artificial intelligence
A collective term for computer systems that solve tasks requiring human thought: they understand text, recognize images, prepare decisions, and learn from examples.
In plain terms
Artificial intelligence is not a single machine or program but a collective term, much like the word 'vehicle' encompasses bicycles, trucks, and airplanes. This includes your phone's face recognition, spam filters, translation software, and chatbots.
With traditional software, the programmer writes step by step what it should do. Most modern AI systems instead derive rules from countless examples. It is a bit like when a child learns to speak not from a grammar book but from what they hear around them.
It is good for entrusting tasks to machines that cannot be described with precise rules: for example, reading a handwritten address, summarizing a long contract, or pre-screening an X-ray image. It still does not "think" like you. It recognizes patterns and calculates with probabilities, which is why it can be surprisingly clever and surprisingly unreliable at the same time.
How it works
- Data collection: developers gather large amounts of examples, such as texts, images, audio recordings, or transaction data.
- Training: a mathematical model practices on examples. A computer procedure automatically adjusts the model's internal settings during the process until it makes fewer and fewer mistakes. These settings can number in the billions in large models.
- Generalization: a trained model provides answers to cases it has not seen before by applying patterns derived from examples to new situations.
- Usage: you enter a question or upload an image, and the model generates the answer based on probabilities. This can be a label, a prediction, or a written text.
- Action: In the latest systems, models are given tools (search, calendar, code execution) and, as agents, can independently complete multi-step tasks.
An everyday example
When your phone unlocks the screen based on your face, the map suggests a new route due to expected traffic, and your bank sends an SMS about a suspicious card payment, artificial intelligence is working behind all three, at least partially. The same applies if you have a chatbot draft an official letter for you.
Who is affected
A common misconception
Limits and risks
- It can make mistakes and even invent facts, so always verify important facts, numbers, and references from other sources.
- It adopts biases and prejudices present in the training data, so it can make unfair decisions, for example in credit assessments or resume screening.
- Often even developers cannot say exactly why a model gave a certain answer, which complicates error detection and clarifying responsibility.
- What you enter into an internet-accessible AI service usually goes to the provider's servers, so share personal, medical, or business secrets only with caution.
Similar, but not the same
- machine learning – Machine learning is one of the currently dominant methods of artificial intelligence, meaning it is part of AI, not a synonym.
- generative AI – Generative AI is only the subfield that produces new text, images, audio, or video, whereas AI as a whole also includes recognition, forecasting, and decision support.
- general AI – Artificial general intelligence refers to a system that reaches human level in any intellectual task. Most experts say no such system exists yet, and the precise definition of the concept is also disputed. Today's AI is strong only in certain tasks.
Related concepts
computer science, automation
machine learning, deep learning, generative AI, large language model, AI-agent
algorithm, neural network, general AI, robotika, data science
A brief history
The question of the thinking machine was raised by Alan Turing in his 1950 paper. The term "artificial intelligence" itself comes from John McCarthy, who used it in the 1955 proposal for the Dartmouth summer research workshop. The workshop held in 1956 is considered the birth of the field. Certain forms of AI had already been part of daily life earlier, such as in search engines and voice assistants. However, the wider public truly took notice with the launch of ChatGPT in November 2022, particularly regarding generative AI.
Where things stand
By the end of 2026, AI has become a standard part of phones, search engines, office programs, and customer service, and agents that carry out tasks autonomously are also spreading. The European Union's AI Act entered into force in August 2024. Its rules become applicable in stages from 2025, and some provisions will only become mandatory in the coming years.
In the news
- Three men charged over $129 million fraud using AI impersonation of CEOs and lawyersNov 6, 2026 · Jog
- Yettel uses software robots to replace repetitive administrative workNov 5, 2026 · Work & HR
- Gránit Bank: customers increasingly open to banking AI for managing spendingNov 5, 2026 · Money
- AI models may predict sovereign defaults years ahead, and credit insurance could benefitNov 5, 2026 · Money
- Eötvös Loránd University rector's directive regulates the use of AI in educationNov 4, 2026 · Education & Family
- National Conference on Educational Research in Eger examines pedagogical impact of AI and digitalisationNov 3, 2026 · Education & Family
- 90% of companies face a wave of AI-written employee complaints, HR Portál reportsNov 2, 2026 · Work & HR
- Hungarian survey: 58% of shoppers use AI for shopping, and trustworthiness is decisiveNov 2, 2026 · Marketing
Also known as
MI · AI · artificial intelligence · artificial intelligence · A.I. · machine intelligence · MI-rendszer · AI-rendszer
Sources
This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.