Solve Challenges with AI Companions
Welcome to Little Learning Machines—a live simulation where you train your real AI companions to solve challenges together, across whimsical islands. Shape their abilities and personalities while leveling up your skills with the games’ no-code AI system.
Welcome to the world of
Little Learning Machines
The Little Learning Machines game tells the story of Imogen, a crystal from the cosmos that crash lands on an island planet full of ancient inanimate robots. Imogen can use pieces of their crumbling form to bring them back to life in the hope of helping them traverse this strange new world.
These newly discovered robots are clueless by default, so they require training, which takes place in a magical place called the Training Cloud. This is where positive/negative reinforcement, presented through the intuitive love/fear rewards system, comes into play, and as the neural network of each robot develops, unique personalities and skills will begin to emerge.
Each quest brings a new themed island to explore, opportunities to train your robots in increasingly complex skills with different items, and more brain-bending challenges to overcome.
Meet Imogen
Discover Animo - Little Learning Machines
Train your robots with Love and Fear
The Training Cloud is where you train the real neural networks of your robots. This is where positive/negative reinforcement, presented through the game’s love/fear rewards system, comes into play. The better the training, the better they will perform on their own during your quests together.
Each neural network develops in a bespoke fashion, allowing unique skills and personalities to emerge
How Love and Fear Work
Animo (little learning machines) learn based on the rewards they receive. They are only able to collect rewards while training in the Training Cloud.
LOVE is a positive reward. FEAR is a negative reward. The sum of these two is their total reward.
As you move your mouse over the graph, you will see how much of each reward that the Animo has received.
How Animo Train
In this example, the Animo has collected three crystals for ten points each, but stepped on flowers twice subtracting ten points each time, and has held the watering can for five turns for one point each turn. Its total reward would be 15.
The Reward Graph
When you select an Animo in the Training Cloud, you will see its Reward Graph. In this graph:
LOVE is represented by rewards above the purple line
FEAR is represented by rewards below the purple line.
The total reward is represented by the white line.
When Animo train, because of reinforcement learning they will always try to optimize towards getting more total reward. In other words, they want to get the white line to increase as the training progress. They can do this by either minimizing their FEAR or increasing their LOVE.
The Training Cloud runs 10 simulations in the background at the same time the simulation you are monitoring is running. This means that while you only see one Animo on screen, there are additional versions of the Animo learning in the background. What you see on screen and in the graph are an average of those versions. So for example, when you hover over the graph and it says that the Animo collected 6.5 points, that means the Animo in the background collected anywhere between 5 and 8 crystals and 6.5 was the average among them.
Play at your own pace
As you spend time together, your little learning machines and their neural networks will evolve. Their personalities and skills will begin to emerge, and bonds will be developed with each other and with you! While playing and succeeding on quests together across island worlds, you will also evolve your skills with reinforcement learning.
Special Thanks!
We are so thankful for the generous support of Canada Media Fund and Ontario Creates. Without them, this innovation would not be possible. Thank you!
