The Most Neglected Part of using AI
When it comes to the art and science of using AI, those teachers who use AI often focus on prompt engineering.
Just in case you’re new around here, McKinsey & Company define prompt engineering as:
The practice of designing inputs for generative AI tools that will produce optimal outputs.
When AI burst onto the scene, it was closely followed by:
Prompting workshops
Prompting frameworks
Collections of prompts for sale
There’s nothing wrong with this. Indeed, I’ve attended the workshops, used the frameworks, and bought the books. The issue is when we focus on prompting to the exclusion of iteration.
In my estimation, far too much time is spent on prompt engineering and far too little emphasis is placed on iteration.
Iteration – for those playing at home – is where you read the output AI gives you, analyse it, make adjustments, make further requests, or add context.
Just as the warmup is only part of the exercise and the first draft is only part of the novel; the first prompt is only part of using AI.
Not only this, but I’d argue that it’s not actually the most important part.
Getting good material from AI is all about iteration.
There’s an old saying that writers use: ‘All good writing is editing’. Writing is simple – you take the ideas from your head and get them out onto a page. Editing is difficult. You analyse, criticise, synthesise, and evaluate.
In the same way, prompt engineering is simple. You write a prompt (or use someone else’s) and fire it into the machine. Iteration is more complex. You have to marshal your higher-order faculties to understand and critique the output, then push the AI in the direction you want.
Focusing on iteration was a game changer for me.
To be honest, it’s what really unlocked the power of AI.
I combed through 150 of my most recent ChatGPT chats; here are the most common iterations I used:
Please rewrite in fewer words.
Please clarify the meaning of _______.
Please rewrite this in less formal language.
You missed one part of my prompt; don’t forget to _____.
Rewrite the output, focus less on ______ and more on ________.
What further questions arise from _______ that I should explore?
How does cultural context influence our perception of __________?
What are the underlying assumptions behind mentioned in your answer?
You have presented one view on ______. What are some other perspectives?
Could you organise your response into clearly labelled sub-headings for me?
The first prompt is important, but iteration is where you will do your heavy lifting.
This week, as you use AI, change your emphasis from the first prompt to the iteration. See how that shapes your expectations and experience with AI.
Recommended Resources:
AI Assessment framework: https://aiadvisoryboards.wordpress.com/2024/01/26/smart-ai-olmc/
AI Assessment scale: https://leonfurze.com/2023/04/29/the-ai-assessment-scale-from-no-ai-to-full-ai/
VINE Generative AI guidelines: https://vine.vic.edu.au/resources/Documents/GAI_Guidelines/VINE%20Generative%20Artificial%20Intelligence%20Guidelines.pdf
Until next week,
Happy Teaching!
Paul Matthews
