Embedded Files
Activities and Demonstrations
Below is a collection of short activities and interactive demonstrations can help students explore artificial intelligence concepts in engaging ways. Many of these tools are browser-based and require little setup, making them easy to incorporate into a single class period or short lesson.
If you are looking for structured one-hour lessons, explore the Hour of AI section.
AI Duet
AI Duet is a browser-based experiment created using Google's Magenta project. When a user plays notes on a keyboard, the AI responds with its own musical continuation of the melody.
AI Takeaway: This activity demonstrates generative AI. The system has learned musical patterns from large collections of music and can generate new sequences that respond to what a user plays.
In the Classroom: Students can experiment with short melodies and observe how the AI responds. Teachers can use this activity to introduce generative AI and discuss how machines can learn patterns from data to create new content.
Akinator
Akinator is an online guessing game where an AI system attempts to identify a character, object, or animal that a player is thinking about. The system asks a series of yes or no questions to narrow down the possibilities.
AI Takeaway: This activity demonstrates how AI systems use decision-making strategies and large datasets to make predictions. It also shows how AI can improve over time as more users interact with it.
In the Classroom: Students can play several rounds and analyze how the system chooses questions. Teachers can use the activity to discuss decision trees, probability, and how AI systems use information to make predictions.
ArtBot
ArtBot is an educational game where players navigate a dungeon to retrieve stolen art. To succeed, players must "train" an AI helper to recognize objects and find the best path through a maze, seeing firsthand how machine learning models are built and tested.
The AI Concept: This activity provides a playable demonstration of Supervised Learning (tagging images to help the bot recognize art) and Reinforcement Learning (rewarding or punishing the bot's pathfinding decisions). It shows that an AI’s success depends entirely on the quality and variety of the data provided by the trainer.
In the Classroom: ArtBot is an ideal "lab" for middle schoolers to move from being AI users to AI trainers. Teachers can use the game to spark discussions on Bias and Data Quality—for example, what happens if the bot is only trained to recognize "paintings" but then encounters a "sculpture" in the dungeon?
Google Quick, Draw!
Google Quick, Draw! is an online game where players have 20 seconds to sketch a specific object while a neural network tries to guess what they are drawing. The system analyzes thousands of sketches to learn how people represent different objects.
The AI Concept: This activity demonstrates how machine learning systems recognize patterns. Students can see how the AI improves as it compares their drawings with examples from its training data. It also highlights how AI systems rely on large datasets to make predictions.
In the Classroom: This works well as a short classroom demonstration or warm-up activity. After several rounds, teachers can lead a discussion about how the AI recognized certain drawings and why it sometimes makes incorrect guesses. This opens conversations about training data, pattern recognition, and model limitations.
Say What You See
Say What You See is an AI-powered game from Google Arts & Culture where players must describe an image in 120 characters or fewer. The goal is to write a prompt that allows an AI system to recreate the original image.
AI Takeaway: This activity demonstrates how generative AI systems respond to prompts. Students learn that the quality and clarity of a prompt strongly influence the output produced by the AI.
In the Classroom: This activity works well when introducing prompt writing. Students can experiment with different descriptions and discuss how small changes in wording affect the generated image.
Semantris
Semantris is a set of word association games powered by Google’s natural language processing. In the game, players type words that are related to the ones on the screen, and the AI ranks them based on how closely they are associated in its database.
AI Takeaway: This activity demonstrates how AI models understand language through mathematical relationships. It shows students that AI does not "understand" words like a human, but instead sees them as data points that are either close together or far apart in a "map" of language.
In the Classroom: This is a helpful tool for introducing the concept of how large language models (LLMs) organize information. Teachers can use it to discuss how AI calculates the relationship between different concepts and why it might group certain words together based on its training.
Survival of The Best Fit
Survival of the Best Fit is an educational simulation that puts students in the role of a hiring manager. They begin by hiring candidates manually and eventually transition to using an AI tool to help automate the process.
AI Takeaway: This activity illustrates the concept of algorithmic bias. It shows how an AI system can learn and amplify human biases present in training data, leading to unfair outcomes in real-world scenarios like hiring or admissions.
In the Classroom: This simulation is ideal for starting a conversation about AI ethics and responsibility. Teachers can lead a discussion about why the AI made certain choices and how human oversight is necessary to ensure that automated systems are fair and equitable.
Teachable Machine
Teachable Machine is a free web-based tool that allows users to train simple machine learning models without writing code. Students can create models that recognize images, sounds, or body poses using a webcam or microphone.
AI Takeaway: Students learn how machine learning systems are trained using labeled examples. They can observe how the quality and quantity of training data affect how well the model performs.
In the Classroom: Students can build a simple classifier in a short activity by training the computer to recognize different objects, gestures, or sounds. This works well as a small group exploration where students test their models and discuss how to improve accuracy by collecting better training data.
Page updated
Google Sites
Report abuse