Episode 26: How AI learns from examples

1. 1 Learning from examplesTraining means showing the tool lots of examples. It works out patterns from them.Try: Picture cards: cat, cat, dog, bird, cat.Google ML course: a model is trained on data and finds patterns through training.
2. 2 One example is rarely enoughOne cat photo can't show what makes a cat a cat. The tool needs many and varied examples.Try: Show one cat. Then a dog. It may still say cat.Our illustration, general good practice.
3. 3 Examples can carry answersEach example can come with the right answer, called a label. The tool compares its guess to the answer.Try: A photo, plus the word cat.Google ML course: labeled examples.
4. 4 Not a rule listNobody types a rule for every cat. The tool finds patterns in the examples instead.Try: Ask: did a person write rules, or were there examples?Many AI tools work this way, not all.
5. 5 Ask what it learned fromWhat a tool saw in training shapes what it gets right. Odd or new cases can still fool it.Try: Does it handle a cat in a costume?General good practice.

Tap a step when you have done it. Your ticks stay on this device.