A scientist, Peter Ma, has applied machine learning and artificial intelligence to data collected by the Search for Extraterrestrial Intelligence (SETI) Institute, a press statement reveals.
Based on initial results, there is a slight chance the new method may have unearthed non-Earth-based "technosignatures". That would mean it had achieved SETI's goal of finding signs of extraterrestrial intelligence.
Algorithm finds 8 promising signals that could be of alien origin
SETI was founded in 1984 to scan the skies for radio signals that could originate from technology developed by intelligent alien civilizations.
The search so far has turned up empty-handed, though there is a slight chance that we may have seen a breakthrough.
In a new paper published in the journal Nature Astronomy, Ma describes how he trained a machine-learning algorithm on 480 hours of telescope data from 820 stars collected in 2016. The algorithm identified eight signals of interest that previous algorithms had failed to detect.
Ma, an undergraduate at the University of Toronto, told VICE in an interview that their method completely removes humans from the equation, unlike previous machine learning algorithms applied to SETI data.
"This work relies entirely on just the neural network without any traditional algorithms supporting it and produced results that traditional algorithms did not pick up," Ma explained to VICE.
Why did the algorithm single out those 8 signals?
The result of Ma and colleagues' experiment is that we now have eight signals that may have originated from advanced extraterrestrial species.
Ma's algorithm specifically pinpointed signs that "are narrow band, doppler drifting signals originating from some extraterrestrial source."
As the SETI Institute's press statement points out, "signals caused by natural phenomena tend to be broadband,". In contrast, the ones picked up by the algorithm "were narrow band, meaning they had narrow spectral width, on the order of just a few Hz."
The signals also exhibited a number of properties that suggest they aren't caused by Earth-based interference, such as the fact they had non-zero drift rates.
Machine learning could crack the SETI code
Essentially, Ma and colleagues' algorithm filters out Earth-based interference while also identifying signals with specific features that could mean they are being used for communication or other technological applications.
Scientists are increasingly using machine learning and artificial intelligence to sift through massive amounts of cosmic data that could otherwise take humans years to investigate. Machine learning is a subset of AI that uses data to refine its search parameters and improve its capabilities.
Ma chose to use a machine learning neural network because this allows adaptability that is not afforded by more traditional artificial intelligence algorithms. "The issue is that the nature of an ET signal is not completely known," he told VICE, "hence our proposed approach is to just learn it."
Though Ma and colleagues' algorithm pinpointed eight unique signals, there is no guarantee that these signals did, indeed, originate from alien civilizations.
The next step is for researchers to investigate the signals in more detail and determine whether it's worth carrying out follow-up observations on the regions of space from which the mystery signals originated.
This article originally published in Interesting Engineering.
Read also: Is It Possible To Terraform Mars For Humans?