World Model in plain English.

A world model is an AI system trained to predict how an environment — physical or simulated — changes over time in response to actions, rather than simply predicting the next word in text. Instead of learning "what word comes next," a world model learns "what happens next": if a ball rolls off a table, where does it land; if a robot arm moves this way, what does the scene look like afterward. World models underpin self-driving simulation, robotics training (where testing in a simulated world model is far cheaper and safer than testing in the real world), and video generators like Sora 2 and Veo 3, which implicitly learn a rough world model of physics and object permanence to produce coherent video. Genuinely robust world models — ones that generalize to situations very different from their training data — remain an open research problem and are widely seen as a missing piece on the road to more general AI.