What are some AI/ML concepts or problems you found difficult while learning?

While learning or working in machine learning and deep learning, I’ve noticed that the most useful learning comes from real doubts, debugging issues, and practical problems. I’m curious to know what kinds of questions people here have faced, especially in areas like: * Model training issues * Overfitting/underfitting * Transformers and BERT * Data preprocessing * Optimization problems Examples: * Why does my model overfit even after regularization? * What happens if the learning rate is too high? * Why is my model not converging? Would be interesting to hear the kinds of challenges others have encountered.