New approach enables physically consistent AI models of chaotic systems

Impact 7 MIT News - Artificial Intelligence

Summary

  • MIT researchers developed a new machine learning approach to enforce physics-based hard constraints on models of chaotic systems.
  • The method uses a modular projection layer to enforce energy dissipation, ensuring predictions remain bounded over long forecast windows.
  • Conventional machine learning models often drift into physically impossible predictions over time when simulating complex nonlinear systems.
  • The new technique preserves chaotic behavior while preventing runaway trajectories without requiring arbitrary limits.
  • The team published their findings and open-source code in the Proceedings of the National Academy of Sciences.

Why it matters

Enforcing physical constraints as hard guarantees prevents AI models from drifting into impossible predictions, improving reliability for weather forecasting, climate modeling, and industrial fluid dynamics.

Who should care