Consider a company that stores user data and trains machine learning models on it to improve the user experience. Now suppose a user deletes their account. While it is straightforward to remove their data from the database, their data may already have been incorporated into trained ML models. Is it possible to efficiently update the model so that it behaves as if it had never been trained on that user's data in the first place? This is the problem of machine unlearning. In this talk, we give a gentle introduction to this problem and discuss some key ideas and results from existing approaches.