In plain words
Most computer programs follow exact instructions that a person wrote: if this happens, do that. They are fast, but they can only do what someone thought of in advance.
Artificial intelligence, or AI, is the wider goal of getting computers to handle tasks we link with human thinking: recognising a face, understanding a sentence, translating, planning a route, spotting fraud.
There are two big ways to get there. One is to write the rules of thinking by hand, called symbolic AI. The other, which powers almost everything today, is to let the computer learn patterns from lots of examples. That is machine learning, and its most powerful form is deep learning.
AI is a field, not a single thing. A chatbot, a spam filter and a chess program are all AI, built in very different ways.
Everyday example
Your phone's photo search
Type "beach" into your phone's photo search and it finds your Goa trip pictures, even though you never tagged them. Nobody wrote a rule for what a beach looks like. The phone uses a model that learned from millions of labelled photos. Recognising a beach is easy for you and was once very hard for a computer. Doing it is AI.
Where the example breaks: You recognise a beach because you have been to one and know what it is for. The phone only matches patterns of colour and shape it learned, which is why it sometimes files a blue wall under "sea".
How it actually works
AI is best understood as a family of methods nested inside each other, plus a few ideas about where it is all heading.
The goal: tasks that seem to need intelligence
The name was coined for a 1956 summer workshop at Dartmouth College. Its 1955 proposal guessed that "every aspect of learning or any other feature of intelligence" could be described precisely enough for a machine to simulate it. That is still the working ambition.
Approach one: rules written by people
Early AI stored knowledge as symbols and rules and searched through them, like a very fast logician. This gave us chess programs and expert systems, but it struggled with messy tasks like seeing or understanding speech.
Approach two: learning from data
Machine learning lets a program find its own rules from examples. The result of that learning is an AI model. Neural networks with many layers, called deep learning, took over from about 2012 because they kept improving with more data and computing power.
What we use today
Most AI in daily life is narrow AI: one system, one kind of task. Generative AI, such as chatbots and image makers, creates new content. Artificial general intelligence, a system as broadly capable as a person, does not exist yet and its definition is debated.
- Name first used
- 1955, in the proposal for the 1956 Dartmouth summer workshop
- Who proposed it
- John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon
- Main branches
- Machine learning, computer vision, language processing, robotics, planning
- Dominant method today
- Deep learning: neural networks with many layers
In everyday talk "AI" now often just means chatbots. On this site we use it in its full, older sense.
Questions people ask
Is AI the same as machine learning?
No. AI is the whole field of making computers do tasks that seem to need intelligence. Machine learning is one way to do it: learning from examples instead of following hand-written rules. Almost all modern AI uses machine learning, which is why the two words get mixed up.
Who invented artificial intelligence?
No single person. The name comes from a 1955 proposal by John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon for a 1956 workshop at Dartmouth College. Alan Turing had asked "Can machines think?" in 1950, and many researchers built the field after them.
Is ChatGPT AI?
Yes. ChatGPT is an app built on a large language model, which is a kind of generative AI, which is a kind of machine learning, which sits inside AI. It is one example of AI, not the whole of it.
Does AI think like a human?
Not in the way we do. Today's AI finds statistical patterns in data and uses them to make predictions. It can be very capable at specific tasks while lacking the understanding, experience and common sense people take for granted. Whether machines could ever truly think is still debated.
What are the main types of AI?
One common split is by breadth: narrow AI does one kind of task, general AI would match people across most tasks, and superintelligence would far exceed them. Only narrow AI exists today. Another split is by method: rule-based versus learning-based.