Summary: What generative AI actually is

A coworker turns six messy bullet points into a confident two-page memo in four seconds, and your stomach drops. That drop is the feeling of not knowing what you are looking at. This lesson replaces the knot with a picture: what the machine is, where it came from, and why understanding it is faster than you think.

Core ideas

• On November 30, 2022, OpenAI released ChatGPT to the public as a free website, and it crossed a reported million users within five days. The technology existed before; what changed is who could touch it.
• Treat that date as a historical marker, like the launch of the web browser. The specific tools have changed many times since and keep changing; the lesson teaches the part that stays true underneath.
• A generative language model does one thing: it predicts the next word, then does it again, one word at a time, until a whole answer has poured out. A reply appearing word by word on your screen is the machine actually working, not a typing effect.
• The predictions come from trillions of words of human writing. The model learns the deep statistical patterns of how people write, not memorized sentences; most sentences you say have never been said before by anyone.
• Prediction sounds too small until you notice what good prediction requires: facts, grammar, feeling, plot. As Sharad Goel puts it, “It’s a complicated prediction problem but it’s still really just prediction.”
• Hold both halves of the core insight: genuinely powerful, and not a mind. People who forget the first half ignore a tool their colleagues already use; people who forget the second half trust it too much.
• Raw prediction machines are tuned afterward toward being helpful, honest, and safe. Deciding whose values that tuning reflects is one of the field’s hardest open problems, and this track returns to it in its later lessons.
• Generative AI breaks from the older, categorizing AI you already know (spam filters, fraud flags, commute predictions) in three ways: it is general, it is generative, and it is approachable. The framing comes from the Harvard course, which borrowed it from a 2023 Goldman Sachs analysis.
• Approachable is the quiet revolution: you operate the tool in plain English. Clear thinking in ordinary words is the entry skill, which means you are already qualified.
• The prediction insight is a practical tool: it explains why answers sound confident even when wrong, and why the same question can get different answers twice. You will handle both cases in this track.
• The track answers two questions in order: how do I use this well (lessons 2 through 5), and what does it mean for my world (lessons 6 through 9). The Harvard admissions story shows why you need both; the admissions office is every office.

What changes for you

The next time a confident, slightly wrong AI answer crosses your screen, you will know what you are looking at: a prediction machine writing one word at a time, powerful and fallible for the same reason. That accurate picture protects you from hype and from dread at once. The next lesson starts the first arc with the most useful skill in the field, asking well: a good prompt has an anatomy, a small set of parts you can learn in an afternoon. Bring a real task from your own week.