The term "Artificial Intelligence" was coined at the Dartmouth Conference in 1956, and the field has gone through multiple cycles of hype and disappointment ("AI winters") since then. The current wave, driven by deep learning and massive compute, began around 2012 with AlexNet's breakthrough in image recognition and accelerated dramatically with the Transformer architecture in 2017 and ChatGPT's public launch in 2022.
Everything that exists today is narrow AI (also called "weak AI") — systems designed for specific tasks. Your spam filter is AI. Your voice assistant is AI. Claude is AI. But none of them can do everything a human can. Artificial General Intelligence (AGI) — a system with human-level capability across all domains — remains a research goal, not a product. The timeline debate ranges from "a few years" to "never," and the honest answer is that nobody knows.
Most modern AI is machine learning: instead of programming explicit rules, you provide data and let the system learn patterns. Deep learning (neural networks with many layers) is a subset of ML. LLMs are a subset of deep learning. This nesting matters because not all AI is ML (expert systems use hand-coded rules), and not all ML is deep learning (random forests, SVMs, and logistic regression are still widely used for tabular data where they often outperform neural networks).