Unsupervised learning means an AI finding patterns in data that has no answers attached. Nobody tells it what to look for; it sorts things into groups of similar items by itself, and a person then decides what each group means.
Tip a big jar of mixed buttons onto a table and ask someone to sort them, without saying how. They might end up with piles by colour, size or number of holes. Nobody gave them the categories; they found them. Unsupervised learning is a computer doing that sorting with data.
In a sentence: “The shop used unsupervised learning to sort its customers into groups with similar buying habits, then decided for itself what to call each group.”
What the word means
The plain point: “unsupervised” means there is no teacher marking the work. The data comes without labels, the right answers.
Google’s glossary for developers defines it as “Training a model to find patterns in a dataset, typically an unlabeled dataset.”1 IBM, the computer company, adds that this happens “without the need for human intervention.”2
The most common job is clustering: putting similar items into groups. Google points out the catch. The model does not know what its groups mean; working that out is up to you.3
A simple example
The plain point: it spots groups and links that people did not think to look for.
Google gives two everyday examples. A model can sort songs into groups by how the music sounds, and those groups can feed a music recommendation service. Another method, run on millions of shopping baskets, might show that baskets with lemons often also contain antacids.1 IBM adds spotting unusual data, which can flag faulty equipment or security breaches.4
Unsupervised learning finds patterns, not explanations. A group it finds can be useful, meaningless or misleading, and a person still has to judge which.
Supervised vs unsupervised vs reinforcement learning
The plain point: the three main ways to teach a computer differ in what they learn from.
| Kind | Learns from | 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 itself. | Like: sorting a jar of mixed buttons into piles. |
| Reinforcement | Learns from: rewards and penalties for its own actions. | Like: training a dog with treats. |
Stanford’s Institute for Human-Centered AI makes the same three-way split: in supervised learning a computer learns to predict “human-given labels”, unsupervised learning “does not require labels”, and reinforcement learning learns from rewards.5
Where you meet it
The plain point: it is the first stage of the chatbots you use.
That Stanford example, predicting each next word in a sentence, is how a large language model starts out. IBM describes the “pre-training” behind GPT models as “a form of unsupervised learning”: the model reads huge amounts of unlabelled text and learns its patterns.6 Some people call this self-supervised learning, because the model makes its own practice questions from the text.1 See GPT meaning.
IBM also lists its weak spots, including a “Higher risk of inaccurate results” and the need for people to check what comes out.4
Often confused with
- Classification: sorting into groups you chose in advance, such as “spam” and “not spam”. That is supervised learning. Clustering finds the groups for you.
- Learning with no human at all: people still pick the data, check the groups and decide what they mean.
- Reinforcement learning: learning from rewards for actions, not from patterns in a pile of data. See reinforcement learning meaning.
Questions people ask
What is an example of unsupervised learning?
Grouping songs that sound alike for a music recommendation service, or finding things that are often bought together, such as lemons and antacids: both are examples Google gives.1
Is ChatGPT supervised or unsupervised?
Both, at different stages. IBM describes the first training of GPT models on unlabelled text as “a form of unsupervised learning”.6 In OpenAI’s InstructGPT research, GPT-3 was then fine-tuned on labelled examples “using supervised learning”, and after that with human feedback.7 IBM says that work led to the models “that powered the launch of ChatGPT”.8
Learn more on this site
- The bigger picture: machine learning meaning.
- The opposite approach: supervised learning meaning.
- What it learns from: training data meaning.
- The words: the AI glossary.
QUICK CHECK · 3 QUESTIONS
Pick an answer to see if you got it. Nothing is recorded.
1. What makes learning ‘unsupervised’?
- A person checks every single step
- The data has no answers attached
- The AI learns from rewards and penalties
Show the answer
Answer: The data has no answers attached. The page says unsupervised means no teacher marking the work: the data comes without labels, and the model finds patterns itself.
