Think Together · Start with a Short Story
The girl notices that the robot often gets things wrong before training but does better on cards it has never seen afterward. How can we observe this “learning”?
Lay Out the Process
- Before trainingRecord performance
- Provide dataTrain the model
- After trainingTest new examples
- CompareDid performance improve?
In machine learning, “learning” usually means using data to improve a system’s performance on a task.
The key is not claiming that it tried hard, but whether it performs better in a suitable evaluation. Working well on new examples is called generalization; memorizing practice questions is not enough.
This is a technical meaning that can be measured. It does not require a machine to have the same experiences as a person.
Now You Try
The robot has memorized every practice card, but cannot recognize a new card of the same kind. What might that tell us?
Ready? Open this to see the explanation
It may have memorized the training examples and failed to generalize. The training data and method need another look.
One Idea to Take Away
In machine learning, “learning” usually means using data to improve a system’s performance on a task.
Try covering the diagram above and explaining each step to someone in your family. If one step is hard to explain, return to that card and look again.