There is no universally accepted definition of artificial intelligence. A useful modern description is a machine-based system that infers from the information it receives how to produce outputs—such as predictions, generated content, recommendations, or decisions—that can influence a physical or virtual environment. That definition does not require a machine to think or feel like a person.
What artificial intelligence means
The OECD’s revised definition describes an AI system as a machine-based system that, for explicit or implicit objectives, infers from its inputs how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments. The OECD Council adopted this definition on 8 November 2023 (OECD Recommendation).
In plain language, an AI system takes in information, applies a method for working out what to produce, and returns an output that may affect what happens next. The output could be a forecast, a written response, a suggested option, or a decision. The definition focuses on the system’s function, not on whether it possesses human-like understanding or consciousness.
Input, inference, output, and influence
- Input: Information the system receives, such as text, images, or data from its operating environment.
- Inference: The process of working out how to produce an output in pursuit of an objective. The objective can be explicitly specified or implicit.
- Output: A prediction, piece of content, recommendation, or decision, among other possibilities.
- Influence: The output may affect a physical or virtual environment—for example, by informing a person’s choice or guiding a system’s next action.
This framework describes a broad range of systems, not a single technology or level of capability.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →#1 Best Overall
How an AI system can work
One useful conceptual model breaks a system into sensors, operational logic, and actuators. Sensors collect raw data; operational logic processes that information in relation to objectives; and actuators can change the environment. The OECD presents this as a way to understand AI systems, not as a checklist every AI product must meet (OECD, Artificial Intelligence in Society).
Some systems receive information digitally and return a recommendation or generated text, with no physical actuator. Others may be part of a system that can act on its surroundings. The model helps explain how input and output connect without implying that all AI operates in the physical world.
Rank #2
Why AI does not mean one human-like mind
AI is a broad field whose systems can be designed for different tasks, including perception, language, planning, learning, communication, and physical action. NIST’s glossary records multiple definitions from different source documents; some emphasize learning or operation under changing conditions, while others describe tasks associated with human capabilities. The range reflects differences in context and purpose, rather than a single mandatory checklist (NIST Glossary: Artificial Intelligence).
Systems also differ in autonomy and in how much they adapt after deployment. A system may produce an output for a person to review, or it may have more freedom to act within its operating context. Adaptiveness likewise varies; the word “AI” alone does not tell you whether a system continues learning after it is put into use. The OECD definition explicitly allows for these differences.
What the Turing test can—and cannot—show
In a 2019 primer, the OECD attributes to computer scientist John McCarthy the 1956 definition of AI as “the science and engineering of making intelligent machines” (OECD, Hello, World: Artificial Intelligence and its Use in the Public Sector). This is a concise historical description, but it does not resolve what intelligence means in every context.
The same primer describes the Turing test as a setup in which a human evaluator asks questions of a human and a machine, then judges whether the machine’s typed answers can be distinguished from the human’s. It is one way to examine conversational behavior. A machine that imitates human conversation would not thereby prove that it is conscious or generally intelligent.
More broadly, success on a narrow test is evidence about that test, not automatic proof of broad ability. An OECD capabilities report observes that a system might excel at a particular IQ-style test yet be unable to do anything beyond that test battery (OECD, Is AI Intelligent? Insights from Benchmarking of AI and the Future of AI). Evaluating broader capability requires evidence across the tasks and conditions relevant to the claim.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.A practical way to identify what an AI system does
When you encounter a claim about an AI system, ask what it actually does rather than relying on the label alone. A useful description names:
Best Value
- the task or problem it is intended to address;
- the inputs it receives and the outputs it produces;
- the environment in which it operates and how its outputs can affect that environment;
- how much autonomy it has, and whether it adapts after deployment; and
- the evidence used to evaluate its performance, including whether that evidence covers the claimed range of tasks.
Those details make it possible to distinguish a system built for one specific task from a broader claim about capability—without assuming that either one thinks like a human.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
