Training vs Using a Model
An AI model has two stages in its life. First it is trained, which is slow and expensive. Then it is used, which is fast and happens every time you tap a button. Knowing the difference helps you understand how AI apps work.
New words
- Training
- The stage when a model learns by studying many examples.
- Inference
- The stage when a trained model makes a prediction for a new input.
- Prediction
- The answer a model gives for an input it is shown.
- Train/test split
- Keeping some examples hidden from the model, to use later as its exam.
- Overfitting
- When a model memorises its practice examples and fails on new ones.
- Deploy
- To put a finished model inside an app so people can use it.
- Retrain
- To train a model again on fresh data, because the old one is out of date.
What you will learn
- Learning to cook, and cooking dinner
- Training: learning from lots of examples
- Testing: the exam it has never seen
- Try it: memoriser or learner?
- Inference: using the trained model
- Try it: the model lifecycle
- Where you see it every day
- Explain it to a friend
- Key takeaways
10 quiz questions at the end.
The full lesson, with interactive demos and the quiz, opens after you sign in with Google. It is free.