Week 2 · Day 4 — The Hand
You can't prompt your way out of a model's defaults. You point it at real people instead.

TL;DR
- Day 4 of the #HOWNOTTOUSEAI challenge involves building something and sending it to a live recipient.
- The session aims to compare the output of an AI tool with an original question, revealing why surveys and prompts fail in similar ways.
- It identifies five layers where decisions are made, noting that interviews and analytics capture these decisions at different stages.
- A live browser coach helps identify closing words in one's own questions.
- Overcoming AI model bias requires directing the AI towards real human expressions, not just better prompts.
- Sources of real human expression include the internet, support tickets, call recordings, reviews, and emails.