Free explainer

Types of AI: narrow, general, generative and real examples

AI can be sorted by how capable it is, by the technique it uses and by what it does. Here is each grouping in plain English, with everyday examples and a short timeline of how we got here.

Start with What Is AI? Browse the AI glossary

Written by the AI Understanding editorial team, a 501(c)(3) nonprofit. Free, no sign-up.

Types of artificial intelligence

The short answer

AI is usually grouped two ways. By capability: narrow AI built for specific tasks, which is all AI today, plus hypothetical general AI at human level and superintelligence beyond it. By technique: machine learning, deep learning and generative AI. Most tools you use daily are narrow, machine-learning-based systems.

Types of AI by capability

This is the grouping most articles mean by 'types of AI'. It asks how broad a system's abilities are.

Step 1

Narrow AI (weak AI)

Built for one task or a small set of tasks, such as spam filtering, route planning or image recognition. Every AI system in use today is narrow, including chatbots, though some are very broad within language.

Step 2

Artificial general intelligence (AGI)

A hypothetical system that matches human ability across most intellectual tasks. Whether and when it could exist is debated among researchers.

Step 3

Superintelligence

A hypothetical system far beyond human ability in nearly every domain. It is a topic of safety research and speculation, not a current technology.

Types of AI by technique

These terms describe how a system is built, and they nest inside each other: deep learning is a kind of machine learning, which is a kind of AI.

Step 1

Rule-based (symbolic) AI

Hand-written rules and logic. Predictable and explainable but hard to scale to messy real-world data.

Step 2

Machine learning

Systems that learn patterns from data instead of following fixed rules. Includes supervised, unsupervised and reinforcement learning.

Step 3

Deep learning

Machine learning with many-layer neural networks. It powers modern image, speech and language AI.

Step 4

Generative AI

Models that create new text, images, audio, video or code. ChatGPT and image generators are examples.

Step 5

Reinforcement learning

An agent learns by trial, reward and penalty. Used in game-playing systems and robotics research.

Another common grouping: by how the system behaves

You may also see four 'functional' types: reactive machines that only respond to the current input, limited-memory systems that use past data (most modern AI), and two hypothetical stages usually called theory of mind and self-aware AI. This is an informal teaching framework rather than a technical standard, but it is useful for seeing that today's systems do not have understanding or consciousness.

Examples of AI in everyday life

Most people already use narrow AI many times a day, often without noticing.

  • Search, maps and navigation that estimate routes and arrival times.
  • Recommendations for videos, music and products.
  • Spam and fraud filters in email and banking.
  • Voice assistants and speech-to-text dictation.
  • Photo apps that group faces or enhance images.
  • Chatbots and writing assistants such as ChatGPT.
  • Translation tools that convert text between languages.

A short history of AI

AI is older than most people expect. These are the milestones that most histories agree on.

Step 1

1950: the Turing test

Alan Turing publishes 'Computing Machinery and Intelligence' and proposes judging machine intelligence by conversation.

Step 2

1956: the term 'AI'

The Dartmouth workshop brings researchers together and popularizes the name artificial intelligence.

Step 3

1997: Deep Blue

IBM's Deep Blue defeats world chess champion Garry Kasparov in a match.

Step 4

2012: deep learning breakthrough

The AlexNet neural network wins the ImageNet image-recognition contest by a wide margin, starting the deep learning boom.

Step 5

2016: AlphaGo

DeepMind's AlphaGo beats Go champion Lee Sedol, a game long thought out of reach.

Step 6

2017: the transformer

Researchers publish 'Attention Is All You Need', the architecture behind today's language models.

Step 7

2022: ChatGPT

OpenAI releases ChatGPT, bringing generative AI to a mass audience.

Where to go next

Pick the thread that interests you. If you are new, follow the beginner roadmap; if you want to test yourself, try a quiz.

ChatGPT guide

How the best-known generative AI works and how to use it.

Frequently asked questions

What are the 3 types of AI?

By capability they are narrow AI (task-specific, the only kind that exists today), artificial general intelligence (hypothetical, human-level across tasks) and superintelligence (hypothetical, beyond human ability).

What are the 4 types of AI?

A common teaching framework lists reactive machines, limited-memory AI, theory of mind and self-aware AI. The first two exist today; the last two are hypothetical. It is an informal classification rather than a technical standard.

Is ChatGPT narrow AI or general AI?

ChatGPT is narrow AI. It is very flexible within language tasks, but it is not general intelligence and it has no human-level understanding across all domains.

What is the difference between machine learning, deep learning and generative AI?

Machine learning is AI that learns patterns from data. Deep learning is machine learning with many-layer neural networks. Generative AI uses such models to create new content such as text and images.

What are some examples of AI?

Examples include recommendation systems, spam filters, voice assistants, navigation apps, translation tools, image generators and chatbots like ChatGPT.