Teaching Kids about Artificial Intelligence

August 13, 2019

#Community #Kids AI for Kids Coding for Kids Postcards from a Creative Coder

This is part of our a series – Postcards from A Creative Coder – Dispatches from Silicon Valley – by GP, Chief Tinkerer at Saturday Kids and a creative technologist based in Silicon Valley. Through this series, GP shares his insights on innovations at the intersection of tech, education play. Check out his first piece exploring inquiry-based learning for kids here.

What does A.I. have to do with our kids?

I was reading an issue of the MIT Technology Review which included an interesting list of 35 innovators under 35 years old. I was curious to see what they are working on, so I did a quick tally of their areas:

Artificial Intelligence	11
Biotech and medicine	10
Robotics	4
Energy	3
Materials	3
All others	4

Note: Some of the innovators worked in more than one area, and I have removed the overlaps.

Clearly, Artificial Intelligence is a key area of focus for young innovators. If this is the case, how can we prepare our very youngest innovators – our children – to understand and use this technology?

Image via www.vpnsrus.com

In this article, I’ll go into some background about A.I. and why it matters for our children to start to understand it.

What is Artificial Intelligence?

Artificial Intelligence (A.I.) is the ability for computers to do things that are considered attributes of intelligence: processing language, understanding pictures, detecting patterns, etc. In the past few years, the combination of the Internet and ever increasing computing power have enabled an approach to A.I. based on statistical analysis of enormous data sets, allowing the development of accurate predictive models.

These capabilities are now being applied in many different applications – from recommending purchases, to self-driving cars.

Why A.I. matters for our kids

The environment in which our children are growing up includes instances of A.I. driven devices and services. As of late 2018, nearly a quarter of US households owned a “smart speaker” like Amazon Alexa or Google Home. What children think about these devices is quite remarkable.

There is a pragmatic reason for teaching kids about A.I. which is its direct bearing on employability and future career success. This is a powerful way of robot-proofing your kids, as Dr. Vivienne Ming so eloquently puts it in her TED talk.

However, a more profound reason is equipping them with the ability to understand what A.I. is (or isn’t) capable of, and some of the basic principles behind how they operate. Some of the issues raised are profound, since exposure to A.I. will affect children’s mental models of what intelligence is, and how it manifests. Young children in particular tend to overestimate the capabilities of A.I. when they encounter it.

Teaching children how A.I. works is a way to “pull back the curtain” and enable them to make more accurate critical judgements.

Tools for kids to learn about A.I.

There are some great online resources available, suitable for every age group – from young children to teenagers. Many of these are free of charge, or very affordable. Here are some you can start with:

Cognimates

Cognimates is a project from the MIT Media Lab, and it extends the Scratch block-based programming system to include A.I. tools and capabilities.

Cognimates is extremely comprehensive. In addition to all the capabilities of Scratch, it includes blocks for typical A.I. applications like recognizing speech, translation, computer vision. It can interface with the real world using devices like the micro:bit, which opens the door to a variety of homebrew robots, among other things.

Moreover, it includes features for actually training models for computer vision and textual analysis.

The Cognimates team has done extensive research on A.I. and children, and their papers are worth reading for educators and parents.

Machine Learning For Kids

Machine Learning For Kids is another extremely impressive resource. Like Cognimates, some exercises are built using the Scratch block-based programming system. However, it uses the IBM Watson A.I. system behind the scenes to power applications, and some exercises use the Python programming language.

Like Cognimates, it goes “behind the scenes” to explain the process of training models, and using them in an application. The worksheets are very thorough, and support students and teachers who are not experts in the subject.

The A.I. Family Challenge is a free program that uses the Machine Learning For Kids system in a structured way, to get families and communities working together to apply A.I.

A.I. Experiments With Google

This collection of browser-based A.I. Experiments developed by Google runs entirely in your web browser, and is a good introduction to the applications of A.I. The experiments cover things like learning, music, drawing and a variety of other creative areas. Of course, they use Google’s powerful A.I. systems.

Unlike the previous two resources, these are mostly finished examples of A.I. in use. However, they are very fun to use, and could be a great starting point and inspiration for children using other systems.

This video is a good overview of the experiments: https://youtu.be/oOwfiYnRi5c

Google AIY projects

A combination of “AI” + “DIY”, these are hardware kits that allow for easy (and affordable) experimentation with computer vision and language processing. They are based on the Raspberry Pi computer, and require some familiarity with Python programming, so they are probably a good project for older kids.

One thing I like about them is that they use a cardboard enclosure, so it is very easy to customize them and add them to other projects, like a robot or a home device.

Tensorflow Playground and Machine Learning Crash Course

Google is clearly very interested in having as many people as possible learn about machine learning and A.I., so they’ve made available a lot of interesting tools. One of these is the Tensorflow Playground, which allows you “tinker” with a Neural Network.

Related to this, and for more advanced students, the Machine Learning Crash Course is a good way to become genuinely knowledgeable.

Zümi

Zümi is a very interesting project, basically a tiny self-driving car that uses real machine learning tools and APIs for programming. While it is still not released (shipping sometime in early August 2019!) it looks very promising, and claims to use Google’s Tensorflow system.

Getting started

With all these resources, what is a good way to start? There is certainly a lot to digest, so you don’t want to jump straight in. Here’s a progression that could be effective:

Start with relevant questions and projects: A great way to learn about new tools is to focus on a goal, something you want to achieve. Inquiry-based teaching programs start by identifying a problem to solve first, even before learning to use tools.
Build some basic skills: Many of these systems use some of the same basic tools and skills. Developing familiarity with Scratch (block-based) or Python programming is a good way to build a foundation.
Experiment with A.I. applications: Using some of the online experiments or even the services and devices that are commercially available is a good way to start understanding what is possible.
Get started with A.I. coding: Depending on the student’s age and abilities, pick one of the programming systems, like Cognimates, and set out to solve some real-world problems!

…

Did you find this article useful? If you have any comments or questions to share with us or GP, we’d love to hear from you! Let’s connect over email, here.