Researchers in Switzerland have built an AI model to uncover potentially habitable worlds that are hiding from view.As detailed in a new study published in the journal Astronomy and & Astrophysics, the machine learning algorithm identified 44 star systems that it suspects harbor Earth-like exoplanets we haven't detected yet, in a promising demonstration of an approach that could accelerate the search for planets teeming with life.It hasn't outright confirmed that the Earth-like planets are actually there — but it's teed up astronomers to investigate those stellar neighborhoods in the future. In simulations, the model achieved an impressive precision value of up to 0.99, meaning that 99 percent of the systems identified have at least one Earth-like planet."It's one of the few models worldwide with this level of complexity and depth, enabling predictive studies like ours," co-author Dr. Yann Alibert, codirector of the University of Bern's Centre for Space and Habitability, said in a statement, as quoted by Forbes. "This is a significant step in the search for planets with conditions favorable to life and, ultimately, for the search for life in the universe."Exoplanets are notoriously difficult to spot, because they're tiny compared to stars and produce little light of their own. So far, scientists have confirmed the existence of just over 5,800 planets outside our solar system — and the data we have on most of these is scant.That doesn't give a lot of material to train a pattern-seeking algorithm on, which require huge data sets. Instead, the astronomers fed their model synthetic planetary systems generated with the Bern Model of Planet Formation and Evolution, which comprehensively simulates the development of hypothetical planets as far back to their inception from a protoplanetary disc."The Bern Model is one of the only models worldwide that offers such a wealth of interrelated physical processes and enables a study like the current one to be carried out," Alibert said in a statement about the research.During these tests, the AI model revealed that the strongest indicators of an Earth-like planet could be found in a system's innermost detectable planet, particularly its mass and orbital period, the researchers wrote in the study.From there, the team applied the machine learning algorithm to a sample of nearly 1,600 systems with at least one known planet and either a G-type, K-type, or M-type star, with G-types being Sun-like stars, and the remaining two classifications describing stars that are smaller and cooler. That revealed that nearly four dozen of them likely harbor an Earth-like world.But the model isn't infallible. It hasn't reproduced certain characteristics of star systems that astronomers have observed, such as the strong correlation between so-called Super Earths and Cold Jupiters, which often appear together around Sun-like stars, the authors note. And the synthetic planets tend to be found closer to their stars than real ones.Still, it doesn't need to be perfect: anything to narrow down astronomers' hunt for Earth's cousins through the the unfathomably vast cosmos could be a gamechanger.Share This Article