AI Lessons
Lesson 2: AI and Images - How does AI recognise images?

Purpose: To understand how AI recognises images by identifying patterns and similarities from past examples.

No. of Classes

1 - (Time : 1 hour 30 minutes, Laptops/desktops : 10, Students strength : 15 to 20).

Materials Required

Laptop / Desktop with Internet connections / Wi-Fi.

Curriculum Content

1. AI Literacy

  • 1.3 Using AI tools

2. AI and ML Concepts - A Simplified Overview

  • 2.3 Pattern recognition
Prior knowledge
  • Browsers
  • Drawing tools
  • Lesson 1 - AI predicts from patterns rather than understanding
Exercises

Exercise (1) - Unplugged activities


  • Part 1: Identify a simple drawing by observing its key features. Discuss how we recognise images using visual patterns. Click here to open the instruction document for this part.
  • Part 2: Repeat the exercise with a few abstract lines and shapes. Discuss how we use patterns to identify objects. Click here to open the instruction document for this part.

Exercise (2)


  • Part 1: Use AutoDraw to explore how AI recognises drawings. Observe how AI matches a rough drawing with similar objects. Click here to view the sample drawing for this activity.
  • Part 2: Use Google Lens to explore how AI identifies photographs. Observe how AI matches a photo with similar objects. Click here to open the pictures to scan for this activity.

Solutions


Teacher's Instruction:
  1. Emphasise the words ‘recognise’ and ‘pattern’.
  2. Explain to students that AI guesses the best matching pattern.
  3. Explain why messy images make it difficult for AI to recognise objects accurately and demo that using google lens activity.
  4. Reiterate that AI guesses based on probability, not understanding.
  5. Google Lens looks up images that already exist rather than creating a new one. Note this difference for the students.
  6. Go through the instruction documents for Exercise 1 and Exercise 2 before taking the lesson.
  7. Points to Ponder:
    • Show a photo of a real cat, Tom from Tom and Jerry, another Ghibli drawing of a cat etc and ask children to recognise the animal. Children can recognise it is a cat from any of these images. Can AI do the same? Why or why not? Discuss.
    • If AI sees 1000 pictures of cats, does that guarantee it recognizes all cat images?