What AI can and can't do

Knowing AI's real strengths and limits is what separates people who get great value from it from those who get burned. Here's the honest map.

What AI does genuinely well (with a human checking):

• Writing and rewriting — drafts, emails, summaries, rephrasing, changing tone. This is its home turf.
• Summarizing long text down to the essentials.
• Brainstorming — dozens of ideas, angles, names, outlines in seconds.
• Explaining — at any level you ask ("explain like I'm 10," "explain like an expert").
• Coding help — even for non-programmers doing small tasks.
• Generating images (and increasingly audio and video) from descriptions.
• Translation and other language tasks.

What AI genuinely can't do reliably (respect these limits):

• It makes things up — confidently. It can state false facts, invent quotes, and fabricate sources in a fluent, authoritative voice. This is called hallucination, and it's built into how the technology works — not a bug that's been fixed. (Module 3 covers spotting it.)
• It doesn't truly understand. It recognizes patterns; it doesn't know what's true or grasp meaning the way you do. It also can't reliably tell when it's unsure or wrong.
• It has a knowledge cutoff. Its training ended at some date, so it may not know recent events — unless it can search the web (many now can, but then it depends on what it finds).
• It's bad at math and precise counting. Because it predicts text rather than calculating, it fumbles arithmetic and counting. (There are fixes — later courses cover them.)
• It can be biased. It learned from human-written text, so it can reflect and repeat human biases and stereotypes.

The practical takeaway: treat AI as a fast, knowledgeable, tireless assistant — not an oracle. Lean on it hard for the language-shaped things it's great at. Keep a human (you) checking anything where being wrong matters: facts, numbers, names, dates, anything you'll publish, send, or act on. The people who thrive with AI aren't the ones who trust it blindly or the ones who dismiss it — they're the ones who know exactly which jobs to hand it and which to double-check. That judgment is most of the skill, and this course builds it.