A new study published in Nature reveals that artificial intelligence systems are demonstrating an unexpected ability to develop internal models of the physical world, similar to how human infants learn. Researchers trained a deep learning system on thousands of hours of video and sensor data depicting simple object interactions. The AI, without explicit programming for physics, began to predict outcomes like object stability and motion trajectories with high accuracy. This emergent capability suggests that foundational concepts of intuitive physics may be a natural byproduct of learning from sensory data in complex neural networks. The findings could influence future approaches to machine learning and developmental robotics. Read the full article at https://example.com/full-article.