AI for research
Episode 5 · 3 minutes 32 seconds · Free to learn · Words printed on screen · Watch above, then try the exercise below.
What you'll learn
- How research with AI works: asking in layers instead of one giant question
- How to turn a wall of information into a short, plain-words summary you can act on
- Why AI sometimes invents confident, tidy answers — and what a hallucination is
- The three checks — source, compare, ask again — that catch invented answers
Key points
Research just means learning before you decide. AI has read enormously, and it answers by predicting the most likely response from patterns in everything it read — usually right, but sometimes it builds a confident, tidy answer it simply invented, and it sounds exactly like the truth. In the video, an answer claims tomatoes need twelve hours of direct sun every day; the seed packet says full sun — six to eight hours — and a trusted garden book agrees. So run the three checks every time it matters: ask for the source (real facts have homes; invented ones don't), compare with something you trust, and ask again a different way. And some things get checked no matter what: dates and numbers, names and quotes, and anything you are about to repeat, buy, or act on. Camp rule: trust AI to start the search — check before you act.
Try it
- Pick something you're curious about — a hobby, a trip, a purchase.
- Ask in layers: start broad, then follow the thread. Then ask for the plain-words summary — the few things a beginner must know.
- Take one important claim from the answer and run the three checks on it: ask for the source, compare it with something you trust, and ask again a different way. See how it holds up.
The words in this video
Prefer to read, or want to follow along while you watch? Here is everything the narrator says in this lesson, word for word.
Welcome back to AI Camp! Today: AI for research — asking bigger questions, and getting answers you can actually use. And the second half matters most: how to catch AI when it makes things up. Because sometimes, it does. Research sounds grand, but here it just means learning before you decide. Planning a garden. Comparing two washing machines. Choosing a gift. Any time you'd ask a knowledgeable friend — you can start with AI. AI has read enormously — books, articles, manuals, more than any of us could read in ten lifetimes. Ask a question, and it pulls the threads together in seconds — in plain words, and at whatever depth you want.
Let's research something real: your first tomato plants. When do they go in the ground? How much sun do they need? How much water? Three honest questions — a perfect little research project. Don't ask one giant question. Ask in layers. Start broad: what matters most for a beginner's tomatoes? Then follow the thread — sun, then water, then timing. Each answer tells you what to ask next. And here's where AI really earns its keep: ask it to sum up. The key things a beginner must know, in plain words. A wall of information becomes a short checklist you can act on.
But. AI doesn't look things up the way you do. It predicts the most likely answer from patterns in everything it read. Usually that's right. Sometimes, it builds a confident, tidy answer that it simply invented. There's a name for that: when AI invents, people call it a hallucination. And here's the catch — it doesn't sound unsure. It sounds exactly like the truth. That's why the checking habit in this lesson matters. Watch for it in our garden. The answer says: tomatoes need at least twelve hours of direct sun, every day. Confident. Specific. And wrong. You can't tell from how it sounds — so we check.
First check: ask where it came from. What source says that? If AI can't point to one — or the source sounds invented — your guard goes up. Real facts have homes. Invented ones don't. Second check: compare it with something you trust. The seed packet says full sun — six to eight hours. The garden book on your shelf agrees. Two trusted sources beat one confident guess. Third check: ask again, a different way. Asked fresh, the same question often gets the honest answer. And when two asks disagree, the truth needs a third source. Go find it.
So: three checks, every time it matters. Ask for the source. Compare with something you trust. Ask again, a different way. Thirty extra seconds — and invented answers rarely survive all three. Some things get checked no matter what. Dates and numbers. Names and quotes — the exact details AI is most likely to invent. And anything you're about to repeat, buy, or act on. If it matters, it gets checked. Today's camp rule: trust AI to start the search — check before you act. Fast for learning, careful for deciding. That balance is the whole skill.
Your turn. Pick something you're curious about — a hobby, a trip, a purchase. Ask in layers, then ask for the plain-words summary. Then take one important claim and run the three checks on it. See how it holds up. Next time at camp: AI with documents and numbers. We'll hand it a long report and a messy spreadsheet — and get back the short version, and the story hiding in the numbers. Ask big, check small — and AI becomes the research partner it should be. See you next time, right here at AI Camp.