About 540 million years ago, various life forms suddenly began to emerge from the muddy floor of planet Earth. This period is known as the Cambrian Explosion, and these aquatic creatures are our ancient ancestors. All complex life on Earth evolved from these underwater creatures. Scientists believe this requires only a slight increase in the oxygen level in the ocean above a certain threshold. We may now be in the midst of the Cambrian explosion of artificial intelligence (AI).

Over the past few years, numerous incredibly powerful AI programs such as Midjourney, DALL-E 2, and ChatGPT have demonstrated our rapid progress in machine learning. Artificial intelligence is now used in almost every field of science to assist researchers with routine classification tasks. It’s also helping our team of radio astronomers expand the search for extraterrestrial life, and the results so far are promising.

Detection of extraterrestrial signals using AI

As scientists seeking evidence of intelligent life beyond Earth, we have created an artificial intelligence system that outperforms conventional algorithms in signal detection tasks. Our artificial intelligence has been trained to look for signals in radio telescope data that cannot be produced by natural astrophysical processes. When we fed our AI with a pre-learned dataset, it discovered eight interesting signals that the classical algorithm missed. To be clear, these signals probably don’t come from extraterrestrial intelligence and are most likely rare cases of radio interference. With all that said, our findings are published today in the journalism. Nature Astronomyemphasizes that artificial intelligence technologies will continue to play a role in the search for extraterrestrial intelligence.

not that smart

AI algorithms don’t “understand” and “don’t think”. They’re great at pattern recognition and have proven extremely useful for tasks like classification, but they’re not problem solvers. They only perform the specific tasks they were trained to do. So, while the idea of ​​an AI detecting extraterrestrial intelligence might sound like the plot of a thrilling science fiction novel, both terms have their flaws: AI programs are not smart, and searches for extraterrestrial intelligence cannot find direct evidence of intelligence.

Instead, radio astronomers look for radio signatures. These hypothetical signals would indicate the existence of technology and therefore the existence of a society that can use technology to communicate. For our research, we created an algorithm that uses artificial intelligence techniques to classify signals as candidates for radio interference or a true techno signature. And our algorithm works better than we expected.

What does our AI algorithm do?

The search for a techno signature has been likened to searching for a needle in a cosmic haystack. Radio telescopes generate huge amounts of data, and that means massive amounts of interference from sources like phones, Wi-Fi, and satellites. Search algorithms must be able to distinguish true techno signatures from “false positives” and do so quickly. Our AI classifier meets these requirements. It was developed by Peter Ma, a student at the University of Toronto and lead author of our paper. Peter added simulated signals to real data to create a training dataset, and then used that dataset to train an artificial intelligence algorithm called an autoencoder. As the autoencoder processes the data, it “learned” to identify distinctive features in the data.

In a second step, these features were fed into an algorithm called a random forest classifier. This classifier creates decision trees to decide whether a signal is notable or just radio interference – essentially it separates the “needles” of a techno signature from the haystack. After training our algorithm, we fed it more than 150 terabytes of data (480 hours of observations) from the Green Bank Telescope in West Virginia. It identified 20,515 interesting signals that we then had to manually check. Eight of these had techno signature marks and could not be attributed to radio interference.

Eight signals, no redetection

To try to confirm these signals, we returned to the telescope to re-observe eight relevant signals. Unfortunately, we were unable to re-detect any of them during our follow-up observations. We have already been in similar situations. In 2020 we detected a signal that turned out to be harmful radio interference. While we will be observing these eight new candidates, the most likely explanation is that they are unusual manifestations of radio interference rather than aliens. Unfortunately, the problem of radio interference will not go away. However, as new technologies emerge, we will be better prepared to address this issue.

narrow your search

Our team recently installed a powerful signal processor on the MeerKAT telescope in South Africa. MeerKAT uses a technique called interferometry to combine 64 dishes to act like a single telescope. This technique makes it possible to better determine where the signal is coming from in the sky, which significantly reduces false positives from radio interference. If astronomers manage to detect a techno signature that cannot be explained as interference, it would strongly suggest that humans are not the only technology creators in the Galaxy. This will be one of the most profound discoveries imaginable.

At the same time, just because we don’t discover anything doesn’t mean we’re the only “intelligent” species that is technologically capable. Undetectability may also mean that we are not looking for the right kind of signals or that our telescopes are not sensitive enough to detect faint emission from distant exoplanets. We may need to cross a threshold of sensitivity before the Cambrian Explosion can be discovered. On the other hand, if we are truly alone, we must consider the unique beauty and fragility of life on earth.