AI Applications
Senior 2 · ICT
AI Applications
Lesson 3 of 4
22 min
Type "a goat wearing sunglasses riding a moto in Kigali" into an image generator and, seconds later, you have a picture that no camera ever took. Text-to-image and text-to-video tools such as DALL-E, Adobe Firefly, Stable Diffusion, Midjourney, Veo and Sora are used for posters, adverts, storyboards and school projects. The same technology can also make convincing fakes of real people, so you need to understand both sides.
| Tool type | Input | Output | Typical use |
|---|---|---|---|
| Text-to-image | A written prompt | A new picture | Posters, illustrations, design ideas |
| AI image editing | A photo and an instruction | The edited photo | Removing a background, extending a picture, sharpening a blurred image |
| Text-to-video | A prompt, sometimes with a picture | A short video clip | Storyboards, short animations |
| AI voice | Text or a voice sample | Spoken audio | Narration, translation, dubbing |
These are all generative AI tools: they learn patterns from millions of images, videos or recordings and generate new ones that fit the prompt.
Keza's school club is making a hand-washing poster.
Weak: hand washing
Strong: Flat, colourful cartoon illustration (style) of two secondary school students in school uniform washing their hands with soap at an outdoor school tap (subject and setting), bright daylight, simple background with space at the top for a title (composition), no text or letters in the image (what to avoid).
Why "no text"? Image generators often produce misspelled or meaningless lettering. It is better to add the title yourself in a design or presentation program, where you control the spelling.
Worked example
A class asks a tool for "a picture of an engineer" 10 times and gets 9 men and 1 woman: 9 ÷ 10 = 90% men, a pattern the tool repeated from its training data. The class writes a more specific prompt, "a young Rwandan woman engineer checking solar panels", and chooses images that show the real variety of engineers.
A deepfake is an image, video or audio recording that has been generated or edited with AI, often to show a real person saying or doing something they never said or did. Deepfakes are used for fake news, fraud and bullying. For example, a cloned copy of a relative's voice might "phone" a parent and ask for money to be sent urgently by MoMo.
Signs of a possible deepfake: lips that do not quite match the words; unusual blinking or a frozen face; blurred edges around the face or hair; odd lighting or strange hands; a voice with an unnatural rhythm; and, above all, a source you cannot trace and content designed to shock. When in doubt, check with the person through another channel, for example by calling them back on the number you already have, and look for reports from trusted news outlets.
Tip
Label it, don't fake it. Every AI image you share gets a label, and no real person appears without consent. When a picture or clip makes you feel shocked or angry, stop and check the source before you believe it or share it.
Turn "a market" into a strong image prompt for a Geography poster.
Solution (one good answer): "Realistic photo-style image (style) of a busy fruit and vegetable market in a Rwandan town, with stalls of bananas, avocados and tomatoes and traders serving customers (subject and setting), morning light, wide view from a slightly raised position (composition), no text in the image (what to avoid)."
An AI-generated picture of Rwanda's flag shows a red stripe. What is wrong, and what should you do?
Solution: Rwanda's flag is blue, yellow and green with a golden sun; it has no red. The AI got the details wrong. Do not use the image; use an accurate flag from a reliable source instead.
Mugisha's mother receives a voice note in her brother's voice asking her to send 200,000 Frw urgently to a new number. What should she do?
Solution: Not send anything yet. She should call her brother back on the number she already has for him, or ask a question only he could answer. If it turns out to be fake, she should report the number to her mobile network and the police.
Give two reasons for labelling AI-generated images.
Solution: Any two of: it is honest with viewers; it stops the image being mistaken for a real photograph or for evidence; it follows school and platform rules; it helps slow the spread of fakes.
Summary
Sources and credits
The prompts, worked examples and practice items were written by Ganzaa.
This lesson adapts CC BY-SA material and is shared under CC BY-SA 4.0.
