Xenith Guru
Practice
One thing to try and a few questions for each lesson. Open a question to see its answer.
1. What does a chat assistant actually do?
Try this. Ask an assistant something you already know well: your town, your work, a hobby. Read the answer line by line and mark anything wrong, vague or invented.
Does a chat assistant look facts up when it answers?
Not unless it is connected to search or to documents you give it. Otherwise it writes from patterns learned in training.
Why can a wrong answer sound confident?
Because the model is choosing likely-sounding words. The fluency comes from the pattern, not from having checked the fact.
2. How do I ask clearly?
Try this. Take something you would normally ask in one line. Add the why, the who and the length, then compare the two answers.
What are the four things that make an answer better?
What you want, why, for whom, and an example or the material to work from.
The answer is nearly right. Should you start again?
Usually not. Say what to change; the conversation so far is part of what the assistant reads.
3. How do I check an answer?
Try this. Ask a question on a topic you care about, with sources. Open every source. For each claim, note whether it is found as stated, found but different, or not found.
Which details most often need checking?
Exact figures, dates, names, quotations and references.
The assistant names a source. Is that enough?
No. Open it and find the claim. A source can be real and still not say what was claimed, or it may not exist.
Is asking "are you sure?" a check?
No. It may change the answer, but only a source you have read, or a reference you trust, is a check.
4. What should I hand over?
Try this. List five tasks from your week. Mark each one: hand over, hand over and check, or keep.
Should an assistant decide whether someone gets a job?
No. Decisions about people should stay with people who can explain them and answer for them.
Can you rely on its summary of a document you gave it?
It is a good start. Check it against the document for anything you will act on.
5. How do I stay responsible?
Try this. Find your school's, employer's or publisher's guidance on using AI. If there is none, write three sentences you would be comfortable following.
Is it fine to paste a friend's private message into a chatbot to get help replying?
Think first: it is their words and their privacy. Leave out names and details, or ask them.
Who is responsible for a mistake in AI-assisted work you publish?
You are.
6. What happens to what I type?
Try this. Open the settings of an assistant you use. Find out whether your conversations are kept and whether they are used for training, and change what you can.
Is it safe to paste a password if you delete the conversation afterwards?
No. Deleting the conversation on your screen may not delete the provider's copy. Never paste passwords.
What is a local model?
One that runs on your own device, so what you type does not leave it.