Lesson 1: What is AI?
Learning with Huge Amounts of Data
During 2012 to the 2020s, neural networks and deep learning significantly improved performance of AI systems.
Neural networks are computer systems inspired by how the brain works. They learn by looking at lots of examples and adjusting tiny connections inside the network until they get better at the task. This all happens using some pretty complex maths.
These networks are made of layers, which are groups of the tiny connections. Each layer learns different things and works together with the other layers to complete a task. For example, in a neural network that learns to recognise a handwritten number from a photo, there could be three layers, where:
- the first layer recognises whether lines are curved or straight
- the second layer recognises the number of lines in the image
- the third (and last) layer identifies a number based on the information from the previous layers
Deep learning is when neural networks have many layers stacked on top of each other. More layers let the AI learn more complicated patterns than previously. Thus more complex maths!
AI developers used neural networks and deep learning to create AI that performed more complex tasks, such as recognising faces, understanding speech and spotting objects in photos. These things are easy for humans to do but had always been hard for computers.
At its core, deep learning is a bit like a child learning to recognise shapes, but on a massive scale and using maths instead of human intuition. When we say an AI "learns" from huge amounts of data, this means it is finding patterns and mathematical relationships within that information (similar to what our brains do).
Some examples of applications developed using deep learning include:
If you would like to learn more about these topics, the following videos from Code.org are worth watching: