#What are GANs?

In this module, students learn about a new kind of AI that generates art. Students participate in an activity around classifying colors and generating new colors. We use this analogy to discuss classifying algorithms vs generative algorithms, and discuss generative AI techniques. We introduce GANs as one such generative technique. Students witness many instances of GAN generated art in the GANs or not activity. Now that students have learned what GANs can make, they are now set up to learn how GANs work in the next module.

#Slideshow

This slideshow is the slide deck teaching the important concepts for this entire topic.

Get a slideshow copy.

#Activities

The following activities are supplementary activities that learners should complete to reinforce these important concepts. These activities are listed in the order we suggest you follow.

#Activity 1. Exploring GANs

• Links:
  1. Exploring GANs Worksheet
  2. Exploring GANs Slides
• Pre-Reqs: "What are GANs?" Slideshow
• Suggested Length: 30 min
• Tech Dependencies: Google Chrome
• Key Concepts: GANs
• Group Reqs: None
• Description: Students explore a variety of GANs and complete questions or activities for the GANs they explore.

#Activity 2. GANs or Not

• Links: GANs or Not Slides
• Pre-Reqs: "What are GANs?" Slideshow
• Suggested Length: 10 min
• Tech Dependencies: Google Chrome
• Key Concepts: GANs
• Group Reqs: None
• Description: Students identify whether or not various entities were generated by a GAN.

#Generator Vs. Discriminator

In this section, we look at how GANs work, examples of GANs and identify what went into creating them.

#Teacher’s Guide

Check out the Teacher’s Guide here.

#Slideshow

Get a slideshow copy.

#Activity: GAN Game

• Links: GAN Game
• Pre-Reqs: "How do GANs Work?" Slideshow
• Suggested Length: 30 min
• Tech Dependencies: Google Chrome
• Key Concepts: GAN, Generator, Discriminator
• Group Reqs: Students will need to be split into two groups. One group of students will be the "generator" with one instructor leading, and the other group of students will be the "discriminator", also with at least one instructor leading.
• Description: Students engage in an unplugged activity in pairs, when one of them tries to guess what image the other one has within a pixel map. The guessing student makes multiple attempts and gets feedback from their partner in order to try to reach the target image. We use these roles as analogies for Generator and Discriminator - the two competing parts of a GAN.