5 Deep learning
Deep learning is a subset of machine learning that utilises artificial neural networks to excel at identifying patterns. Deep learning uses large datasets and complex algorithms to achieve higher levels of accuracy and performance.
In the video, you’ll learn how deep learning expands what AI is capable of and how it’s changing technology today.
Download this video clip.Video player: Video 8
Transcript: Video 8
Imagine you are standing at the edge of a data ocean, an ocean that is growing every second with waves of information from all around the world.
Traditional machine learning models are like small boats, they can only handle so much before they start to sink. But what if we had a bigger boat, like a ship?
Well, that is deep learning. Deep Learning is our powerful ship, designed to navigate the vast ocean of data.
It is inspired by the most complex system we know - the human brain.
Just as our brain consists of billions of interconnected neurons working together to make sense of the world around us, deep learning uses neural networks to learn from data and make informed predictions.
Have you ever wondered how the human brain works? How do we learn?
Think back to when you were a child. You probably played games that involved image recognition at school.
Remember those cards? You would decide if it was a dog or a cat, and your teacher would confirm.
This is similar to how a computer learns. The neural network makes assumptions and can be, let’s say, 70% certain that the image is correct or not.
Instead of guessing, it adjusts its parameters and retrains over time.
So, why does this matter?
Well, the more data the computer has to train on, the faster it will be able to correctly recognise an image- whether it is a dog, a cat or even a flower.
This is why the topic of data volume is so important. Now, let’s think about cooking.
When you start learning, you begin with simple recipes, like frying an egg or cooking rice. Each time you cook, you learn something new - how high to turn the heat, how long to cook the eggs, how much water you should put in your rice.
Over time, you become more proficient, and you can cook these dishes without even thinking about it. This is like to how traditional machine learning works.
But what if you want to learn to cook a complex dish, like fried chicken? There are so many variables to consider - how you bread the chicken, what temperature you fry it and what sort of oil you use.
It is not enough to just practise; you need to understand how all these factors interact. This is where deep learning comes in.
Deep learning, like frying chicken, involves a lot of trial and error. The neural network makes an assumption (or a guess), checks how close it was to the right answer, and then adjusts its parameters for the next guess. This process is repeated over and over again, each time getting closer to the correct answer.
Just like having more recipes can enhance your cooking skills, the more data the computer has to train on, the better it will be at making accurate predictions.
And as a child learns to recognise images from animal cards, a computer can learn to generate human language.
Another branch of AI is natural language processing, or NLP, which uses similar principles as deep learning but focuses on language.
The more textual data a computer learns from, the more proficient it becomes at producing human-like language.
Video 8
Interactive feature not available in single page view (see it in standard view).