Generalization
An AI model's ability to perform well on new, unseen data rather than just memorizing its training examples.
In Plain English
Generalization measures whether a trained AI system can apply what it learned to situations it has never encountered before. Think of it like learning to cook from recipes: a good cook doesn't just repeat the exact same dish perfectly, but can adapt techniques to new ingredients and variations. In AI, poor generalization happens when a model memorizes its training data so thoroughly it fails on anything slightly different—like a student who memorizes test answers but can't solve similar problems. Companies care deeply about generalization because an AI trained on last year's customer data needs to work well with this year's customers too.
💡Real-World Example
A spam filter trained on thousands of old emails learns to spot junk mail patterns. If it generalizes well, it catches new spam it has never seen before. If it doesn't generalize, it might work perfectly on emails from its training set but fail completely on today's spam, which uses different tricks and language.
Related Terms
What did you think of our explanation?