ABOUT AI - Part 1: What Actually Is AI?

Welcome to Part One of our five-part About AI series, where we explain AI simply, without assuming you have any technical knowledge.

So, what actually is AI?

Artificial intelligence is a broad term for computer systems designed to perform tasks that usually require some form of human intelligence, such as recognising patterns, understanding language, solving problems, making predictions or generating content.

Unlike traditional software, where every rule has to be programmed in advance, many modern AI systems learn patterns from large amounts of data and use those patterns to respond to new information.

AI is not one machine, one intelligence or one system. It is an umbrella term covering a family of technologies that do very different things.

A Very Short History of AI

When we talk about AI today, we usually think of systems such as ChatGPT, Gemini or Claude.

But AI has been around far longer than that.

The term artificial intelligence dates back to the 1950s. Early systems focused on logic, problem-solving and rule-based decision-making. By the 1980s, expert systems were being used commercially, and during the following decades AI became increasingly useful for analysing data, recognising patterns, making forecasts and automating tasks.

For most people, however, AI remained something running quietly in the background.

That changed dramatically in 2022 when ChatGPT was released to the public. Suddenly, millions of people could interact with AI directly through ordinary conversation. AI moved from something largely hidden behind systems and industries to something people could actually talk to.

AI, Machine Learning and Deep Learning

These terms are often used interchangeably, but they are not quite the same thing.

Think of them as layers. AI is the broadest category. AI → Machine Learning → Deep Learning

Machine learning sits inside AI. Instead of being told every rule, a machine-learning system learns patterns from data.

Deep learning sits inside machine learning. It uses layered neural networks to work with large amounts of complex information and has driven many advances in areas such as language, image recognition, speech and generative AI.

You do not need to understand the maths behind any of this to understand the basic relationship:
 

Different Levels of AI

You may also hear AI described by its level of capability.

Artificial Narrow Intelligence — ANI  
This is the type of AI we currently use. ANI is designed to perform particular types of tasks or operate within defined areas. That can include recognising images, translating language, filtering spam, recommending content or generating text. Current systems can perform a wide range of tasks, but they are still considered narrow rather than truly general intelligence.

Artificial General Intelligence - AGI
AGI is the idea of an AI system that has broad, human-like cognitive ability. Able to learn, reason and adapt across many different kinds of tasks rather than being limited to particular domains. There is no universally agreed definition of AGI, and whether or when it will be achieved remains debated.

Artificial Superintelligence - ASI
ASI is a hypothetical future form of AI that would surpass human cognitive ability across virtually all intellectual domains. This does not currently exist.

Different Ways AI Can Be Used

Under the AI umbrella, systems can also be described by what they are designed to do.

These categories often overlap.

Large Language Models — LLMs
A Large Language Model is an AI model designed to work with language. LLMs are trained on very large amounts of text and learn patterns in language so they can generate, transform, summarise and respond to text.

Many conversational AI systems are built around large language models.

Generative AI
Generative AI creates new content based on patterns learned during training.
That content might be:

• text
• images
• video
• audio
• music
• code

An LLM can be generative AI, but generative AI is a much broader category than language alone.

Multimodal AI
Multimodal AI can work across more than one type of information. For example, one system may be able to understand text, analyse an image, respond to spoken language and generate audio or visual content.

Systems such as ChatGPT and Gemini increasingly combine several of these abilities within one interface.

Predictive AI
Predictive AI uses existing data and patterns to estimate what is likely to happen next. It is widely used in areas such as finance, retail, manufacturing, logistics and marketing. It may help forecast demand, detect risk, predict equipment failure or identify trends.

Conversational AI
Conversational AI is designed specifically for interaction through natural language. That might include customer-service chatbots, voice assistants or more advanced systems built around LLMs.

The main goal is to allow humans to communicate with a system using ordinary language rather than specialist commands.

Agentic AI
Agentic AI takes things a step further. Rather than simply answering a question, an agentic system can carry out a sequence of actions towards a goal. For example, it might:

• organise files
• search for information
• prepare or send messages
• work across apps
• book travel
• complete a multi-step task

The important difference is that it can act, not simply respond.

The AI You Don't Notice

A lot of AI is almost invisible. You may already interact with AI when:

• your bank flags an unusual transaction
• your map changes your route
• a shopping site decides what products to show you
• social media decides which posts appear in your feed
• your phone recognises a face or object in a photo

These systems influence what you see, what gets prioritised and sometimes which choices are placed in front of you. So there are really two sides to our everyday experience of AI.

There is the AI we consciously choose to interact with, such as ChatGPT, Claude or Gemini, then there is the AI quietly operating in the background of many of the digital systems we use every day.

Understanding that distinction changes the picture quite a lot.

AI isn't one machine, one intelligence or one system. “AI” describes a whole family of technologies doing very different things.

NEXT – Part 2: How is Ai built
From training data and neural networks to the data centres behind it.