Can AI predict the universe's first stars from 21-cm radio signals?

Researchers built simulations of the Cosmic Dawn era—when the first stars formed—by training neural networks on detailed hydrodynamic models. They predicted the 21-cm radio signals these early galaxies would emit, finding that Population III stars (the very first) dominated early epochs while Population II stars took over later. The team forecasts these signals should be detectable with existing radio telescopes, offering a direct test of when cosmic reionization actually began.