Course 2, lesson 13 of 100, Ages 6+

Data is AI’s food

Why examples matter

Like I’m 5

Data is what AI eats. Feed it lots of good examples and it grows strong. Feed it junk and it gets things wrong.

The big idea

Data is information collected to learn from: photos, sounds, numbers and words. For AI, data is like food. The better the food, the healthier the learning.

If the data is messy, wrong or one-sided, the AI learns the wrong lessons. An AI trained only on photos of golden retrievers might decide every dog is golden. Good data is plentiful, correct and varied.

Examples

  • Weather records: Years of temperatures help AI forecast tomorrow.
  • Voice clips: Thousands of people saying 'yes' and 'no' teach a speaker to understand.
  • Bad data: If cat photos are wrongly labelled 'dog', the AI gets confused.

How it works

  1. Data means examples: pictures, words, sounds or numbers. AI learns by looking at huge amounts of it.
  2. The more examples, and the more different they are, the better AI gets at its job.
  3. Bad data makes bad AI. If every cat picture shows a ginger cat, AI may decide a grey cat isn’t a cat.

Check your understanding

What helps AI learn best?
Options: One perfect example; Lots of different examples; No examples at all.
Answer: Lots of different examples. Lots of varied examples teach AI what really matters, instead of one-off details.
What makes data good for learning?
Options: It's plentiful, correct and varied; It's tiny and all the same; It's full of mistakes.
Answer: It's plentiful, correct and varied. More examples, correct labels and plenty of variety lead to better learning.

Remember

Good, varied data makes good AI.

Talk about it

What examples would you show an AI to teach it about dogs?

Go deeper

Training data shapes everything a model can do. Engineers clean it, label it, split it into training and test sets, and check it for gaps and errors. ‘Garbage in, garbage out’ is still the first rule of machine learning.