Artificial intelligence · Curated list
Machine learning, Neural network, Natural language processing and 27 more, ranked by how much of the world has written about each one.
Most written-about first. Each description summarises the idea or technique's Wikipedia article, which its name links to.
1
Subset of artificial intelligence
Often shortened to ML, a field of artificial intelligence that develops and studies statistical algorithms able to learn from data and generalise to data they have not seen, and so to do tasks they were never explicitly programmed for.
Wikidata · Photo: Original: Alisneaky Vector: Zirguezi, CC BY-SA 4.0
2
Computational model used in machine learning
Also called an artificial neural network (ANN), or NN for short, a computational model of machine learning inspired by how biological neural networks are built and work.
Wikidata · Photo: Dake , Mysid, CC BY 1.0
3
Processing of natural language by a computer
Usually shortened to NLP, the processing by computers of information in natural language. It is a branch of computer science, closely tied to artificial intelligence.
Wikidata · Photo: Siobhán Grayson, CC BY-SA 4.0
Type of machine learning model
Usually shortened to LLM, an AI model trained on huge amounts of text for tasks of natural language processing, above all generating language. LLMs can usually produce, translate, summarise and analyse text in many settings.
5
Branch of machine learning
Often shortened to DL, the part of machine learning that uses neural networks of many layers for tasks such as regression, classification and representation learning. Inspired by biological neuroscience, it centres on stacking artificial neurons in layers and "training" them to process data.
Wikidata · Photo: Sven Behnke, CC BY-SA 4.0
6
AI that generates content
Often shortened to GenAI, a branch of artificial intelligence (AI) whose generative models produce text, images, video, audio, computer code or other digital data. The models learn the patterns and structures behind their training data and use them to create new data in response to input, often a prompt written in natural language.
Computerized information extraction from images
The field whose tasks include methods for capturing, processing, analysing and understanding digital images, and for drawing high-dimensional data from the real world to produce numerical or symbolic information, such as decisions. "Understanding" here means turning what is seen into accounts of the world that thinking can make sense of and that can prompt suitable action.
Computer system emulating human expert
A computer system, in artificial intelligence (AI), that imitates how a human expert makes decisions. Such systems tackle complex problems by reasoning over stored knowledge, expressed mostly as if–then rules instead of conventional procedural code.
Type of AI with wide-ranging abilities
Usually shortened to AGI, a hypothetical kind of artificial intelligence that equals or exceeds human abilities in almost every cognitive task. Where artificial narrow intelligence (ANI) is competent only within well-defined tasks, an AGI system could generalise what it knows, carry skills from one domain to another and solve new problems without being reprogrammed for each task.
Machine learning paradigm
Often shortened to SL, a paradigm of machine learning in which an algorithm is taught, from example pairs of inputs and outputs, to turn input data into a particular output. A statistical model is trained on labelled data, meaning that each input comes with the correct output.
Field of machine learning
Often shortened to RL, the part of machine learning and optimal control concerned with how an intelligent agent should act in a changing environment to maximise a reward signal. With supervised and unsupervised learning, it is one of the three basic paradigms of machine learning.
Statistical modeling method
A statistical model estimating how a scalar response relates, through a linear combination, to one or more explanatory variables. With exactly one explanatory variable it is a simple linear regression; with two or more, a multiple linear regression.
Paradigm in machine learning that uses no classification labels
A framework of machine learning in which, unlike supervised learning, algorithms learn patterns only from data without labels. Other points on the spectrum of supervision include self-supervision, and weak or semi-supervision, where a small share of the data is labelled.
Automated recognition of patterns and regularities in data
The task of assigning an observation to a class according to patterns drawn from data. Though similar, pattern recognition (PR) differs from pattern machines (PM), which may be able to do PR but mainly exist to tell apart and generate emergent patterns.
Categorization of data using statistics
Classification of data by computer, where the algorithm is usually built with statistical methods.
Artificial intelligence (AI) and its subfields have found uses across academia and industry. Machine learning has served various scientific and commercial ends, among them e-commerce, credit scoring, decision-making, image recognition and translation between languages.
Cognitive science approach
An approach to studying cognition and human mental processes with mathematical models called artificial neural networks or connectionist networks.
Autonomous artificial intelligence agent
A program built on artificial intelligence that can pursue goals, take actions and use software or other tools, with a degree of autonomy. Agentic AI differs from using AI as a tool for narrow, specific jobs like answering questions, as chatbots commonly did in their non-agentic forms in 2023, or as traditional machine learning algorithms do.
