Talking to Machines
Inside AI's Brain
Tricky Contradictions
Proper Prompting
Who Owns it?
200

A question, instruction, or statement given by a user to guide AI's response.

A prompt

200

This interconnected network makes decisions using weights and hidden layers.

A neural network

200

This is when AI generates conflicting responses to similar input, so both can't be correct.

A contradiction

200

The five key elements of a strong prompt: Task, Context, Format, Reference, and this.

Tone

200

This is the legal right granted to a creator to use, distribute, and modify their original work.

Copyright

400

Generative AI picks its next word based on this measure of how likely an outcome is.

Probability

400

The information — text, images, or code — used to teach an AI model is called this.

Training Data

400

Asking "Why is it good to…" versus "Why is it bad to…" is this prompt-refinement strategy.

Opposite framing

400

This element means telling the AI exactly, and actionably, what you want it to do.

Task

400

This is the act of giving credit to the original creator of a work.

Attribution

600

This technology mimics human intelligence, performing tasks such as understanding language, recognizing patterns, and making decisions.

Artificial intelligence (AI)

600

This term describes what happens when training data contains errors or unbalanced examples — the AI's behavior and biases end up shaped by them.

Biased training data / the data shapes AI behavior

600

Additional information surrounding a prompt that helps AI respond more accurately.

Context

600

Specifying "write it as a haiku" refines which element of a prompt?

Format

600

This is the practice of openly disclosing how something works — especially data use or AI decision-making.

Transparency

800

This term means focusing only on the essential details of a problem and ignoring everything else.

Abstraction

800

This general-purpose form of AI is trained on a very, very large body of knowledge.

A large language model (LLM)

800

In the telehealth activity, students fixed contradictory AI advice using these two strategies.

Fact-checking with reliable sources and refining the prompt

800

Refining a prompt is similar to this software development process — test, observe, update, repeat.

Debugging

800

In 2023, an AI song went viral using the voices of these two artists — without their permission.

(Drake and The Weeknd)

1000

When students added this to their gift-idea prompt — describing who the gift was for — AI's suggestions became far more useful. Name the prompting lesson here.

Be specific / add details to your prompt

1000

When AI finishes a sentence, it doesn't know the right answer — it selects the next word based on patterns in its training data and this concept, which measures how likely each word is to come next.

Probability

1000

Explain why AI models contradict themselves while a medical professional usually wouldn't.

AI learns from broad, generalized datasets that may be conflicting or outdated; professionals have deep, current expertise.

1000

"Write in a professional tone, as if you are a college admissions counselor" refines two elements at once. Name both.

Tone and context

1000

Distinguish between copyright and attribution: which one gives the creator control, and which one just gives credit?

Copyright = control; attribution = credit/recognition