Building brains like ours

Computer scientists have invented algorithms for learning: ways a computer can learn to do new things just by following rules. One way is to copy the way animals learn: copy their brains (but without blood and gore).

Our brains are made of billions of cells called neurons.

neuron

Each is connected to lots of others. They work by sending messages to each other. Each neuron fires (sends a message) only when it gets enough messages from other neurons.

Everything you think, feel and do (whether being happy, reading this magazine or catching a ball) involves neurons sending messages to each other. When we learn, our neurons are each changing how easily they fire. A neural network is a program based on an algorithm that mimics neurons and their messages. To teach a neural network to tell cats from dogs, say, you show it lots of pictures, telling it which are cats and which dogs.

With each picture it changes when its neurons fire so that eventually cat-like things make the ‘like a cat’ neuron fire, and dog-like things make the ‘like a dog’ neuron fire.


Clever critterness: learning like us


Fun to Do: Doodle Game

Have a friend draw something and try to work out what it is as quickly as possible.Work out what bits of objects your neurons are spotting as you work it out….ears? …a window?… a sail?


Cute RobotMore AI stuff →