2. After a model sorts data into groups, what still needs doing?
- Nothing, the model explains every group
- Each group must be marked spam or not spam
- A person decides what each group means
Show the answer
Answer: A person decides what each group means. The page says the model does not know what its groups mean, so a person has to judge whether each one is useful, meaningless or misleading.
3. How does a chatbot’s first stage of training work?
- It reads unlabelled text and learns its patterns
- It copies answers written by teachers
- It chats with users from the very first day
Show the answer
Answer: It reads unlabelled text and learns its patterns. The page says IBM describes the pre-training behind GPT models as a form of unsupervised learning on huge amounts of unlabelled text.
Related terms: Alignment · Attention · Checkpoint · Computer vision · Constitutional AI · Distillation
- 1Google for Developers, Machine Learning Glossary, last updated 10 Apr 2026, read at source 5 Oct 2026“Training a model to find patterns in a dataset, typically an unlabeled dataset. The most common use of unsupervised machine learning is to cluster data into groups of similar examples.” ↑a“an unsupervised machine learning algorithm can cluster songs based on various properties of the music. The resulting clusters can become an input to other machine learning algorithms (for example, to a music recommendation service).” … “shopping carts containing lemons frequently also contain antacids.” ↑bself-supervised learning is “A family of techniques for converting an unsupervised machine learning problem into a supervised machine learning problem by creating surrogate labels from unlabeled examples.” ↑c“an unsupervised machine learning algorithm can cluster songs based on various properties of the music. The resulting clusters can become an input to other machine learning algorithms (for example, to a music recommendation service).” … “shopping carts containing lemons frequently also contain antacids.” ↑d
- 2IBM, What is unsupervised learning?, published 23 Sep 2021, updated 2 Mar 2026, read at source 5 Oct 2026“Unsupervised learning, also known as unsupervised machine learning, uses machine learning (ML) algorithms to analyze and cluster unlabeled data sets. These algorithms discover hidden patterns or data groupings without the need for human intervention.” ↑
- 3Google for Developers, What is Machine Learning?, last updated 27 Jan 2026, read at source 5 Oct 2026“Clustering differs from classification because the categories aren’t defined by you.” … “An unsupervised approach doesn’t know what the clusters of data mean. Based on your understanding of the data, it’s up to you to define them.” ↑
- 4IBM, What is unsupervised learning?, updated 2 Mar 2026, read at source 5 Oct 2026“Unsupervised learning models can comb through large amounts of data and discover atypical data points within a dataset. These anomalies can raise awareness around faulty equipment, human error, or breaches in security.” ↑achallenges include “Higher risk of inaccurate results”, “Human intervention to validate output variables” and “Lack of transparency into the basis on which data was clustered”. ↑b
- 5Stanford HAI, Artificial Intelligence Definitions, Sep 2020, read at source 5 Oct 2026“In supervised learning, a computer learns to predict human-given labels” … “unsupervised learning does not require labels, sometimes making its own prediction tasks such as trying to predict each successive word in a sentence; reinforcement learning lets an agent learn action sequences that optimize its total rewards”. ↑
- 6IBM, What is GPT (generative pretrained transformer)?, read at source 5 Oct 2026“Generative pretraining is a form of unsupervised learning, where the model is fed unlabeled data and forced to make sense of it on its own.” ↑a“Generative pretraining is a form of unsupervised learning, where the model is fed unlabeled data and forced to make sense of it on its own.” ↑b
- 7Ouyang et al. (OpenAI), Training language models to follow instructions with human feedback, arXiv:2203.02155, submitted 4 Mar 2022, read at source 12 Oct 2026 (as cited on supervised learning meaning): “we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning”, then “reinforcement learning from human feedback”↑
- 8IBM, What is reinforcement learning from human feedback (RLHF)?, published 10 Nov 2023, read at source 12 Oct 2026 (as cited on RLHF meaning): “This was a crucial step in bridging the gap between GPT-3 and the GPT-3.5-turbo models that powered the launch of ChatGPT.”↑
The comparisons on this page (sorting buttons, flashcards, 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