Artificial Intelligence

What is Artificial Intelligence (AI)?

An analogy to help understand what Artificial Intelligence is


AI is an umbrella term that refers to the development of computer systems capable of performing tasks that would typically require human intelligence. As the diagram shows, these sub terms include Machine Learning (the ability to learn without explicit programming), Deep Learning (a subset of Machine Learning which requires no human intervention) and Generative AI (where computers are revolutionising content creation).

Read more on these terms in our articles on Artificial Intelligence vs Machine Learning and AI Algorithms: Traditional Machine Learning vs. Deep Learning

This article explains these terms using day to day scenarios to make these often complex sounding terms, easy to understand.

AI vs ML vs DL vs Gen AI
What is Artificial Intelligence (AI)?

Imagine you've misplaced your keys, and you need to find them. Think about the skills you'd use to search:

Identify your keys: You've seen lots of keys before, but you know which ones are yours because they're on your special keyring.

Make a search plan: You think about where you might have left them. Maybe they're on the kitchen counter, or you dropped them by the front door? You'd check each spot where you usually put them.

Adjust to changes: What if you had guests over and the keys got moved? If your usual spots turn up empty, you'd think about where they could be now. Did someone put them on a table, or did they slip behind the couch cushions?

React to tips: Say a friend tells you they saw your keys in the bathroom. You know what to do next — head straight to the bathroom, and you won't mix up your keys with any others you might find because you're looking for your familiar keyring.

Your ability to search for your keys is based on intelligence — recognising, planning, adapting, and acting on information.

Now, what if a computer or robot could do all that? We could program it to recognise your keyring, remember common places where you leave your keys, and even adapt when things change. If it's told the keys are somewhere new, like the bathroom, it'll go check there.

This is Artificial Intelligence: giving a machine the skills to solve problems like finding keys, learning from what it encounters, and making decisions just like humans do.

What is Machine Learning?

Following on from our key-finding robot, suppose we want the robot to not only find keys but also distinguish between various household objects which we might lose around the house.

For this, we would present the robot with images of keys, wallets, glasses, and remote controls, each labelled accurately. The robot would be trained to recognise these objects by their shape, size, colour, and even by the sound they make when dropped.

Once trained, the robot would be able to pick out these items from a visual scene. It’s similar to how you learn to differentiate between objects by seeing them repeatedly.

However, just as no two keys are exactly the same, the robot must understand the variety within a category. Showing it numerous pictures of keys, for example, teaches it the essence of 'key-ness'. The more images it processes, the better it becomes at finding keys, no matter where they're hiding. This is the core of Machine Learning — enabling a machine to recognise patterns and make decisions from examples, rather than through direct programming.

Machine learning is a type of Artificial Intelligence.

What is Deep Learning?

Deep Learning is a branch of Machine Learning inspired by the human brain's workings. It allows our robot, initially trained to recognise everyday items, to understand more complex scenarios, like objects partially obscured or viewed from odd angles.

Using neural networks with many layers, Deep Learning processes vast amounts of data to recognise intricate patterns. Each layer in these networks extracts different features from the images, progressing from simple attributes like edges and colours to more complex ones like textures and object recognition, regardless of orientation or lighting.

If our robot is equipped with Deep Learning, it will b able to identify objects with remarkable accuracy. For instance, even if a key is partially hidden under a pillow, it can detect the visible part and accurately deduce it's a key. Deep Learning empowers machines for tasks requiring human-like perception, significantly enhancing their capability to interpret and interact with the world.

Generative AI

This is where creativity comes in. The examples so far have shown human-like intelligence but no creativity.

If you were asked to create a car that could move quickly across any terrain, you would draw on your imagination and merge various concepts into an innovative mode of transport.

Our robot can also do something similar - by knowing about different vehicles like cars, hovercrafts, helicopters, planes etc., it can combine different parts together to create a new vehicle. This invention isn't just a rehash of existing vehicles; it's an entirely new creation born from the robot's ability to synthesise and innovate based on its learned knowledge.

This is Generative AI - where a computer displays apparent imagination to create new concepts such as artwork, music or architectural designs without being explicitly told what to create.

This example below is a vehicle created by ChatGPT 4 when asked to invent a vehicle which moves quickly across different terrain.

March 28, 2024

Read our latest

Blog posts