Ever discovered an important email in your spam? The same principal has some astronomers wondering if aliens might have already tried to make radio contact, but we’ve missed it in piles and piles of unused data created by telescope research every year.
Now a new study has used machine learning to comb through hundreds of hours of radio signals for items of interest which could be the equivalent of an important message in your spam tray.
They identified 115 million items, but then using machine learning tool, were able to narrow the search to eight newly discovered ‘signals of interest’.
Unfortunately, there’s no definitive signs of aliens found yet, but the new method could be used on other large datasets, potentially allowing researchers to quickly search through large swathes of data to maybe find messages from extraterrestrials.
“‘Are we alone?’ is one of the most profound scientific questions that humans have asked,” the international team of researchers wrote in their new paper.
“The search for extraterrestrial intelligence (SETI) aims to answer this question by looking for evidence of intelligent life elsewhere in the galaxy via the ‘technosignatures’ created by their technology.”
The researchers took data from the Robert C. Byrd Green Bank Telescope, located in West Virginia, in the US. The dish is the world’s largest fully steerable radio telescope and does about 6500 hours of radio observations every year.
A whopping 480 of those hours was used in this new study, zooming in on the radio data from 820 stars. This data was given to a machine learning tool that uses a novel ‘β-convolutional variational autoencoder’ or β-VAE for short.
The machine learning model looked at 115 million ‘snippets’, and came back with almost 3 million ‘signals of interest’.
However, of these signals of interest, a lot of them were found to be radio frequency interference. That means that it’s probably radio signals from Earth, not from space. After some human wrangling of the data, they got the number down to just over 20,000 signals of interest.
“[We reduced] the number of candidate signals by approximately two orders of magnitude compared with previous analyses on the same dataset,” they wrote in their paper.
“Upon a visual inspection, we identify eight promising signals of interest that show narrow, drifted signals.”
In May 2022, the team went back to check on these eight sources, which were coming from five different stars. Unfortunately, they didn’t find any similar readings, so this is no smoking gun when it comes to aliens.
However, letting AI look through millions of snippets of radio noise and narrowing down the search for where to look is an enticing prospect for researchers.
With most astronomical data being unused currently, it could be a good way to ‘reuse’ some of that data to find a whole new way of seeing.
“I think one of the underutilised possibilities right now is that we have so much data coming in from telescopes everywhere – optical, radio, infrared, X ray. And if we’re trying to answer specific science questions with them, a lot of the other stuff in the data is noise,” CSIRO’s Dr Vanessa Moss told Cosmos in December.
“I think that’s a missed opportunity … Some code, or machine learning, or intelligence that could look at all the data, ever and then try to find these outliers.”
The research has been published in Nature Astronomy.