AI Model Achieves 99% Precision in Identifying Earth-like Planets in Simulations, Pinpointing 44 Real-World Targets
An AI model, trained on tens of thousands of simulated planetary systems, has demonstrated up to 99% precision in recognizing simulations containing an Earth-mass planet within a broad temperate zone. This model, developed by Jeanne Davoult, Romain Eltschinger, and Yann Alibert, was then applied to 1,567 real planetary systems, identifying 44 potential targets for further observation. The 'Earth-like planet' (ELP) in this study is defined by a mass between 0.5 and 3 times that of Earth and an equilibrium temperature between 160 and 510 Kelvin. The model's high precision was achieved on synthetic data, where the presence or absence of an ELP was known. The researchers masked weaker planets in the simulations to train the AI to infer unseen ELPs from the properties of detectable planets. The study acknowledges that the 99% precision applies to simulated environments and that real-world application will require empirical validation. 
