The broad field of getting computers to do things that normally require human thinking. Recognising a face in a photo, translating a sentence, spotting fraud in a bank transaction, writing an email. It is an umbrella term covering many different techniques, not a single piece of technology, and it has been an academic field since the 1950s.
Technical view
The definition has always been a moving target, a problem sometimes called the ‘AI effect’. Once a task is solved reliably it tends to get reclassified as ordinary software. Chess, spam filtering and optical character recognition were all considered AI before they became routine. Historically the field split into two camps: symbolic AI, which encoded human knowledge as explicit rules and logic, and statistical or learned approaches, which infer behaviour from data. The rule-based camp dominated until roughly 2010. Almost everything people mean by AI today comes from the learned camp.
Related
Part of AI Terminology Index, section: The Foundations.