A human guide to artificial intelligence

What you’ll be able to do

• Distinguish a prediction system from a generative system.
• Choose a useful, low-risk first task.
• Explain why a fluent answer still needs human review.

Start with a problem you recognize

Imagine you are organizing a free community workshop. You have scattered notes, a rough agenda, and an audience that includes people who have never tried AI. You could use an assistant to organize the notes, propose an agenda, or explain an unfamiliar term. You remain responsible for the event and the accuracy of its description.

AI is a broad field. Some systems classify inputs or predict outcomes: for example, identifying whether a message resembles spam. Generative systems produce new text, images, audio, or other outputs. A language model uses patterns learned during training and the context available in the current interaction to generate text. Producing a plausible sentence is different from establishing that the sentence is true.

Three kinds of help

Transformation: Turn material you supply into another form. Examples include an outline from notes or a plain-language version of your own explanation.

Exploration: Generate alternatives you can evaluate. Examples include workshop titles or different ways to explain an idea.

Investigation: Gather or interpret information. This requires particular care about sources, dates, and what the assistant actually accessed.

These categories overlap. Their value is helping you choose what to check. For a transformation, compare the result with the original. For exploration, judge usefulness and fit. For investigation, verify the evidence.

A worked example

Your notes say: “Saturday. Library meeting room. AI introduction. Adults. Bring a laptop if possible. Time not confirmed.”

A useful draft might say: “Join an introductory AI workshop for adults at the library on Saturday. Bring a laptop if you have one. The start time will be announced.”

A poor draft might say: “Join our certified AI experts at 10 a.m. for a two-hour workshop.” The source never supplied a certification, time, or duration. Polished wording has introduced unsupported facts.

Try this prompt

Help me turn these fictional planning notes into a short event description.
Use only facts in the notes. Do not invent a time, duration, speaker qualification, or registration link.
After the description, list the details I still need to confirm.

Notes: Saturday. Library meeting room. AI introduction. Adults.
Bring a laptop if possible. Time not confirmed.

Your practice: choose one useful task

1. List three small tasks you do regularly.
2. Choose one with non-sensitive inputs and an output you can easily check.
3. State whether you want transformation, exploration, investigation, or a combination.
4. Try one prompt and save the response.
5. Mark every statement that was not present in your input. Decide which additions are helpful suggestions and which are unsupported facts.

Your finished artifact: The original input, the AI draft, and a corrected version with a short note explaining your changes.

How to judge your work

The corrected version should preserve known facts, flag missing information, and suit its audience. If you cannot tell whether the output is accurate, choose a simpler first task. The goal is a dependable habit, not a dramatic demonstration.

A common mistake

Asking an assistant to “handle the whole thing” before deciding what success means. Start with one bounded contribution. You can expand its role after you know how to review the result.

Reflection

Which part of the task benefited from AI, and which part still depended on your experience? Write one sentence about each.

Go deeper

Google’s introduction to language models provides additional technical background. You do not need the mathematics to complete this course.