AI, in plain words

"AI" used to mean a research program. In 2026 it mostly means ChatGPT, the autocomplete in your IDE, and the "smart" feature your bank app added last month. The word stretched a long way to get there, and that stretch is the first thing to understand.

This is post 1 of 8 in the Foundations series. We start at the word itself and work down: AI → machine learning → deep learning → the things you actually use every day.

Where the word came from

The term was coined in 1955 for a research proposal — the Dartmouth Summer Research Project on Artificial Intelligence, run by John McCarthy in 1956. The pitch was simple and outrageous: get a few mathematicians in a room for a summer and figure out how to make machines that can reason, use language, and solve problems on their own.

They didn't solve it. Nobody has yet. But the name stuck.

For the next 60 years, "AI" mostly meant academic research that didn't quite work. Programs that could play chess (eventually). Programs that could recognise handwritten digits (sometimes). Long winters where funding dried up. The term carried a slight whiff of overpromising.

What the word means now

Then ChatGPT launched on November 30, 2022. It reached 100 million users in two months, faster than any consumer product before it. After that, "AI" in everyday speech means roughly: a thing that can read or write text and respond like a person.

That's not the textbook definition. The textbook still says "machines that perform tasks requiring human intelligence". But that's not what your friend means when she says "I asked AI to plan my trip". She means an LLM. Probably ChatGPT, Claude, or Gemini.

The slippage matters. When a marketing page says "AI-powered", it could mean a 70-billion-parameter language model, or a 10-line if-statement that happens to use the word "smart". Both ship under the same label.

Why everything's AI now

Two reasons.

One: the post-ChatGPT goldrush. Every company with a SaaS product wants to say it has AI, because investors and customers reward that word right now. Calendar apps "use AI" to suggest meeting times. Toothbrushes "have AI" to detect brushing technique. Most of these are not LLMs, and many aren't even machine learning.

Two: "AI" is genuinely a fuzzy word. There's no agreed line between "rule-based software" and "AI". The spam filters of 2005 used statistical learning and were called software. The same code in 2026 would be called AI. The label moves with the times.

When you read "AI" on a product page, treat it as a marketing word. When you want to know what's actually inside, ask: is it a language model? A smaller machine-learning model? A hand-written rule?

Inside the umbrella

The cleanest way to think about it: AI is the umbrella term. Underneath sits machine learning (ML), and underneath that, deep learning (DL). Every modern "AI" thing you actually use lives at one of those nested levels.

That's where post 2 picks up: what each layer actually means, and why the three terms get used interchangeably even when they shouldn't.

What to take away

• "AI" started in 1956 as a research program. For 60 years it mostly meant "things that don't quite work yet".
• After ChatGPT (Nov 2022) the everyday meaning shifted to "things that can read and write like a person" — which today mostly means LLMs.
• "AI-powered" on a product page tells you nothing about what's actually inside.
• AI is the umbrella. ML and deep learning are what sit underneath. Post 2 opens them up.