A new study published in 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 MIT and Stanford trained a large language model on vast amounts of text and code, then tested it on classic physics puzzles. The AI demonstrated a surprising capacity to infer cause and effect, predict object trajectories, and understand basic mechanical concepts, despite never having been explicitly programmed with physical laws. This emergent capability suggests that scaling up AI models may lead to more generalized intelligence, though experts caution that the reasoning is still shallow compared to human cognition. The findings could accelerate the development of robots and AI assistants that better interact with the real world. For the full details, read the complete article at https://technologyreview.com/2024/05/ai-physical-reasoning-study.