Fun Machine Learning Resources and Activities for Kids
A review of our favorite online ML learning resources as well as some DIY activities you can try at home
As with so many other families and companies during the COVID-19 pandemic, Anno.Ai team members have had to adjust to new patterns of working, child care, and schooling for our children. One of the many ways we’ve sought to support each other and our Anno.Ai family during this time is by sharing educational activities and resources with each other.
As an AI/ML company, many of these resources and ideas have included ways to involve our kids in fun machine learning and computer programming activities. Here, we wanted to share a few of our favorites as well as some DIY activities you can try at home. Enjoy!
Online ML Learning Resources for Kids
Here are some of our favorite online platforms for training models and exploring machine learning. These have simple user interfaces and are designed for kids to learn and experiment.
Google Teachable Machine
The Teachable Machine website provides an easy to use interface for training image, sound, and pose classification models. No login is required to get started. Training data files can be loaded directly from your computer or from your computer’s webcam or microphone. The initial view keeps things simple for the user, but if they want, kids can dig deeper to explore advanced options like adjusting the number of epochs, batch size, and learning rate, or get tips on common ML vocabulary. Models can be exported to use in other projects, and the FAQ includes links to read more about fairness and inclusion in ML.
IBM’s Machine Learning for Kids Activity Kit
IBM’s Machine Learning for Kids website provides a range of free step-by-step guides and hands-on model building activities designed to support both teachers and parents who are involved with teaching their kids at home. The online interface provides an easy way to train a model to recognize text, images, numbers, voices, and sounds. To get started, you will need to create a free account on IBM Cloud and an API key for Watson cloud services to train machine learning models.
Google’s Quick, Draw!
The Quick, Draw! website is a fun game for all ages to draw simple sketches of everyday objects and contribute those to training a neural network. The data can also be searched and displayed by class, which we’ve found is a fun way for even toddlers and younger kids to interact with the data and point out drawings of their favorite objects (e.g., teddy-bears, bulldozers, tractors, cats, and dogs).
The Machine Learning Playground is a web-based demo that lets you draw point patterns, configure parameters, and analyze/classify the data using a range of different ML algorithms (KNN, Perceptron, SVN, ANN, and Decision Tree). The ML Playground is probably most suited for older kids who have some basic familiarity with these terms and want to explore and visualize how the algorithms behave on real data.
DIY Data Collection and Models
At Anno.Ai, we occasionally get our extended families involved in data collection, especially around topics and objects that are interesting for kids. For example, during a project on handwritten text recognition, we had some of our team members and relatives help with creating test data by writing out the preamble to the United States Constitution. This also ended up being a fun way for our youngest team members to practice handwriting, learn a little more about the history of the Constitution, and get involved in the ML process!
A fun DIY activity you can do with your kids is to have them collect images of things that they’re interested in (horses, garbage trucks, LEGOs, cartoon or game characters, kittens… the list goes on) and train their own image classifier using one of the in-browser platforms such as Teachable Machine. We’ve provided some images that you can use to get started with a LEGO detection and color sorting model here.
As parents interested and involved in AI/ML, this extra time with our kids has also been an opportunity to observe and focus more on how children learn, and the ways we can advance AI/ML by integrating principals from human learning and cognition. For further reading on these types of approaches, we recommend Alison Gopnik’s work on learning to learn, and have included some additional references below.
What are some of your favorite learning and educational resources for getting kids engaged in machine learning? We’d love to hear from you — drop us a note in the comments or send us a message on Twitter.
Bambach, S., et al. (2018) Toddler-Inspired Visual Object Learning. NIPS 2018.
Stretcu, O., et al. (2020) Coarse-to-Fine Curriculum Learning for Classification. Published as a workshop paper at “Bridging AI and Cognitive Science”, ICLR 2020.