The next level

Understanding Prompt Engineering

Alex Banks

Founder & Educator

Infusing persona

Cartoon illustration of person delivering a speech

¹ Image: DALLE_3

Infusing persona

Photograph of Ernest Hemingway

¹ Image: Pixabay

Infusing persona

Photogrpah of Ernest Hemningway

¹ Image: Pixabay

Making it personal

Screenshot of a personalized response from ChatGPT

Useful ways to structure your prompt

Cartoon illustration of a person eating breakfast

¹ Image: DALLE-3

Training techniques

Cartoon illustrations of a person, first as a young boy, then as an adolescent, and finally as a grown man

¹ Image: DALLE-3

Training techniques

Film posters from the films Gladiator, and Sex and the City

¹ Images: Wikipedia

Training techniques

A helping hand for problem-solving

Chain-of-thought (COT) prompting

Zero-shot - No structure provided. ChatGPT solves on its own
One/few-shot - Roadmap of steps ChatGPT can follow to solve problems

Cartoon illustration of a young footballer striking the ball

¹ Image: DALLE-3

Limitations

Biases - The model presents stereotypes or misinformation
Hallucinations - The model confidently states incorrect information
Overfitting - The model is only as good as the data it's trained on

Cartoon illustrations of a child sat at a desk with an award while being watched by other children, a person hallucinating, and a young person with comically large trousers

¹ Images: DALLE-3

Anticipating complex workflows

Cartoon illustration of a child thinking about sustainable technology

¹ Image: DALLE-3

Anticipating complex workflows

Cartoon illustration of a person looking thoughtful

¹ Image: DALLE-3

Let's practice!

Understanding Prompt Engineering