Memorized and still failed
A student memorizes the answers to old exams and fails the new one. AI makes this mistake too; it's called overfitting.
MEMORIZING
Remembering useless details
Overfitting is when a model fits its examples too tightly, remembering useless details instead of the real pattern.
EXAMPLE
Dogs always on grass
If AI only sees dogs on grass, it thinks grass means dog, and gets a dog on sand wrong.
FOR YOU
Varied examples
When you give AI examples, give a few different kinds, not copies of one.
This article is based on the video Overfitting from the Mark học AI channel. Watch the video (in Vietnamese) to see the animations.