Artificial intelligence can help scientists in the search for extraterrestrial life: Canadian specialists have developed neural networks that can help distinguish between alien signals and Earth signals.
People began searching for signs of extraterrestrial life in space quite a long time ago: the Search for Extraterrestrial Intelligence (SETI) program, which began back in 1984, serves this purpose. The problem is that in recent years, the development of wireless technology has begun to interfere with the search for radio signals in space.
As Vice.com says, Canadian scientists decided to connect to the search for extraterrestrial signals in space neural networks, which based on the library of machine learning TensorFlow and Python Keras library trained to analyze signals from space and look among them potentially interesting anomalies.
The machine learning algorithm analyzed more than 480 hours of data from 820 stars collected by the telescope in 2016 and found 8 signals that had not been seen by previous algorithms and that could potentially have extraterrestrial sources.
In theory, the signals detected by the algorithm – if they do have an extraterrestrial source – could include information about the development of technology or even a set of technosignatures of an alien civilization. However, as the leader of this project, Peter Ma, a third-year physics and mathematics student at the University of Toronto, noted, he and his colleagues don't count on much of that.
According to Ma, the algorithm developed in his project works twice as fast as traditional algorithms. Moreover, using a neural network to learn data also allows for out-of-the-box "thinking" that more traditional human-driven algorithms don't have.
Traditional algorithms, as Ma explained, work with a given set of human-designed instructions. And that means that the algorithm will only look for and find what the human tells it to find. The problem, however, is that the nature of the alien signal is not fully known, hence the human cannot tell the algorithm exactly what it needs to look for. Here, however, the algorithm simply studies and analyzes, filtering out Earth signals that are already familiar to it and drawing scientists' attention to those signals that, for whatever reason, may seem anomalous.