren't quite what they appear to be, causing people to potentially perceive the same erroneous information. Things uncovered by AI risk biasing humans who may already believe they've found some type of evidence of extraterrestrial life.