MIT and UBC researchers develop WorldParticle, one AI model for physical simulation

Researchers at MIT, UBC and collaborators have developed WorldParticle, a deep-learning system that simulates many physical processes with one architecture instead of separate tools for each.

The system represents materials as particles and predicts their motion one step at a time. According to MIT News, it was demonstrated on six phenomena, including deformable objects, granular materials and proteins.

The researchers say the model could help engineers simulate phenomena ranging from sliding sand to folding proteins.