I didn't grasp the similarity between human and machine learning until I took a machine learning class at Berkeley. Listening to lectures or reading the textbook is akin to fitting a machine learning model: You receive just enough information to begin understanding, but not enough to master it. True mastery requires you to validate your knowledge through homework, quizzes, exams, iteratively tuning your approach based on feedback. Without this practice, similar to a machine, your learning is merely overfitting to training data - you might handle familar problems well, but struggle with new challenges.