Human
TOKYO: In a landmark achievement, researchers have used artificial intelligence to simulate every estimated star in our galaxy, offering the most detailed virtual model of the Milky Way ever created. The simulation, representing more than 100 billion individual stars, was unveiled in November 2025 by a team led by RIKEN Center for Interdisciplinary Theoretical and Mathematical Sciences in Japan.
The breakthrough hinges on a hybrid approach: deep-learning neural networks trained on high-resolution simulations of stellar explosions (supernovae) now replace the most computationally intensive parts of traditional physics calculations. Whenever a star in the simulation goes supernova, the AI surrogate predicts how surrounding gas will behave, bypassing the need for tiny individual time-steps and allowing the simulation to efficiently track galaxy-wide dynamics alongside fine-scale events.
By adopting this AI-powered method, what would have taken conventional supercomputers roughly 36 years of computation can now be accomplished in days or months. The result is a "digital twin" of our galaxy: a dynamic model that captures the birth, motion, and death of stars across 10,000 years of simulated cosmic time.
Scientists say the simulation opens a new era for studying galactic evolution, from the formation of spiral arms to how supernovae shape star formation, chemical enrichment, and the overall structure of the Milky Way. As computing and AI continue to advance, this could become a standard tool for testing theories about how galaxies grow and evolve across cosmic time.
