The Central Problem of ML
Here's the challenge: we want models that work on NEW data they've never seen before. But we can only train on OLD data we do have.
🎯 The Goal: Make good predictions on future, unseen data
But the model only has access to past data during training.