Machine learning means teaching a computer by showing it lots of examples, so it finds the patterns itself instead of following rules a person wrote. It is the way almost all of today’s AI is made, from spam filters to chatbots.
You could teach a child what a dog is by writing a rulebook: four legs, fur, a tail. But cats have those too. Instead, you point at hundreds of dogs and say “dog”, and the child works out the difference alone. Machine learning teaches computers the second way.
In a sentence: “Our email app uses machine learning, so the spam filter gets better the more junk mail people mark.”
Machine learning vs AI vs deep learning
- AI is the big umbrella: any computer doing thinking-type tasks.
- Machine learning is the main way AI is built today: learning from examples instead of hand-written rules.
- Deep learning is one kind of machine learning, using large, many-layered neural networks. Chatbots are built this way.
What the word means
The plain point: the computer “learns” from data, rather than being told every step.
Google’s course for developers defines it in one line: “Machine learning (ML) is a way to train software, called a model, to make predictions or generate content using data.” The thing being trained is called a model: a file of learned patterns that a computer runs.1
IBM, the computer company, stresses the difference from normal programming: nobody types in the rules.2
The learning stage is called training. Using the finished model afterwards is called inference.
How does machine learning work?
The plain point: three steps, the same for a spam filter or a chatbot.
- Data. Collect lots of examples, often with the right answer attached, such as emails already marked “spam” or “not spam”. See training data meaning.
- Training. The computer adjusts the model’s internal numbers until its guesses on those examples are as good as it can get them.3
- Predictions on new data. The finished model is used on examples it has never seen, like a brand-new email, and makes its best guess.4
A simple example
The plain point: show it the answers, and it learns to guess new ones.
Take a spam filter. People mark thousands of emails as “spam” or “not spam”. The model studies them and learns what spam tends to look like. When a new email arrives, it makes a guess. Google uses spam as an example of this kind of learning.4
A machine-learning model makes a guess based on the examples it saw. Good examples give good guesses; gaps or mistakes in the examples carry through. That is why AI answers are worth checking.
Everyday examples
- Spam filters learn from emails people marked as junk (the example above).
- Maps apps predict your journey time: Google Maps combines past and live traffic using machine learning.5
- Recommendations (“because you watched…”): Netflix says its algorithms estimate how likely you are to enjoy each title from your viewing and from people with similar taste.6
- Voice assistants wake up when they hear their name: Apple’s “Hey Siri” detector is a deep neural network.7
- Chatbots such as ChatGPT, Claude and Gemini: large language models trained on huge amounts of text.
The main kinds
The plain point: there are a few ways to teach, depending on what examples you have.
| Kind | How it learns | Everyday comparison |
|---|---|---|
| Supervised | Learns from: examples that come with the right answer. | Like: flashcards with the answer on the back. |
| Unsupervised | Learns from: examples with no answers, finding groups and patterns itself. | Like: sorting a box of mixed buttons into piles that look alike. |
| Reinforcement | Learns from: trying things and getting rewards for good results. | Like: training a dog with treats. |
IBM describes the three the same way: learning to predict the “correct” output, finding patterns in data, and learning by reward.2 Google’s course also lists generative AI, models that make new content, as a type of machine-learning system.1
Where the word comes from
The phrase is usually traced to Arthur Samuel, who wrote about teaching a computer to play checkers (draughts). IBM says “The origin of the term (albeit not the core concept itself) is often attributed to Arthur L. Samuel’s 1959 article in IBM Journal, ‘Some Studies in Machine Learning Using the Game of Checkers.’”8 The idea in that quote is the one on this page: a program that learns to do better, rather than only following fixed rules.
Often confused with
- AI: the bigger umbrella. Machine learning is the main way AI is built today, but the word AI covers more. See artificial intelligence meaning.
- Deep learning: one kind of machine learning that uses large, many-layered neural networks. See deep learning meaning.
- An algorithm: the step-by-step method used to learn. The model is what that method produces. See algorithm meaning.
- Learning while you chat: the main learning happens before you use a chatbot. Whether your chats are used for later training is a setting. Common AI misconceptions explains.
Questions people ask
Is ChatGPT machine learning?
Yes. ChatGPT runs on a large language model, which is a deep learning model, and deep learning is a kind of machine learning: it learned from huge amounts of text rather than being programmed with answers.910
What is the difference between machine learning and traditional programming?
In traditional programming a person writes every rule. In machine learning the person supplies examples and the computer works out the rules itself, which IBM describes as making predictions without explicit, hard-coded instructions.2
Does machine learning need coding?
Professional machine learning usually does, but you can try it without. Google’s free Teachable Machine lets you train a model on your own pictures, sounds or poses in a web browser with no code.11
Learn more on this site
- Start here: AI explained for complete beginners.
- What it learns from: training data meaning.
- Make it your own: fine-tuning meaning, extra training on your own examples.
- The big build: how AI is made, step by step.
- The words: the AI glossary.
- Related: AI vs Machine Learning vs Deep Learning.
- Related: Generative AI vs Predictive AI.