Methods in artificial intelligence research
In artificial intelligence (AI), a set of methods built on high-level representations of problems, logic and search in symbols that humans can read. Its tools included production rules, semantic nets, frames and logic programming.
20
The use of artificial intelligence (AI) on medical and healthcare data, in fields such as diagnosing disease, planning treatment, monitoring patients, developing drugs and systems that support clinical decisions.
Wikidata · Photo: Setzner1337, CC0
Trial and error problem solvers with a metaheuristic or stochastic optimization character
Often shortened to EC, a family of algorithms from computer science for global optimisation, modelled on biological evolution, and the part of soft computing and computational intelligence that studies them. Technically, they solve problems by trial and error across a population of candidates, and have a stochastic optimisation or metaheuristic character.
Subfield of computer science and logic
The area of computer science, specifically of metalogic and of knowledge representation and reasoning, devoted to understanding the various aspects of reasoning. Its study helps create computer programs that let computers reason entirely, or almost entirely, on their own.
Machine learning paradigm
Often shortened to SSL, a paradigm of machine learning in which the data itself supplies the signals of supervision a model is trained on, instead of labels provided from outside. With neural networks, it aims to exploit structures or relationships already present in the input data to create useful training signals.
Type of algorithm, produces approximately correct solutions
In computer science and mathematical optimisation, a technique for solving problems faster when classic methods are too slow to find an exact or approximate answer, or cannot find any exact answer in a search space. It gets its speed by giving up some optimality, completeness, accuracy or precision.
Iterative simulation method
Often shortened to PSO, a method of computational science that optimises a problem by repeatedly improving a set of candidate solutions against a given measure of quality. The candidates, called particles, interact and move around the search space by simple mathematical formulae that adjust each one's position and velocity.
AI whose outputs can be understood by humans
Usually shortened to XAI, a field of artificial intelligence (AI) research into methods that let humans oversee AI algorithms intellectually; it largely overlaps with explainable machine learning (XML), interpretable machine learning and interpretable AI. It concentrates on the reasoning behind what AI algorithms decide or predict, to make them more transparent and understandable.
Machine learning technique
Often shortened to TL, a machine learning (ML) technique that reuses knowledge learned on one task to improve performance on a related one. In classifying images, for instance, what was learned while recognising cars could help in recognising trucks.
Emotion modeling in AI
The study and development of devices and systems able to recognise human affects, interpret and process them, and simulate them. The field is interdisciplinary, spanning cognitive science, psychology and computer science.
Interdisciplinary research area
Often shortened to QML, the field studying how quantum algorithms can do machine learning. The term often means quantum algorithms for machine learning tasks that analyse classical data, sometimes called quantum-enhanced machine learning.
Statistical techniques analyzing facts to make predictions about unknown events
A range of statistical techniques, drawn from machine learning, predictive modelling and data mining, that analyse present and past facts to predict unknown events, including future ones.
Every Wikidata item that is a kind of artificial intelligence or of machine learning, or a part of either, and that has its own English Wikipedia article, ranked by the number of Wikimedia sites with a page about it — Wikipedia's language editions, mostly, and sister projects such as Wikiquote. That counts how many communities independently thought it worth describing, and nobody can buy a place on it. The top 30 are shown.
Wikidata files these under this list's query, but they are not what the list is about:
These belong on the list but Wikidata does not file them where the query looks:
Fetched from Wikidata on 2026-09-25. Each description is this site's summary of the entry's Wikipedia article, which its name links to: the wording is ours, the facts are Wikipedia's, licensed CC BY-SA 4.0. To correct an entry, correct it there; the next refresh carries the change.
If you are reading this on the way somewhere: Computatus eSIM sells prepaid data eSIMs for 190 countries, installed by QR code before you leave, so maps, tickets and messages work without a roaming bill.
| Country | Cheapest plan | Price | Order |
|---|---|---|---|
| United States | 1GB per day | $2.33 | Get it → |
| France | 2GB per day | $2.15 | Get it → |
| Italy | 2GB per day | $1.81 | Get it → |
| United Kingdom | 2GB per day | $1.81 | Get it → |
| Germany | 2GB per day | $1.81 | Get it → |
| Canada | 1GB per day | $2.47 | Get it → |
| India | 1GB per day | $3.26 | Get it → |
| Australia | 1GB per day | $1.79 | Get it → |
See all Computatus eSIM plans →
A commission on plans bought through these links is what pays for this site. Nothing above depends on it: every entry is chosen and ranked before any price is looked up.
