Exploring Neural Networks with Quick, Draw!

Introduce students to machine learning concepts by engaging them in a fun hands-on activity using Google’s Quick, Draw!

Assignment Details


AI Theme





Develop a foundational understanding of how neural networks can recognize images based on a user’s input
Reflect on the implications of training a neural network, data rights, and the iterative process behind developing a machine learning model


Machine Learning: A subfield of artificial intelligence where computers are trained to learn from data and improve their performance over time without being explicitly programmed for specific tasks. Instead of following predefined rules, machine learning algorithms analyze patterns in data to make predictions and recognize patterns. This process involves feeding the computer large amounts of data and allowing it to adjust its operations based on the results, much like how humans learn from experience.
Neural Network:
A machine learning program modeled after the human brain’s structure and function. It consists of interconnected layers of nodes, or “neurons,” that work together to process and analyze data. Each neuron takes input, processes it, and passes the output to the next layer. Neural networks are designed to recognize patterns, learn from examples, and make decisions based on complex data.


Similar to the classic word-guessing game Pictionary, Google’s Quick, Draw! challenges players to draw an object in a limited period of time with the goal of correctly guessing what the drawing represents. The key difference is that in Quick, Draw!, the thing doing the guessing is a neural network. Start this activity by providing an overview of Google’s Quick, Draw! as a tool that utilizes a machine learning model to recognize doodles and images. Explain that the model has been trained on millions of drawings created by everyday Internet users in order to identify common patterns and make predictions.

  1. Direct students to visit Quick, Draw! and spend 10-15 minutes familiarizing themselves with the platform by participating in several rounds of the drawing game. This step can be conducted individually, or you might set up a competition by dividing students into teams and keeping score of which team produces the highest number of correctly guessed doodles.
  2. After playing the game, have students visit the Quick, Draw! data set of over 50 million drawings and find representations of some of the images they were asked to doodle. Review them in groups and consider the detail (or lack thereof) in many of the drawings and how this impacts the neural network image recognition.
  3. After students have completed their exploration, initiate a group discussion by posing the following questions. Divide the students into groups of 3-4 and ask them to discuss at least three of these questions. Alternatively, if you wish to incorporate a take-home element into this activity, students might write a brief structured reflection on their experiences with using the Quick, Draw! platform.
    • How accurate was the AI model in recognizing your drawings? Were there any instances where the model struggled to identify your doodles? If so, why do you think that happened?
    • How do you think the AI model makes predictions about the drawings? What aspects of the drawings might the model use to determine what has been drawn?
    • Can you think of any real-world applications for image recognition technology similar to that used in Quick, Draw!? How might these applications benefit society?
    • Are there any ethical considerations or potential drawbacks associated with the deployment of image recognition technology in various contexts?
    • What do you make of Google’s decision to open-source millions of doodles that have been created by over 15 million players? How do data rights come into play when using a tool like Quick, Draw!? Who “owns” the drawings, and how might they be used beyond their initial purpose? Here you might encourage students to further investigate Sketch-RNN, a neural network that has learned to draw by being trained on the millions of doodles in the Quick, Draw! data set.
    • How does the diversity of the training data affect the AI model’s performance? Do you think the model would perform equally well across different cultures and artistic styles? Why or why not?
    • Consider the balance between human creativity and AI prediction. Do you think AI could ever truly understand or replicate human artistic intent? Why or why not? How do you think humans “understand” and how is that different from AI?


DuPont Analysis of Coca-Cola

Conduct a DuPont analysis of Coca-Cola using real-world data, comparing and contrasting your results with an analysis conducted by Microsoft’s chatbot Copilot.


How to Train Your AI Dragon

Empower educators to craft innovative activities by training AI tools to enhance student learning outcomes, generate educational content, and design evaluation methods.