What Is AI?
What today's AI is, what it can't do, and why it sometimes states false things with confidence.
AI tutor is turned off for this class
Use the AI10 lectures, notes, and assignments below to keep learning.
Lesson overview
What today's AI is, what it can't do, and why it sometimes states false things with confidence.
Teacher script · 45 min
- 0–5
Hook
Modern AI is not a brain and does not 'know' things the way you do. It has seen enormous amounts of text or images and learned which patterns tend to follow which. When it answers, it is predicting what usually comes next. That is powerful, and it is also why it can sound sure while being wrong.
- 5–15
Explore
Students do the activity in pairs: Sort ten photos into 'dog' and 'not dog' and write the rule you used. Now imagine finding that rule from a million photos without being told it. That process is called training.
- 15–30
Explain
Because AI predicts likely patterns, it has predictable weaknesses. It can HALLUCINATE (state something false but plausible), because 'plausible' is exactly what it optimizes for. It reflects BIAS in its training data. And it has no built-in sense of truth. Using it well means treating it as a fast, fallible assistant you always check, not an oracle.
- 30–40
Connect to the summit
Read aloud the "Where this course leads" section below. Ask students which ideas from today they recognize in it.
- 40–45
Check
Run the formative check below. Anyone who can explain a key term in their own words has it.
Student activity
Sort ten photos into 'dog' and 'not dog' and write the rule you used. Now imagine finding that rule from a million photos without being told it. That process is called training.
Slides
Formative check
- 1.In your own words, what is "Machine learning"? (Looking for: Software that finds patterns from examples instead of following rules a person wrote.)
- 2.In your own words, what is "Training data"? (Looking for: The examples an AI learns from. Its strengths and biases both come from here.)
- 3.In your own words, what is "Prediction"? (Looking for: AI's core move: guessing the most likely next word, label, or pixel.)
Carry-away concepts
- Machine learning
- Software that finds patterns from examples instead of following rules a person wrote.
- Training data
- The examples an AI learns from. Its strengths and biases both come from here.
- Prediction
- AI's core move: guessing the most likely next word, label, or pixel.
- Hallucination
- When AI states something false but plausible-sounding, because plausibility is what it optimizes.
Where this course leads
From here the university courses open up: how machines search, learn, see, and use language (CS221, CS230, CS131, CS124). Before those, it's enough to know that AI learns patterns, predicts, and can be wrong, so you neither dismiss it nor over-trust it.
This is a beginner course written for Feynman, not a Stanford course.