Human cognition has a ``large-scale first'' cognitive mechanism, therefore possesses adaptive multi-granularity description capabilities. This results in computational characteristics such as efficiency, robustness, and interpretability. Although most existing artificial intelligence learning methods have certain multi-granularity features, they do not fully align with the ``large-scale first'' cognitive mechanism. Multi-granularity granular-ball computing is an important model method developed in recent years. This method can use granular-balls of different sizes to adaptively represent and cover the sample space, and perform learning based on granular-balls. Since the number of coarse-grained "granular-ball" is smaller than the number of sample points, granular-ball computing is more efficient; the coarse-grained characteristics of granular-balls are less likely to be affected by fine-grained sample points, making them more robust; the multi-granularity structure of granular-balls can produce topological structures and coarse-grained descriptions, providing natural interpretability. Granular-ball computing has now been effectively extended to various fields of artificial intelligence, developing theoretical methods such as granular-ball classifiers, granular-ball clustering methods, granular-ball neural networks, granular-ball rough sets, and granular-ball evolutionary computation, significantly improving the efficiency, noise robustness, and interpretability of existing methods. It has good innovation, practicality, and development potential. This article provides a systematic introduction to these methods and analyzes the main problems currently faced by granular-ball computing, discussing both the primary applicable scenarios for granular-ball computing and offering references and suggestions for future researchers to improve this theory.