Video Activity: Advanced Prompt Engineering

An Expert Deep Dive with Sander Schulhoff

📹 Video Learning: Prompt Engineering & AI Red Teaming

📋 Before You Watch

List two times you have used prompts to guide an AI tool. What worked? What failed?
Note the difference between helpful experimentation and unsafe behaviour when using AI.
Preview the vocabulary: prompt injection, red teaming, guardrails.

🤔 Pātai While Watching

He aha ngā mōrearea ka puta i te "prompt injection"?
What risks are created when a model follows a malicious prompt?
How does Schulhoff describe the role of red teams in keeping AI systems safe for communities?
Which prompt engineering strategy do you think would be most useful for classroom projects, and why?

💭 After Watching

Think-Pair-Share: Design a three-step protocol for your class that keeps AI use ethical and safe. Spend 2 minutes planning individually, 3 minutes sharing with a partner, then contribute highlights to a class charter.

Critical Thinking Questions

1. Schulhoff describes prompt injection as a major threat. In your own words, what is it, and why does it matter for schools using AI?

2. Summarise how AI red teams operate. How could a student leadership group adapt those ideas to monitor safe AI use at kura?

3. (Extension) This talk was filmed in 2023. Research one new development in prompt engineering or AI safety since then. Evaluate how it could strengthen tikanga-based guidelines for AI in Aotearoa classrooms.