The most important thing to know: AI can be confidently wrong

This is the single most important lesson in the course, so read it carefully: AI assistants can be confidently, convincingly wrong. They state false information in the same fluent, authoritative tone as true information — and if you don't know this, you can be seriously misled. Understanding this, and the habit it requires, is what separates safe AI users from unsafe ones.

What "hallucination" means. AI assistants sometimes generate information that is simply false — invented facts, wrong dates, made-up statistics, fake quotes, even fabricated sources and citations that look real but don't exist. This is called hallucination, and here's the crucial part: it's not an occasional glitch you can spot — it's a fundamental property of how these systems work. They generate plausible-sounding text, and plausible-sounding is not the same as true. The AI isn't lying on purpose; it genuinely doesn't "know" what's true — it predicts what words should come next, and sometimes those words are wrong.

Why it's so dangerous — the confidence. The reason hallucination is a trap is that AI states wrong information just as confidently as correct information. There's no hesitation, no "I'm not sure" — a fabricated fact comes out sounding as authoritative as a real one. In fact, research has found AI can sound more confident when it's wrong. So "it sounds sure" is not a reliability signal. You cannot tell truth from hallucination by how confident the answer sounds — which is exactly why you must develop the verification habit.

How often does it happen? It varies a lot by the task. For summarizing text you gave it, errors can be rare. For recalling specific facts, dates, statistics, or details about obscure topics, error rates can be high — studies auditing chatbots on factual questions have found meaningful and sometimes rising rates of false claims. The point isn't a specific percentage (which varies and changes) — it's that it happens often enough that you cannot assume any specific fact is correct.

The habit this requires — verify anything important. Here's the practical rule that keeps you safe:

• For anything that matters — health, money, legal, safety, medication, travel bookings, anything you'll act on or repeat as fact — verify it with a trustworthy source before relying on it. Don't take the AI's word for an important fact.
• Ask for sources, but check them — you can ask "what's your source for that?", but be aware AI can fabricate sources too, so actually look them up.
• Lower-stakes, creative, or reversible things (brainstorming, drafting a message, explaining a concept you'll think about) — you can generally use directly without verifying every word.

The dividing line is stakes: the more a wrong answer could hurt you (health, money, safety, reputation), the more you must verify.

What this doesn't mean. This isn't a reason to distrust AI entirely or not use it — AI is genuinely useful, and most everyday uses (writing help, brainstorming, explanations you'll think through) are low-risk. It's a reason to use it wisely: enjoy its help, but verify anything important rather than assuming it's right. That one habit prevents almost all the harm hallucination can cause.

The mindset: the most important thing to know about AI is that it can be confidently, convincingly wrong — hallucination is a fundamental property, not a rare glitch, and AI states false facts (even fake sources) in the same authoritative tone as true ones, so "it sounds sure" means nothing. The habit that keeps you safe: verify anything important — health, money, legal, safety, anything you'll act on — with a trustworthy source before relying on it, while using AI freely for low-stakes, creative, or reversible tasks. This single habit — matching your verification to the stakes — is what lets you get AI's real benefits without being misled by its confident mistakes. It's the most valuable thing in this whole course.