But Aliens might send something more complex than a single loud note. How do you scan SETI data for something that just seems anomalous or weird? Researchers have been trying to enlist artificial intelligence (AI), but it hasn’t been easy. One species of AI, natural language algorithms, can recognize key words in the flow of human speech think of Amazon’s Alexa, or eavesdroppers at the National Security Agency after being trained on vast speech data sets. But the huge number of narrow frequency channels in SETI data overwhelms these algorithms. Converting the data stream into 2D diagrams that resemble images works better, at least in tests, in which machine vision algorithms picked out strange pictures from a torrent of similar ones. “We have to guess what an anomaly might look like and train the algorithm to look for this, or look for things that look similar,” says Steve Croft of UC Berkeley’s SETI Research Center. How Big Money is Powering a Massive Hunt for Alien Intelligence
