ntworld.ink
Roper Gulf Regional Council · Session 5 of 5

Wrap Up, Questions, and AI Image Generation

A relaxed close. We pull the course together and leave plenty of room for questions. Then one memorable activity: the whole group generates the same picture, and no two come out the same. It is the clearest way to show what generative AI really does.

// by the end of this session

You'll leave with the course consolidated, your questions answered, and a clear, memorable sense of what generative AI is, including the important part: it does not look up an answer, it makes a fresh one every time.

// a look back

A light recap of the whole course

A walk back through the five sessions: staying safe and getting started with Copilot and Word, running and hosting meetings and writing minutes in Teams, presenting your deck and sharing files and Copilot for email, and the Excel basics with Copilot as a patient tutor. The orientation deck you built along the way is yours to keep and reuse.

The thread running through all of it is the same: these tools do good work in less time, and a person still reads, checks and signs off on what they produce.

// the activity

Everyone generates the same picture

At the same time, everyone asks Copilot for a picture using the exact same words. Then we compare screens around the room. No two pictures are the same, and that is the whole point.

Everyone types

Show me a picture of a crocodile driving a Hilux.

// why they differ

Why is everyone's different?

Because generative AI does not look anything up. It learned what things look like from millions of labelled pictures, then paints a fresh one from random speckle every time, guided by your words. Different starting speckle means a different crocodile for everyone. The companion explainer, and a real labelled dataset, show what learning from examples actually means.

How Image Generation WorksOpens in a new window CIFAR-10 labelled image setCat, dog, bird, car, truck · new window

With enough labelled examples, a machine can learn to tell a crocodile from an alligator, the snout, the teeth, the colour, even though nobody wrote those rules down. That is the same idea, running at a huge scale, behind every picture Copilot makes.

// open floor

Open floor

The heart of the session: requests, another look at any tool, and real questions from people's own jobs. All the practice tools are one click away in the catalogue, and they stay online for whenever you want them.

All Simulation ToolsOpens in a new window
// where to get help

Where to get help, and close

The practical safety net: how and where to report an AI or cyber incident, who to ask when you are stuck, and a warm thank you for the work you have put in across the course. The tools stay online for whenever you want them, and the habits you have built, using the approved tool, keeping personal and community information out, and checking what the AI gives you, are the ones that carry into everyday work.

// one last game

Be a champion Prompt Critic

Eight everyday prompts from around the region, each one about to be sent to Copilot with something wrong: missing context, invented facts, a mood best kept private, or personal details that should never go in a prompt. Call the flaw out before the send button gets pressed. First played live in the course; it stays here for any time you want a refresher, and the checklist at the end is worth keeping.

Play: Prompt CriticsOpens in a new window · built for a group, works solo too