Artificial Intelligence

Artificial intelligence (AI) is the field of computer science devoted to building systems able to perform tasks that normally require intelligence: perceiving, reasoning, learning, planning, deciding or using language. The discipline's reference textbook, by Stuart Russell and Peter Norvig, organizes it around the idea of an agent: an entity that perceives its environment and acts upon it to achieve goals. A widely used institutional definition is the OECD's (updated in 2024), which describes an AI system as a machine that, for explicit or implicit objectives, infers from the input it receives how to generate outputs—predictions, content, recommendations or decisions—that can influence physical or virtual environments.

Where the term comes from

The expression “artificial intelligence” was born in 1955, in the proposal for a research workshop to be held the following summer at Dartmouth College. It was signed by John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon, and it started from a bold conjecture: that “every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it.” That 1956 gathering is considered the founding event of AI as a field, and McCarthy is credited with coining the term.

Narrow, general and superintelligence

It helps to distinguish three levels. Narrow AI (or weak AI) is designed for a task or a bounded set of tasks; it is the only kind that exists today, and it includes conversational assistants, recommenders and fraud-detection systems. General AI (AGI) would be a system able to generalize and transfer its knowledge across domains at a human level: it remains theoretical. And superintelligence—machines that would surpass people in every domain—is a hypothetical stage. In 2026, despite the power of frontier models, there is broad consensus that all AI in production is narrow: AGI has not been reached.

Major branches

Historically, symbolic AI (sometimes called GOFAI) dominated, based on rules and logic written by humans: traceable reasoning, but rigid. The paradigm that prevailed was machine learning, in which the system learns patterns from data instead of explicit rules, and within it deep learning, based on neural networks. The most recent surge is that of generative AI and large language models, built on the Transformer architecture (2017); their breakthrough into mass use came with the launch of ChatGPT in late 2022.