Animated video · Knowledge check
AI Predicts. You Decide.
A 90-second animated explainer on how generative AI writes, the three ways it goes wrong, and the check to run before using its output.
The challenge
Staff treated AI answers as search results. A short, shareable video had to change one mental model, "it predicts, it doesn't look up", fast enough to watch between meetings, and still be accessible without audio.
My approach
- Wrote a six-scene script around one character, Priya, carried over from the Foundations course.
- Built every frame as a code-driven animation, then rendered it to H.264 and WebM, so edits are a script change, not a re-shoot.
- Kinetic typography instead of voice-over: all meaning is on screen, so it works muted, and a full text version sits under the player.
- Showed next-word prediction visually, with likely and unlikely word candidates scored as bars.
Objectives and how each is assessed
| Learning objective | Assessment |
|---|---|
| Explain that generative AI predicts likely words rather than looking up facts. | Questions 1 and 2 |
| Recognise invented facts, bias and out-of-date content in AI output. | Question 3: classify a flawed draft |
| Apply the Verify, Review, Own check before using AI output. | Question 4: put the check in order; pass mark 3 of 4 |
Results and reflection
Add pilot results, reviewer quotes or what you would do differently next time.
Design artefacts
- Download the video (MP4, 1.3 MB)
- Script and storyboard (available on request)