Beginner level · Module 2: How AI Learns from Data · Lesson 3 of 3
Why AI Makes Mistakes
Measure how often AI is right, understand the kinds of errors and why they happen.
In this lesson you will
- Explain test data and accuracy
- Identify common causes of AI errors
Key concepts
test dataaccuracyfalse positives and negativescauses of error
The lesson
To know how good a model is, test it on examples it never saw during training. Accuracy is the share it gets right — but not all errors are equal: missing a disease is worse than a false alarm.
Errors come from gaps in training data, wrong labels, unusual inputs, situations that have changed since training and tasks that are genuinely ambiguous.
No AI system is perfect. Responsible use means knowing the error rate, the kinds of errors and having people check important decisions.
Step by step
- Test your classifier on 20 new examples.
- Calculate its accuracy.
- Classify the errors by cause.
- Decide which errors would matter most in real use.
Practice activity
Write a short accuracy and error report for your classifier.
Expected result: A report with accuracy, error causes and their real-world importance.
Check your understanding
-
Why test a model on new data?
Show answer
To see how it performs on cases it has not memorised.
-
Name two causes of AI errors.
Show answer
For example gaps in training data and wrong labels.
Taught live by an instructor — online or in person. Free course lessons are for self-study and do not lead to a certificate — see registration options.