Scientists have discovered eight potentially extraterrestrial signals using a machine-learning algorithm as published in the journal Nature Astronomy on January 30th. Although the study doesn't provide solid evidence of intelligent alien life, the authors suggest that the use of AI is a viable method in searching for extraterrestrial intelligence. However, a subsequent search for the signals yielded no results.
Cherry Ng, an astronomer at the University of Toronto and co-author of the study, expressed their admiration for the performance of the AI-based approach in searching for extraterrestrial intelligence. They are hopeful that l, with the assistance of AI, it will be possible to accurately assess the probability of extraterrestrial signals from other civilizations.
Extraterrestrial Signal Sources
The new method involves what lead author Peter Ma, an undergraduate from the University of Toronto, refers to as "semi-unsupervised learning." This approach combines aspects of both supervised and unsupervised machine learning. In supervised machine learning, the algorithm makes predictions based on data that is labeled by humans. On the other hand, unsupervised machine learning extracts patterns from vast data sets without human direction.
The researchers first taught the algorithm to distinguish between signals originating from Earth and those from extraterrestrial sources by using radio waves, which are frequently sought after in SETI searches due to their capability of traveling long distances in space. The researchers experimented to reduce false positives by testing various algorithms. They analyzed data collected by the Green Bank Telescope in West Virginia, which consisted of 150 terabytes of information and covered observations of 820 stars located near Earth.
As a result of the analysis, they found eight signals that had previously gone unnoticed from five stars located between 30 light-years and 90 light-years away from Earth. According to the scientists involved with Breakthrough Listen, a significant SETI project, the signals discovered have two characteristics that align with those that could potentially be created by intelligent extraterrestrial life.
Machine learning could help radio telescopes scour the cosmos for signs of extraterrestrial intelligence.
Random Alien Frequency
The signals are only present when the telescope is pointed towards the star and absent when it is not, unlike local interference which is generally constant. Additionally, the signals change the frequency over time, giving the appearance of being distant from the telescope. This was stated by Steve Croft, the project scientist for Breakthrough Listen at the Green Bank Telescope, in a statement.
To Peter Ma, the lead author of the study, the features observed in the signals could also occur randomly. To make any conclusions about extraterrestrial life, the researchers need to observe the same signals multiple times. However, a follow-up observation using the Green Bank Telescope did not yield any signs of the signals, as reported by Live Science.
The research team intends to use their algorithm on data from powerful radio telescopes like MeerKAT in South Africa or the proposed Next Generation Very Large Array in North America. Peter Ma stated that with their new technique and the advanced telescopes in the future, they are optimistic that machine learning can expand their search from hundreds of stars to millions.
RELATED ARTICLE: In Search For Alien Life: Wild Paper Claims Gravitational Wave Observatories Could Search for Warp Drive Signatures
Check out more news and information on Extraterrestrial Life in Science Times.
