Machine Learning (not only for) for Kids
There is only one effective learning principle when it comes to machine learning: "Hands-on learning is key." Neither kids nor adults are able to develop a clear understanding what machine learning is and how it works without trying it for themselves, ideally with use cases that are interesting and relevant to the individual.
|
Try These Experiments and Games
BERT for Kids - Visual tutorial on how to user the BERT language model in Scratch (Scratch 3.0 plugin for Bert and Wikipedia were both created by Dale Lane.) Evolution - Kids can build simple beings and see them evolve over time Jurassic-1 - Kids can ask questions to the Jurassic-1 language model |
The Right Place to Start: Dale Lane's MachineLearningForKids.co.uk
MachineLearningforKids.co.uk is super simple, but at the same time it brings a lot of the machine learning power of IBM Watson to the table. Developed and maintained by Dale Lane from IBM, the site enables kids or basically any age to visually train their own machine algorithm and then use the resulting inference model to solve simple or complex challenges in the visual and fully browser-based Scratch programming language. Dale's concept is as simple as it is brilliant, as he shows kids in one go how to train a machine learning model and then how to use it in their own Scratch program. This trains kids to naturally understand the boundaries and challenges of machine learning, so that they can help their parents see through the marketing jargon of today's software and hardware vendors.
MachineLearningforKids.co.uk is super simple, but at the same time it brings a lot of the machine learning power of IBM Watson to the table. Developed and maintained by Dale Lane from IBM, the site enables kids or basically any age to visually train their own machine algorithm and then use the resulting inference model to solve simple or complex challenges in the visual and fully browser-based Scratch programming language. Dale's concept is as simple as it is brilliant, as he shows kids in one go how to train a machine learning model and then how to use it in their own Scratch program. This trains kids to naturally understand the boundaries and challenges of machine learning, so that they can help their parents see through the marketing jargon of today's software and hardware vendors.
How it works: Kids and adults can train and start using their own machine learning models in 3 quick steps:
1. Find examples of text, images, or number patters the machine learning model should recognize
2. Push the button to train the model (it's that easy)
3. Use the model instantly via a Scratch extension
Examples of what you can do: It's powerful stuff
On his website, Dale provides a wealth of clever kid-friendly work sheets that at the same time teach kids how grown ups create enterprise-grade machine learning models. And most importantly, kids learn at young age that there is no such thing as artificial intelligence. All we do is provide examples of different concepts to the machine and hope that these examples are sufficiently diverse and comprehensive to help the algorithm create an inference model of the desired reliability.
Tell apart friendly from rude text inputs and use Scratch to conditionally display a happy or sad face
Use your webcam and train a model that distinguishes between happy and angry faces by providing 10 examples of each. Use scratch to ring a bell every time your face looks angry
Build a simple chat bot that recognizes and answers natural language questions.
Create an Amazon-like recommendation engine for books or movies
Train a model to play Tic Tac Toe
1. Find examples of text, images, or number patters the machine learning model should recognize
2. Push the button to train the model (it's that easy)
3. Use the model instantly via a Scratch extension
Examples of what you can do: It's powerful stuff
On his website, Dale provides a wealth of clever kid-friendly work sheets that at the same time teach kids how grown ups create enterprise-grade machine learning models. And most importantly, kids learn at young age that there is no such thing as artificial intelligence. All we do is provide examples of different concepts to the machine and hope that these examples are sufficiently diverse and comprehensive to help the algorithm create an inference model of the desired reliability.
Tell apart friendly from rude text inputs and use Scratch to conditionally display a happy or sad face
Use your webcam and train a model that distinguishes between happy and angry faces by providing 10 examples of each. Use scratch to ring a bell every time your face looks angry
Build a simple chat bot that recognizes and answers natural language questions.
Create an Amazon-like recommendation engine for books or movies
Train a model to play Tic Tac Toe
Pros:
Kids can see cause and effect within minutes Fantastically simple user interface for training machine learning models Kids can gradually observe how well their model works based on different sets of training data Kids see the limits of machine learning, as they will experience non-sensical results, as we experience them within our own enterprise context Put your own models to work in MIT's Scratch visual programming language. EMA Quick Take: Dale's MachineLearningForKids.co.uk website is a "must try" for everyone who wants to teach kids about AI/ML. Start by reading Dale's excellent blog post full of examples and ideas for teaching machine learning concepts to kids in a fun and non frustrating manner. |
Cons:
Account creation is not automatic |