Computer-generated animals! + more STEM activities
A curation of weekly projects and resources from Robotics & Beyond.
Hi! I’m journalist Caroline Delbert and this is The Toolkit from Robotics & Beyond 501(c)3. Our goal every Saturday morning at 10 AM is to curate some videos, projects, and products that help kids (and parents) get inspired by science, technology, engineering, math, and creative design.
This week we’re going to talk about animals, especially ones invented by computers. Wait, what?! Click on the projects below or read on for details.
Beginner: Simon’s Cat, a YouTube cartoon series (video)
Intermediate: Ganimals, MIT’s tool to generate animals with AI (platform)
Advanced: Remixing words (video)
How does a computer know what an animal is?
Does your family have pets? I have two cats, and watching them do their normal cat stuff makes me really happy every day. Scientists have studied why we think animals are cute, but they still don’t quite understand. People have kept pets and drawn pictures of animals for many thousands of years. Today we’re going to talk about how animals look, how humans think, and how computers try to think about animals.
All three age groups will look at different stylized animals and languages, including MIT Media Lab’s Ganimals platform, a website where you can generate pretend animals and then feed and care for them. These Ganimals are made by a process called machine-learning: the computers use feedback about the pretend animals to do a better job making more pretend animals. (GAN means “general adversarial network,” it’s a little like two AIs competing to find the best solution to a pattern.) Some Ganimals are really cute, and some are a little scary or creepy. (Young kids might find them frightening.)
The human brain does an incredible job at putting entire ideas together. We can look into the fridge and know right away if there’s something we want to eat. In a Where’s Waldo book, we can pick Waldo out of the crowd. Our brains are great at recognizing objects and patterns, but computers don’t have eyes, or the same parts and sections our brains have. They can only do the jobs people have taught them to do.
Take a look at this GAN-generated cat (from a Twitter account dedicated to GAN cats). It looks kind of like a cartoon, right? And even though it’s not a realistic cat, our brains and eyes tell us that’s what it’s supposed to be. The Ganimal cat is one way computers are learning to recognize different kinds of animals. That means future computers could do a better job doing things like monitoring animals at the zoo or counting endangered animals in the wild.
Ages 5-8: Simon’s Cat
Simon’s Cat, a short cartoon series on YouTube with more than 200 episodes, has very simple artwork, like the outlines you see in a coloring book. The cat in the cartoon doesn’t really look like a real cat at all, but when we watch, we know it’s a cat. That’s because we see cues like triangle ears and a playful tail, we hear the cat meowing and purring, and our brains tell us the overall pattern is a cat.
A computer doesn’t know what a cat is. That means a human can make a program to try to teach the computer about things like pointy ears and playful tails — the “symbols” of a cat. Over time, the program can get better and better until it’s pretty good at “seeing” some kinds of cats.
Ages 8-12: Computer-generated animals from MIT
First, watch this introductory video (3 min):
Using the Ganimals tool from MIT, kids can see some examples of computer-generated animals. The Ganimals are made by combining two real pictures of animals, and the computer tries to find a balance between the two that looks like it could be a real animal, too, as measured by its algorithm.
With Ganimals, you can generate new animals and care for them like digital pets. Experiment with the platform, then try drawing some pretend animals of your own by combining different things, like a bird with a cat’s tail or a teddy bear with pointy ears. Do your pretend animals look more like one animal than the other? What does your brain say?
Ages 12-15: Why letter order doesn’t really matter
Watch this video from HowStuffWorks (4 min):
There’s a meme that comes back every year or two about jumbled letters and how our brains can still read them as words. When we type, we enter characters like letters and spaces just one at a time, but when we read, our brains consider whole words at once.
That means, like a handful of change or Lego bricks, the order of the letters doesn’t matter very much. When we look at a scrambly picture of a computer animal, our brains are doing something similar by taking in the whole picture at once and making sense of it. This ability to quickly, and mostly correctly, identify patterns of many kinds and on almost any scale is something that makes humans unique—and something computer programmers try to emulate.
Send us feedback:
Tell us what you like or don’t like about this newsletter! Or send us what your family is looking for and we’ll curate an activity list for you. Email toolkitnewsletter@gmail.com.
If you draw some pretend animals, please share! Tag us on Twitter @toolkitstem, share on our Facebook page, or tag @roboticsbeyond on Instagram with the results!
Wow that’s cool!
(A random thing we think is cool) Livestream some birds: You can watch a pair of falcons and their babies hanging out in their nest, birds visiting feeders in a special laboratory, and baby owls that will learn how to fly soon.
The Toolkit is written by Caroline Delbert and produced by Robotics & Beyond
Join us on Twitter, Facebook, or Instagram
Learn more: stemtoolkit.org