A new study published in Nature reveals that artificial intelligence systems are developing an unexpected capability: the ability to reason about the physical world in ways that resemble human intuition. Researchers trained a neural network on vast datasets of simple physical interactions, such as objects falling, colliding, and rolling. The AI model, dubbed “PhysNet,” was then tested on scenarios it had never seen before. It successfully predicted outcomes like how a stack of blocks would topple or where a bouncing ball would land, demonstrating a form of intuitive physics previously thought to be uniquely human. The findings suggest that core aspects of common-sense reasoning might emerge from learning statistical patterns in data, challenging long-held views on the nature of intelligence. This research could lead to more robust and reliable AI systems for robotics and real-world interaction. For the complete details, read the full article at https://technologyreview.com/2024/05/15/physnet-ai-intuitive-physics.