QUICK CHECK · 3 QUESTIONS
Pick an answer to see if you got it. Nothing is recorded.
1. What is machine learning?
- A person writing every rule for the computer to follow
- Teaching a computer with lots of examples so it finds the patterns itself
- A computer that repairs its own parts
Show the answer
Answer: Teaching a computer with lots of examples so it finds the patterns itself. Instead of hand-written rules, the computer is shown examples and works out the difference itself, like a child learning what a dog is.
2. How do machine learning, AI and deep learning fit together?
- Deep learning is one kind of machine learning, and machine learning is the main way AI is built
- AI is one small part of deep learning
- They are three unrelated ideas
Show the answer
Answer: Deep learning is one kind of machine learning, and machine learning is the main way AI is built. AI is the big umbrella, machine learning is the main way it is built today, and deep learning is one kind of machine learning.
3. Why are AI answers worth checking, according to the page?
- Machine learning models never make guesses
- Models only work on emails
- Gaps or mistakes in the training examples carry through to the guesses
Show the answer
Answer: Gaps or mistakes in the training examples carry through to the guesses. A model makes a guess based on the examples it saw: good examples give good guesses, and gaps carry through.
Related terms: GPT · Hallucination · LLM (large language model) · Narrow AI (weak AI) · Natural language processing (NLP) · Neural network
- 1Google, What is Machine Learning?, updated 27 Jan 2026, read at source 1 Oct 2026“Machine learning (ML) is a way to train software, called a model, to make predictions or generate content using data.” ↑alists the types of ML systems as supervised learning, unsupervised learning, reinforcement learning and generative AI. ↑b
- 2IBM, Machine Learning, updated 2 Apr 2026, read at source 1 Oct 2026“This pattern recognition ability enables machine learning models to make decisions or predictions without explicit, hard-coded instructions.” ↑a“Supervised learning trains a model to predict the ‘correct’ output for a given input.” … “Unsupervised learning trains a model to discern intrinsic patterns, dependencies and correlations in data.” … “Reinforcement learning (RL) trains a model to evaluate its environment and take an action that will garner the greatest reward.” ↑b“This pattern recognition ability enables machine learning models to make decisions or predictions without explicit, hard-coded instructions.” ↑c
- 3Google for Developers, Machine Learning Glossary: ML Fundamentals, undated glossary, as cited on this site’s training data page, read at source 1 Oct 2026training is “The process of determining the ideal parameters composing a model.” ↑
- 4Google, Supervised Learning, updated 25 Aug 2025, read at source 1 Oct 2026“Supervised learning uses labeled data to train models that predict outcomes for new, unseen data.” ↑a“Supervised learning uses labeled data to train models that predict outcomes for new, unseen data.” … “like identifying spam or predicting precipitation.” ↑b
- 5Google, Google Maps 101: How AI helps predict traffic and determine routes, 3 Sep 2020, read at source 5 Oct 2026“We then combine this database of historical traffic patterns with live traffic conditions, using machine learning to generate predictions based on both sets of data.” ↑
- 6Netflix Help Center, How Netflix’s Recommendations System Works, undated help article, read at source 5 Oct 2026“We estimate the likelihood that you will enjoy a particular title in our catalog based on a number of factors including” “your interactions with our service (such as your viewing history and how you rated other titles),” “other members with similar tastes and preferences on our service”; “All of these signals are used as some of the inputs that we process in our algorithms.” ↑
- 7Apple Machine Learning Research, Hey Siri: An On-device DNN-powered Voice Trigger for Apple’s Personal Assistant, October 2017, read at source 5 Oct 2026“The ‘Hey Siri’ detector uses a Deep Neural Network (DNN) to convert the acoustic pattern of your voice at each instant into a probability distribution over speech sounds.” ↑
- 8IBM, Machine Learning, updated 2 Oct 2026, read at source 12 Oct 2026“The origin of the term (albeit not the core concept itself) is often attributed to Arthur L. Samuel’s 1959 article in IBM Journal, ‘Some Studies in Machine Learning Using the Game of Checkers.’” It quotes the paper: “a computer can be programmed so that it will learn to play a better game of checkers”. ↑
- 9OpenAI (Chatterji et al.), How People Use ChatGPT, 15 Sep 2025, read at source 5 Oct 2026“ChatGPT is based on a Large Language Model (LLM), a type of Artificial Intelligence (AI)”. ↑
- 10IBM, What Are Large Language Models (LLMs)?, 6 Oct 2021, read at source 5 Oct 2026“Large language models (LLMs) are a category of deep learning models trained on immense amounts of data, making them capable of understanding and generating natural language.” ↑
- 11Google, Teachable Machine, Experiments with Google, November 2019, read at source 5 Oct 2026“A fast, easy way to create machine learning models – no coding required.” … “Train a computer to recognize your images, sounds, & poses, then export your model for your sites, apps, and more.” ↑
The comparisons on this page (teaching a child about dogs, flashcards, sorting buttons, training a dog) are this site’s own ways of explaining, not claims from these sources.
Free printable AI cheat sheets, and AI news every week.
Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.
SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK