A new study from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) demonstrates a significant advancement in AI’s ability to understand and reason about the physical world through video. The research introduces a framework where AI models are trained to watch videos and build intuitive ‘physics models’ of objects and their interactions. This allows the system to predict future states, such as how a stack of blocks might fall, or to infer past events, like which block was removed from a tower. The approach moves beyond pattern recognition, aiming to give machines a more human-like, common-sense understanding of basic physical laws. The work represents a step toward AI that can better interact with and navigate complex real-world environments. Read the full article at: https://technologyreview.com/2024/05/20/1093095/ai-learns-physics-by-watching-videos.