Transfer learning is a machine learning approach in which knowledge learned from one task, dataset, or model is reused to improve performance on another task.
It is common in modern AI because pre-trained models can often be adapted more efficiently than training from scratch.