>A new computer model has been developed that can match humans' ability to predict how objects will move. The model could be used to improve the safety of self-driving cars and other autonomous systems, as well as for simulating objects in video games and movies.

> Humans predict object movement by drawing on visual and physical knowledge, as well as common sense. The computer model, developed by researchers at Stanford University, combines machine learning and physics-based simulation to achieve human-like performance on a range of tasks, including predicting how a ball will bounce off a table or how a liquid will flow into a glass.

> "Our model can simulate the world around us in a way that is intuitive to humans," said Peter Abbeel, a professor of computer science at Stanford and director of the Stanford Artificial Intelligence Laboratory. "This opens up a wide range of possibilities for new applications that rely on accurate object prediction, such as self-driving cars and video games."

> The computer model uses a combination of convolutional neural networks (CNNs), which are artificial neural networks that can process spatial information, and a physics-based engine to simulate the movement of objects. The CNNs are used to extract features from the visual input, such as the shape and texture of an object, and the physics-based engine is used to simulate how the object will move based on those features.

> The model was trained on a large dataset of human motion capture data, which allowed it to learn how humans predict the movement of objects. The researchers found that the model could achieve human-like performance on a range of tasks, including predicting the path of a ball, the trajectory of a liquid, and the motion of a human hand.

> "We hope that our model can help to bridge the gap between human intuition and machine learning," Abbeel said. "By combining the best of both worlds, we can create autonomous systems that are more safe, efficient, and user-friendly."