Direct questions

8. What is the difference between machine-learning and generative-AI classes for children?

Ask what the learner will actually investigate: learning patterns from data, generating content with an existing model, or building software around that model. These can overlap, but they produce different learning evidence.

Generative AI creates content such as text, images or audio in response to prompts. Machine-learning education also encompasses training and evaluating models; Experience AI’s challenge, for example, asks learners to develop an application around a model they have trained. (gao.gov)

Proposed course emphasis Questions the child should learn to answer
Classification or prediction What examples trained the model? How does it perform on genuinely new examples?
Generative-AI literacy When is this output unreliable? What evidence would verify it?
AI application development What does our own code do, and what does the external model do?
Research-oriented work What is the question, baseline, evaluation method and limitation?

A useful technical check concerns data leakage: scikit-learn’s documentation warns that using information from test data during model development can produce overoptimistic performance estimates. Ask a research mentor to explain how the project avoids this. (scikit-learn.org)

Our purchasing recommendation is to reject a syllabus consisting only of tool names. Ask for concepts, activities and observable learning goals.

Sources cited on this page

  1. Science & Tech Spotlight: Generative AI | U.S. GAO ↗gao.gov
  2. 12. Common pitfalls and recommended practices — scikit-learn 1.9.1 documentation ↗scikit-learn.org