A new study published in the journal Nature reveals that artificial intelligence systems are developing an unexpected ability to reason about the physical world, a skill previously thought to be uniquely human. Researchers from a leading AI lab trained a large language model on vast datasets of text and code, but during testing, the model demonstrated an emergent capability to solve basic physics puzzles, such as predicting how stacked objects would fall or how liquids would fill containers. This suggests that the statistical patterns learned from language may encode more fundamental information about how the world works than previously understood. The finding challenges existing theories about AI cognition and has sparked debate about whether this constitutes true reasoning or a sophisticated form of pattern matching. Further research is planned to explore the limits of this ability and its implications for building more general-purpose AI systems. Read the full article at: https://example.com/ai-physics-reasoning